# Your Agent Doesn't Need an LLM for Every Decision - Raymond Weitekamp, OpenProse Raymond Weitekamp, OpenProse [Talk page](https://ai.engineer/talks/hFNbPmzzcEs-your-agent-doesnt-need-llm-every-decision) · [Video](https://www.youtube.com/watch?v=hFNbPmzzcEs) ## Your Agent Doesn't Need an LLM for Every Decision Raymond Weitekamp’s Recursive Decision Models put a classifier at the root of an agent. The same idea that helps a game character find a broken part lets a voice interface learn when it can switch workspaces before transcription finishes. ### What does the model actually have to produce? [0:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=0s) Give a coding agent enough tokens and it can keep producing possible changes. Someone still has to decide which change goes into production. Raymond Weitekamp of OpenProse asks how much of an agent’s work has that second shape. If the options already exist, what should a model return? A **decision model** ranks or selects among the options it receives. Its output is a judgment. Application code decides what to do with it. The term *System One* borrows the distinction in *Thinking, Fast and Slow* between quick, intuitive judgments and slower, deliberate problem solving. *Zero-shot classifier* is another name used here. The interesting engineering question is whether a series of fast judgments can accomplish a task that seems to require deliberate reasoning. ### Give the classifier the first call [6:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=385s) A **recursive language model** (RLM) gives a root model the ability to call a child RLM, which can in turn call another. The program caps the depth. In Weitekamp’s example, a branch that reaches depth two ends with an ordinary LLM call. A **recursive decision model** (RDM) applies the same rule to selection. One of the choices at a node can be to ask a narrower question. At maximum depth, a single decision call settles it. [Recursive Decision Models](https://recursivedecisionmodels.com) The position of the first call changes which work the application pays for. - **Decision as a tool.** A coding agent can use a classifier when compacting context, checking whether a proposed tool call may execute, or routing a request to a model. The surrounding agent is already running when it asks for the judgment. - **Decision at the root.** The application starts with the classifier. Calling a language model or delegating to an agent is one of its choices. A branch that can finish with decisions alone can go all the way to an action without generating text. ### Doom needs two views of the same level [9:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=588s) What can a Doom agent learn from the map that it can’t see from the player’s perspective? One group of decisions handles immediate choices such as holding position, walking, or dodging. The program combines the answers into the next player action and repeats as the game changes. A second group uses a bird’s-eye view of the map to choose where to explore. Coupling the two gives the player somewhere to go and a way to react to what happens on the way there. Both jobs use decision calls. Doom illustrates how to compose decisions. Weitekamp adds a further capability for recursion and uses three rules to guide the design. - **Keep context outside the model.** Code holds the full game state and cuts a different slice for each question. The movement controller gets its local view. The exploration controller gets its map view. - **Let a decision request another decision.** Descending into a subproblem is itself a choice the model can make. The program enforces the depth limit. - **Make generation optional.** First get a task to work with decision calls alone. The same architecture should also be able to call an LLM or agent when a task needs one. ### What if the pins on a Galton board could decide? [15:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=907s) On an idealized Galton board, a falling ball has a 50–50 chance of going left or right at each pin. After many drops, the piles form a bell shape. Now give each ball a number and let the pins read it. The first pin sends lower numbers one way and higher numbers the other. The next pin divides its own range again. Successive choices sort the balls into numbered buckets. What happens if a pin can answer two questions at once? The three-dimensional board sorts by color along one axis and number along the other. A cross-section along each axis lets you check each judgment separately. With perfect classification, a ball lands in the bucket for both its number and its color. One classifier call could assign that pair directly. The board earns its keep as a picture of how an early choice restricts the choices farther down. ### Search the children of a category or crop an image [17:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1052s) A product catalog gives the hierarchy a practical job. A search example uses 12,000 Shopify product categories and a query such as *laptop backpack*. Presenting every category at once risks overloading the context window. Give the classifier the query and the children of the current parent category, and its ranking can tell the program which part of the catalog to open for the next call. The original example used beam search, keeping several promising paths alive. Code can also make the search greedy and follow only the highest-ranked child. That saves exploration at the price of committing to an early classification. Keeping the top three, one alternative Weitekamp gives, allows a lower-ranked branch to reach a more plausible category farther down. The classifier supplies the rankings. The programmer chooses how many paths to pursue. For an image, which pixels should the next call see? In a beetle search, a positive judgment prompts the program to zoom in and ask again about a smaller box. Another implementation recursively halves the image and classifies the resulting blocks. In the first, a hit directs the crop. In the second, the subdivision rule determines which regions the classifier can inspect. A capable vision model might answer these particular questions in one pass. Recursion gives the program control over the area each judgment concerns. ### Find the faulty inserter in Factorio [21:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1277s) Which part of a bank of similar production lines needs repair? In Weitekamp’s Factorio factory, the RDM picks the group with a problem, narrows it to one line, then identifies an inserter on that line for replacement. The character runs over to do the job and moves on to the next one. Each selection changes what the next call has to inspect. By the time the character moves, it has a particular part to attend to, with no generative model in the loop. The regular layout helps. The factory is effectively a matrix of repeated components, so each question has a manageable set of answers. The character also runs the long way around to the next repair. Weitekamp considers adding a pathfinding component. The diagnosis gave the character a destination and left the quality of its route to another part of the program. Where do the options come from? A script, the application, or another agent must enumerate what the classifier can choose. Games can expose their state and available moves programmatically. The talk’s slide website offers a smaller example. Its document object model (DOM) exposes three interactions on the page in question. You can go up, go down, or start a video. A desktop has many more visible controls and keyboard shortcuts, but wherever code can enumerate them, selection becomes a plausible way to operate the interface. ### Let a language model edit the smart home’s map [26:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1572s) An outline walk can descend from a document title to a heading, subsection, paragraph, sentence, and word. The document supplies the possible paths. What changes when a model can introduce a new option? In the call trees from Weitekamp’s smart-home demo, squares mark generative calls. A request can narrow down to an existing action or reach a generator that can change the choices available. The branches take different shapes depending on what the request needs. Weitekamp expanded an earlier smart-home demo with more rooms and appliances. A bedtime request implies several changes, including locking the door, leaving a lamp on, and switching other devices off. A request to watch a film is less explicit still. The house needs a way to associate the user’s phrase with a combination of appliance commands. The shared interface is a **map**. An entry has a name, a short description, and a pointer to behavior. `TV on` can point to a function that sets a Boolean. Another entry can point to a whole loop. The decision model judges which named entry fits the request. Code invokes the function behind that entry, so a classifier can select a complicated operation as readily as it selects a switch. The LLM gets a structured way to add or edit entries. If the user’s movie routine means turning the TV on and the lights off, the LLM can record that combination as *movie time*. On a subsequent request, the decision model has a single option pointing to both actions. The learned preference lives in the map. The RDM implementation can also represent entries in several dimensions, including low-dimensional embeddings whose dimensions carry semantic meaning. ### Run judgments in parallel and arbitrate the write [32:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1928s) Fast, inexpensive decision calls make speculation affordable. Several models can read the same state and suggest actions concurrently. Let each one act as soon as it finishes, though, and completion order starts deciding what happens. A fast model can change the state while a slower model is still evaluating the old state. Its answer may call for the opposite change. Which proposal should be able to change the state? For speech arriving in real time, different channels can already favor an action before the sentence finishes. An **arbiter** considers the evidence available across channels at that moment and decides whether it has enough confidence to let an action through. Weitekamp compares this to concurrent reads with a lock on writes. The arbiter adds a judgment about whether a proposed write is justified at all. The first RDM implementation was a module in a programming framework that already supported System One API calls. Its conventions gave Weitekamp a way to check that the modules composed as intended. He also built a standalone Rust package for applications that need a fast, portable implementation. [RDM repository](https://github.com/rawwerks/rdm) ### A laptop you could play by hitting it [36:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2190s) *Operating System One* is a play on learning to operate System One models, with a physical goal behind it. Mousing left Weitekamp with shoulder tension. Graduate-school deadlines brought repetitive strain and a collection of vertical mice, trackballs, and ergonomic keyboards. He wanted to keep doing computer work while spending less of the day holding the same posture. Voice seemed within reach. An earlier experiment turned the laptop itself into a drum kit. The machine had a built-in accelerometer. Hit one of four locations on its case and a classifier distinguished the impact well enough to trigger the corresponding drum. Tilting the laptop controlled an effect. In another mapping, each hit triggered a side-chain compressor, briefly ducking the sound. The same hardware could take a discrete input like a strike or a continuous input like tilt. Weitekamp learned programming and machine learning through the Princeton Laptop Orchestra. Some pieces made an instrument of the trackpad or keyboard. Others synchronized clocks over Wi-Fi so players’ machines could exchange precisely timed messages. Small models such as k-nearest neighbors and single-layer neural networks could learn mappings on the fly, classify gestures, or interpolate between controls. A performer could hear what latency cost. The sound had to arrive in time with the gesture. ### Earshot learns when it can act before transcription finishes [43:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2631s) Earshot gives a voice interface a task you can watch and hear. Spoken numbers select desktop workspaces one through eight. Initially, the system transcribes the audio, identifies the number, and switches to the corresponding workspace. You hear the count and then see the desktop catch up. Could the system recognize enough of a familiar sound to switch sooner? This experiment was a precursor to [oSo](https://oSo.computer). How can the same spoken number operate the desktop and teach a faster response? Two paths listen at once. While the transcription path supplies an action and a teaching signal, a much smaller supervised model takes coarse frequency bins from short slices of the same audio. Weitekamp gives 40, 80, and 100 milliseconds as example slice lengths. The small model starts without a learned association between those features and a workspace. Another layer judges when its prediction is trustworthy enough to follow. That lets training begin well before the learner gets to move the desktop. The counting in *Einstein on the Beach* makes the handoff perceptible. One through four recur more often, so the learner gets more examples of those sounds. Their workspace switches start keeping up with the count while later numbers still lag. By the end of the demonstration Weitekamp reports about 45 milliseconds for the learned response, with an occasional blip. That timing belongs to the repeated-number task, where the fast path only has to distinguish a handful of familiar sounds. How fast can a recognizer respond when the input is a drawing? Hunch plays Pictionary by classifying a user’s doodles locally with an embedding model running through WebGPU. Weitekamp points to a classification time of roughly 150 milliseconds in that demo. Here the input is a sketch and the immediate output is a guess about what the user is drawing. How can one count teach a shortcut and still operate the desktop? The transcription path supplies supervision while serving the initial response. Once the system trusts the short-window classifier for a sound, its prediction can trigger the switch before transcription finishes. - Spoken number — A number in the audio corresponds to one of eight workspaces. - Transcription path — Recognizes the word and identifies the target action. - Short-window classifier — Learns to associate coarse frequency bins with the target. - Trust decision — Determines whether to follow the small model’s prediction. - Switch workspace — The desktop moves to the workspace for the recognized number. - Spoken number → Transcription path (Audio to transcribe) - Spoken number → Short-window classifier (Frequency bins from a short slice) - Transcription path → Short-window classifier (Supervision) - Transcription path → Switch workspace (Initial response from the transcribed word) - Short-window classifier → Trust decision (Predicted action) - Trust decision → Switch workspace (Use the prediction when trusted) ### oSo starts with voice and delegates the open-ended work [48:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2899s) **oSo**, short for Operating System One, was a preview at the time of the talk. Its starting question is what it would take to work on a computer without touching it, and eventually without needing to see it. Voice is the first input. In the proposed design, an RDM can handle a familiar command directly or hand a job to a coding agent or recursive language model when the job needs generation. There are two things for such a system to learn. One is what a particular user wants, as with the household’s movie routine. The other is which model can reliably recognize that intent. oSo aims to move a recurring decision to the smallest, fastest available model once its past performance gives the arbiter reason to trust it. Earshot supplies the working example of that second kind of learning. The desktop starts responding sooner to the numbers it has heard most often. Weitekamp reads engineers’ return to the terminal as a willingness to leave behind some of the constraints of conventional applications. For someone whose shoulder hurts after a day of mousing, typing commands still leaves a physical problem to solve. He invited viewers to join the oSo waitlist for an experiment with a different kind of control. A word could operate the computer as directly as a tap on a laptop once played a drum. [oSo project](https://oSo.computer) ## Key takeaways - At each RDM node, code supplies the relevant state and available choices. A choice can request another judgment, select an action, or delegate work that needs generation. - Training a faster model and letting it act are separate decisions. Earshot can collect examples for an unfamiliar number while already using the faster path for a familiar one. ## Slides ### Recursive Decision Models, or, Operating System One Raymond Weitekamp originally submitted the talk as Recursive Decision Models. He explains that the subtitle, Operating System One, better captures what he wants to discuss. [0:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=0s) [The title “Recursive Decision Models” appears above “or, Operating System One” and the speaker credit for Raymond Weitekamp.](https://assets.ai.engineer/sites/aiecode2025/sha256/d7fbddc36cb8a27c06722a629659540394712fd59f14e055f111b6bd0a648889.webp) Recursive Decision Models or, Operating System One Raymond Weitekamp RAW.works | OpenProse @raw_works recursivedecisionmodels.com oSo.computer ### What if “reasoning” was actually many small decisions? Weitekamp asks whether an AI system could solve a task that seems to require deliberate thought by making many small, fast decisions. [2:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=156s) [Large centered text asks, “What if ‘reasoning’ was actually many small decisions?”](https://assets.ai.engineer/sites/aiecode2025/sha256/1745af66d3ffc16c4ae2aec52139f50fc203de8ea68a7096da7d22f39f0e4880.webp) What if “reasoning” was actually many small decisions? ### RDM / LLM as a decision A recursive decision model sits at the root of both trees. In the lower tree, branch A can call a language model at A2. The other three leaves stay the same, so the LLM call is one choice within the tree. [9:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=575s) [A root RDM branches into two sub-RDMs. At depth two, branch A contains Decision A1 and LLM A2. Branch B contains Decisions B1 and B2.](https://assets.ai.engineer/sites/aiecode2025/sha256/d94a5dc9124e72460341ea6614e6f16111e09335a88e0152369427dfc85bfbfa.webp) RDM / LLM as a decision Root RDM (depth=0) ├── Sub-RDM A (depth=1) │ ├── Decision A1 (depth=2) │ └── LLM A2 (depth=2) └── Sub-RDM B (depth=1) ├── Decision B1 (depth=2) └── Decision B2 (depth=2) Root RDM (depth=0) ├── Sub-RDM A (depth=1) │ ├── Decision A1 (depth=2) │ └── Decision A2 (depth=2) └── Sub-RDM B (depth=1) ├── Decision B1 (depth=2) └── Decision B2 (depth=2) [illegible]h=1) [illegible]th=2) [illegible]th=2) [illegible]h=1) [illegible]th=2) [illegible]th=2) [illegible] tool [illegible]1) [illegible]) [illegible]epth=2) [illegible]2) ### Choosing where to explore in Doom A second set of decisions uses the map to choose where Jev should explore next. With scouting as the goal, the first-person controls handle the next moment: hold fire, carry on and walk. [10:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=637s) [An outlined room and corridor appear beneath the game view. A green player arrow and a purple marker are visible on the map as the caption introduces the exploration decision.](https://assets.ai.engineer/sites/aiecode2025/sha256/341f84109e41262623e6fc3d23fc1a5fc5dc17fc57c00817d9b2f72c88926654.webp) DOOM // OPS Add a second composition that decides where to explore next, GOAL: SCOUTING FOR WHAT'S OUT OF SIGHT JUDGMENTS FIRING Should the player's trigger be held down right now? fire hold_fire conf 0.97 DODGE The player's current top priority is scouting for what's out of sight. What does this exact moment call for? carry_on dodge_left dodge_rig[illegible] dodge_for[illegible] dodge_bac[illegible] conf 1.00 MOVEMENT The player's current top priority is scouting for what's out of sight. Given the current situation, how should the player move right now? walk +0° walk [illegible] walk [illegible] walk [illegible] walk [illegible] walk [illegible] walk [illegible] Jev plays Doom · Diogo Almeida, TypeSafe · Sep 15, 2026 · cut from 0:53–1:02 and 1:31–1:56 Interface text: ← SESSIONS HELM Jev RESTART PAUSE MUTE LINK LIVE last [illegible] / avg [illegible] | 39 calls | [illegible] tok ORDERS > standing order… SITUATION REPORT MAP GRAPH EVALS 50 AMMO 100% HEALTH 2 3 4 / 5 6 7 ARMS 0% ARMOR BULL 50 200 SHEL 0 50 RCKT 0 50 CELL 0 300 helm: research/[illegible]_snowy_elephant [illegible] [illegible] $0.0010 TIME 0 0:15 / 0:30 ### Advancing through the Doom level One Jev agent uses the map to choose where to explore. A second uses the first-person view to decide what to do moment by moment. As the purple path leads into a corridor, its firing and dodging choices are hold_fire and carry_on. [11:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=676s) [The game view shows a corridor. Below it, a purple trace loops through a mapped room and extends toward colored markers in a passage.](https://assets.ai.engineer/sites/aiecode2025/sha256/1f04c565639db3d8df943cbee8044b0409787f165ffd7dea4a7e9fb933f12239.webp) DOOM // OPS GOAL: ADVANCING THROUGH THE LEVEL ORDERS > standing order… JUDGMENTS FIRING Should the player's trigger be held down right now? fire hold_fire conf 0.98 DODGE The player's current top priority is advancing through the level. What does this exact moment call for? carry_on dodge_left dodge_[illegible] dodge_[illegible] dodge_[illegible] conf 1.00 MOVEMENT The player's current top priority is advancing through the level. Given the current situation, how should the player move right now? hold_grou… walk to t… conf 0.96 TURN Where should the player point right now? Looking and aiming are one act: the gun goes where the eyes go. hold the… look [illegible] glance left SITUATION REPORT MAP GRAPH RULES Room 1 Room 2 Room [illegible] lies to the north — seen from afar but never yet entered; about 12 paces away The unmapped edge on the southwest side of Room 1, where about eight paces of floor continue into unmapped territory; about three paces away [illegible] Jev plays Doom · Diogo Almeida, TypeSafe · Sep 15, 2026 · cut from 0:53–1:02 and 1:31–1:56 Interface text: ← SESSIONS HELM Jev RESTART PAUSE MUTE LINK LIVE last [illegible]ms / avg [illegible]ms · [illegible] calls · [illegible] tok 50 AMMO 100% HEALTH ARMS 0% ARMOR BULL 50 200 SHEL 0 50 RCKT 0 50 CELL 0 300 [illegible] · research/[illegible] · ctx [illegible] [illegible] COST $0.0079 TIER 0 0:21 / 0:30 ### Three rules Code gives each decision the slice of external context it needs. A decision can call another, up to a maximum depth. Recursive decision models must be able to run without a generative model and still work with LLMs and agents. [12:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=774s) [Code slices context kept outside the model. Decisions can call other decisions up to a maximum depth. The system can run without a generative model and can compose with LLMs and agents.](https://assets.ai.engineer/sites/aiecode2025/sha256/7325e25356213078cb90540f654cf2e4ad25d24a7d5b08a1950c63802d10f874.webp) Three rules 1 The context stays outside the model, and code cuts it into slices. 2 A decision can choose to call another decision, up to a maximum depth. 3 It works with no generative model at all, and it composes with LLMs and agents. ### Decisions all the way down Weitekamp introduces a Galton board as his first example of a recursive decision model. A ball’s path down the board illustrates a sequence of decisions. [15:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=904s) [Large light text reading “Decisions all the way down” is centered on a dark background, with the presenter in a lower-right inset.](https://assets.ai.engineer/sites/aiecode2025/sha256/ff65ed4378c21a490621244e7c7af568403f3e8eddf697c0d3c7fb6643265d19.webp) Decisions all the way down ### Same number. Any color. Alternating layers of pins steer each falling ball by number along X and by color along Y. The two cross-sections show the same balls: each numbered bucket holds a mix of colors, while the color view groups them by hue. [16:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1003s) [Colored, numbered balls descend through a three-dimensional pin board. Two cross-sections show how the balls sort by number and by color, with bars below counting landed balls.](https://assets.ai.engineer/sites/aiecode2025/sha256/7d9974edc9eca79521e9cccf7549fb05b70f07a6c96c4e6b30d5e194cbef72ea.webp) Same number. Any color. 01 / SPATIAL VIEW The sorting volume 10 × 10 buckets Z ↓ Y / COLOR X / NUMBER X-axis ticks: 1 2 3 4 5 6 7 8 9 10 16 LAYERS · ALTERNATING X AND Y PINS 02 / X–Z SECTION Sort by number Z ↓ X-axis ticks: 1 2 3 4 5 6 7 8 9 10 LANDED BALLS Bar labels: 2 4 5 7 8 6 4 8 5 5 X / NUMBER X 1 → 10 Y collapsed 03 / Y–Z SECTION Sort by color Z ↓ LANDED BALLS Visible bar labels: 8 4 8 3 4 6 7 8 Y / COLOR Y violet → red Interface text: Pause 100 balls Reset Smart pins Random pins Speed 1× Number front Corner Color front Left Right Look down Look up Reset Custom view 54 / 100 balls Number 100% Color 100[illegible] ### Hierarchical search for a laptop backpack At each fork, the search replay follows every branch scoring at least half as high as the leader. Laptop Bags scores .55, so Backpacks at .45 also qualifies. That second branch leads to the selected category, Laptop Backpacks (.33). [17:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1052s) [For a 20-litre backpack with a padded sleeve, the tree follows Laptop Bags at .55 and Backpacks at .45. Under Backpacks, a blue ring marks Laptop Backpacks at .33.](https://assets.ai.engineer/sites/aiecode2025/sha256/de2e35d77c2dacc13f883ae365dade4c3bcfb5ff4475628e443bc1e4194842da.webp) Hierarchical search, without the beam Hierarchical search Twelve thousand categories, and no request shows more than one level of them. Each answer decides which categories to ask about next: every one at least half as likely as the likeliest. The same walk as TypeSafe's hierarchical classification cookbook, without the beam. recursivedecisionmodels.com/hierarchy ↗ A product listing Laptop backpack with a padded sleeve, 20 litres. The search 3 levels asked about, up to 4 followed at once Laya's answers Legend: followed · rated, not followed All products Luggage & Bags 1.00 Laptop Bags .55 Backpacks .45 Laptop Backpacks .33 Hiking Backpacks .25 Military Backpacks .22 School Backpacks .19 Luggage Accessories .00 Garment Bags .00 Apparel & Accessories .00 Office Supplies .00 ✓ Luggage & Bags › Backpacks › Laptop Backpacks A replay, not a live run. The laptop backpack, board game and hiking backpack are Laya's recorded answers (a 450 MB decision model, in the browser). The cat perch and headphones are an oracle replay: the right path, with odds set by [illegible] Interface text: Laptop backpack Board game Hiking backpack Cat window perch Headphones ### Hierarchical search for a family board game Laya’s board-game replay reaches Board Games and also explores Toys. At each step it follows every category at least half as likely as the leader. Across seven levels, it follows up to six branches at once. [17:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1055s) [A family strategy game listing appears above a tree that fans into many toy and game categories. The search summary reports seven levels examined and up to six branches followed at once.](https://assets.ai.engineer/sites/aiecode2025/sha256/8ec5b2ff4b2d10479966a233387630843cc5ec92dc3833f4e25321e0735b5875.webp) Hierarchical search, without the beam Hierarchical search Twelve thousand categories, and no request shows more than one level of them. Each answer decides which categories to ask about next: every one at least half as likely as the likeliest. The same walk as TypeSafe's hierarchical classification cookbook, without the beam. recursivedecisionmodels.com/hierarchy ↗ A product listing A board game for families: a strategy game for 2 to 4 players. The search 7 levels asked about, up to 6 followed at once Laya's answers Legend: followed · rated, not followed All products Toys & Games 1.00 Games [illegible] Board Games 1.00 Battle Top Accessories .00 Battle Tops .00 Toys .22 Remote Control Toy Access... .18 Remote Control Toys .16 Remote Control Boats & Wa... .18 Remote Control Helicopters .18 Remote Control Airships & ... .17 Remote Control Tanks .14 Remote Control Cars & Tru... .12 Remote Control Motorcycles .09 Remote Control Planes .06 Remote Control Robots .05 Play Vehicles .16 Toy Airplanes .20 Toy Cars .19 Toy Trucks & Construction... .15 Toy Boats .11 Toy Motorcycles .09 Toy Race Car & Track Sets .09 Pretend Play .15 Play Money & Banking .44 Toy Tools .13 Pretend Electronics .13 Play Vehicle Accessories .14 Musical Toys .12 Flying Toys .06 Riding Toy Accessories .01 Game Timers .20 Puzzles .11 Outdoor Play Equipment .00 Mature .00 Electronics .00 A replay, not a live run. The laptop backpack, board game and hiking backpack are Laya's recorded answers (a 450 MB decision model, in the browser). The cat perch and headphones are an oracle replay: the right path, with odds set [illegible] Interface text: Laptop backpack Board game Hiking backpack Cat window perch Headphones ### Hierarchical search for a hiking backpack Laya’s recorded search follows every category at least half as likely as the leader at each level. Under Backpacks, all four branches qualify. Hiking Backpacks wins narrowly at .28, just ahead of Military Backpacks at .27. [17:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1066s) [A 40-litre backpack with a rain cover follows Luggage & Bags to Backpacks. Hiking Backpacks is ringed at .28, just above Military Backpacks at .27.](https://assets.ai.engineer/sites/aiecode2025/sha256/5578fce00c08fe822d93abd114aac52581a99f3cf011d81557f15ee8f781232c.webp) Hierarchical search, without the beam Hierarchical search Twelve thousand categories, and no request shows more than one level of them. Each answer decides which categories to ask about next: every one at least half as likely as the likeliest. The same walk as TypeSafe's hierarchical classification cookbook, without the beam. recursivedecisionmodels.com/hierarchy ↗ A product listing Hiking backpack, 40 litres, with a rain cover. The search 3 levels asked about, up to 4 followed at once Laya’s answers ● followed ○ rated, not followed All products Luggage & Bags 1.00 Backpacks 1.00 Hiking Backpacks .28 Military Backpacks .27 Laptop Backpacks .25 School Backpacks .20 Garment Bags .00 Briefcases .00 Hardware .00 Office Supplies .00 ✓ Luggage & Bags › Backpacks › Hiking Backpacks A replay, not a live run. The laptop backpack, board game and hiking backpack are Laya’s recorded answers (a 450 MB decision model, in the browser). The cat perch and headphones are an oracle replay: the right path, with odds set by [illegible] Interface text: Laptop backpack Board game Hiking backpack Cat window perch Headphones ### SAM 3 masks six sheep SAM 3 marks each of six sheep with a different color. Nathan’s Jev demo asks which sheep might “yee haw.” In the beetle test that inspired it, Jev rechecked each box, discarded false hits and rejected one correct detection too. [19:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1188s) [Six sheep in a pasture have different colored masks. The adjacent post compares two annotated views of an insect display and describes feedback on beetle detections.](https://assets.ai.engineer/sites/aiecode2025/sha256/882acdc1d92b07273bbca2f2fce09a1cc11e516ca203f1f58dc44f47cbe6f226.webp) SAM 3: every sheep Nathan LeClaire @dotpem · Oct 18, 2026 Here's an example of stacking jev on jev inspired by Lily. Jev divides and conquers an image to find a bunch of bounding boxes. For straightforward detection it's worse/slower than SAM3 but for something intelligent - like finding all sheep who might "yee haw" - it does well. Jev / Vision Preview SAM 3 6 masks SAM 3 one score per mask Q sheep most likely to yee haw SAM 3 run on the same photo Image support is a preview lillyguisnet @lillyguisnet · Oct 16, 2026 Replying to @lotusfracture First image, increase number of image cuttings allowed up to 15 times. Second, same but also included feedback. Each produced box is checked and asked if correctly on a beetle. If no, repeat cycle, if yes keep the box. Greyed squares, Jev said no beetle. Other squares, Jev said 1 beetle then asked to produce the box. Green dot, ground truth. Pink box, Jev box. Yellow dot, Jev box centroid. Blue dotted box, independent cuts for another test (can ignore here). Overall, the feedback loop helps throw away bad boxes but doesn't help finding more. 1 correct beetle was rejected too. Nice trick! :) Interface text: View on X ↗ View on X ↗ ### Counting sheep before splitting the image Jev counts six sheep. In Nathan LeClaire’s demo, any count of two or more tells the code to split the image. The vertical line through the pasture marks the first cut in a search that repeats on smaller regions. [20:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1211s) [A vertical line bisects the pasture photo beneath a count of six. The chart gives six sheep a score of 0.78 and seven a score of 0.19. A rule calls for a split when the count is at least two.](https://assets.ai.engineer/sites/aiecode2025/sha256/1737a615a424ea564b0d84bbaaed6629f570867ce572af7dd71dbe0e2bcc5222.webp) So ask how many Jev / Vision Preview Photo 1729 × 910 px 6 Jev Call 1 answered in .39 s Q How many sheep are in the picture? Count a sheep that is cut off by the edge of the picture. X-axis ticks: 0, 1, 2, 3, 4, 5, 6, 7, 8+ Bar labels: 6 (.78), 7 (.19) MANY CODE count ≥ 2 → cut in half [illegible] JEV CALLS 01 Nathan LeClaire @dotpem · Oct 18, 2026 Here's an example of stacking jev on jev inspired by Lily. Jev divides and conquers an image to find a bunch of bounding boxes. For straightforward detection it's worse/slower than SAM3 but for something intelligent - like finding all sheep who might "yee haw" - it does well. lillyguisnet @lillyguisnet · Oct 16, 2026 Replying to @lotusfracture First image, increase number of image cuttings allowed up to 15 times. Second, same but also included feedback. Each produced box is checked and asked if correctly on a beetle. If no, repeat cycle, if yes keep the box. Greyed squares, Jev said no beetle. Other squares, Jev said 1 beetle then asked to produce the box. Green dot, ground truth. Pink box, Jev box. Yellow dot, Jev box centroid. Blue dotted box, independent cuts for another test (can ignore here). Overall, the feedback loop helps throw away bad boxes but doesn't help finding more. 1 correct beetle was rejected too. Nice trick! :) Interface text: 0:03 / 0:35 View on X ↗ View on X ↗ ### Cut, drop, keep Jev checks each patch of pasture. If it sees multiple sheep, it splits the patch and checks each half. If it sees none, it drops the patch. If it sees one, it keeps a pink box. The demo kept 8 boxes in 23 calls. [20:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1250s) [White lines divide the pasture into regions, crosses mark discarded areas, and pink boxes outline detections. The final tally shows 11 split decisions, two drops and eight retained boxes.](https://assets.ai.engineer/sites/aiecode2025/sha256/cda8238d7493265935ad3fdb0dddd33697b138c4793472e3382551ded4a55758.webp) Cut, drop, keep Jev / Vision Preview Photo 1729 × 910 px Jev Calls 4–23 round 8 of 8 2 calls at once MANY → cut in half 11 NONE → drop it 2 ONE → keep its box 8 BOXES 8 JEV CALLS 23 A real photo, real recorded calls Image support is a preview Nathan LeClaire @dotpem · Oct 18, 2026 Here's an example of stacking jev on jev inspired by Lily. Jev divides and conquers an image to find a bunch of bounding boxes. For straightforward detection it's worse/slower than SAM3 but for something intelligent - like finding all sheep who might "yee haw" - it does well. lillyguisnet @lillyguisnet · Oct 16, 2026 Replying to @lotusfracture First image, increase number of image cuttings allowed up to 15 times. Second, same but also included feedback. Each produced box is checked and asked if correctly on a beetle. If no, repeat cycle, if yes keep the box. Greyed squares, Jev said no beetle. Other squares, Jev said 1 beetle then asked to produce the box. Green dot, ground truth. Pink box, Jev box. Yellow dot, Jev box centroid. Blue dotted box, independent cuts for another test (can ignore here). Overall, the feedback loop helps throw away bad boxes but doesn't help finding more. 1 correct beetle was rejected too. Nice trick! :) Interface text: X View on X ↗ X View on X ↗ ### It must compose with LLMs At each orange square, the LLM generates options. The decision model keeps one or two, highlighted in blue. For each option it keeps, the LLM generates a smaller set of possibilities. [26:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1568s) [Orange squares fan out into possible paths. One or two highlighted paths reach each blue decision node, then branch again at the next orange square.](https://assets.ai.engineer/sites/aiecode2025/sha256/e5da6aab9ae142b987c2d14d844ee64f5a4e999958b8906cef4a13e183d7b5d2.webp) It must compose with LLMs LLM: diverges decision model: converges decis[illegible] An LLM fans out options, a decision model keeps one or two, and each one kept opens a smaller fan. ### Generation x Decision Demonstrate Emergent Graphs The decision model follows a tree laid out by code. In the smart home demo, LLM commands and the house’s replies shape the tree as it grows. One run kept retrying a command until it hit the limit at 80 requests and 35 LLM calls. [26:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1578s) [Two predefined search trees appear on the left. Four smart-home examples on the right show how different requests produce different trees, with blue decision requests and orange LLM-written commands.](https://assets.ai.engineer/sites/aiecode2025/sha256/aff0b080f369ca3ae3801b169a9fc4c9c5dbc18ac173cf7bcabddda843cb19a8.webp) Generation x Decision Demonstrate Emergent Graphs Decisions only Code lays the whole tree out before the run. The decision model is only asked its way down it. ● one request to the decision model · laid out by code, never asked Halving search 15 requests · 117 pieces of the handbook laid out Outline walk 15 requests · 37 headings and passages laid out A decision model and an LLM Nothing is laid out in advance. Each tree is what one run made, from what the LLM wrote and how the house answered. ● asked about the request, or the house ■ asked about a command the LLM wrote ○ not asked: a limit “Lock the front door.” 7 requests · 1 LLM call The door is open, so the lock refuses. “I am leaving for the weekend.” 15 requests · 2 LLM calls First taken for small talk. “Make it three degrees cooler.” 11 requests · 3 LLM calls Ends one degree too far. A three-part request, one part of which the house cannot do 80 requests · 35 LLM calls The same command is written and tried again at every level, until the request limit. Blue: a request to the decision model. Orange: something the LLM wrote. Left, decisions only; right, with an LLM. After docs.typesafe.ai/demos/smart-home. ### Smart home bedtime routine For “We are off to bed.” the decision model picks the “good night” routine, then selects “finish” after the house reports what changed. The run takes 0.5 s and leaves the entry locked, the oven off and the bedroom lamp on. [29:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1748s) [The request reads “We are off to bed.” The floor plan shows the entry locked and the bedroom lamp on. The run trace lists “good night” followed by “finish.”](https://assets.ai.engineer/sites/aiecode2025/sha256/fc0cc21880749317506462b71253fac7043abd95960a649de691237f76834bb9.webp) Smart home A decision model runs the house. It writes nothing, and it is never shown the house: it is told what the last step did, and points at one name on the map, a room and then a thing in it, until it points at finish. Based on TypeSafe's smart home assistant demo. Tell the house what to do We are off to bed. Ask done in 0.5 s The house stays as each request leaves it Reset off 28° inside 22° target − + 28° outside − + 10° 30° BEDROOM lights off lamp on window open BATHROOM lights off fan off LIVING ROOM lights off tv off fan off window closed KITCHEN lights off oven off coffee maker off dishwasher off ENTRY lights off lock locked alarm off door closed GARAGE lights off washer off dryer off doorway clear door closed entry lock: now locked The map what a request can point at ROOMS each a map of its own entry lights, door, lock, alarm. 0.00 living room lights, tv, fan, window. 0.00 kitchen lights, oven, coffee maker, dishw... 0.00 bedroom lights, lamp, window. 0.27 bathroom lights, fan. 0.00 garage lights, door, washer, dryer. 0.00 thermostat mode, target. 0.02 ROUTINES several steps under one name, hover to read them good night ×1 Going to bed: the rest of... 0.10 good morning Getting up: the front doo... 0.00 movie time Watching a film: the living r... 0.00 leaving Going out and leaving the hous... 0.01 finish ×1 Stop: nothing more should be ... 0.60 Run 2 Code requests 2 / 256 depth 0 / 3 stopped: a decision good night .91 — finish .60 1 good night 0.91 → good night: done (living room tv: alr... 2 finish 0.60 asked by good night The model was shown 490 characters of the 583 that exist task We are off to bed. observation good night: done (living room tv: already off; living room fan: now off; living room lights: now off; kitchen oven: now off; kitchen coffee maker: already off; kitchen dishwasher: now off; kitchen lights: now off; bathroom fan: already off; bathroom lights: a... It answered next finish 0.60 What should happen next for `inputs.task`? It chose among 12 finish 0.60 bedroom The bedroom: lights, lamp, window. 0.27 good night Going to bed: the rest of the house off, th... 0.10 thermostat The thermostat: mode, target. 0.02 leaving Going out and leaving the house empty: every... 0.01 entry The entry: lights, door, lock, alarm. 0.00 living room The living room: lights, tv, fa[illegible] kitchen The kitchen: lights, oven, coff[illegible] and 4 more Stopped at a decision. PRINTED [illegible] garage lights: now off; gara[illegible] lights: now off; entry door:[illegible] locked) finish Interface text: Recursive Decision Models Pinboard Notebook Smart home Hierarchical search EXAMPLES 1 The house Ask it for something. It keeps its state from one request to the next. 2 One request 3 The map 4 With a language model 5 A REPL 6 Making a starter map 7 A starter map 8 A map with labels We are off to bed. Morning! Turn the TV on. Close the garage door. We are leaving for the weekend. Let's watch a film. I left the oven on. Set the thermostat to 20 degrees. Lock the front door. It is too hot in here. Set the alarm. Let's watch a film. recursivedecisionmodels.com ### What a decision model points at The map gives the decision model and the LLM a shared set of names. The decision model reads the descriptions and picks a name. The code uses that name to look up what the model selected. [29:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1779s) [A decision model can only point. A map gives each choice a name, a description for the model and a corresponding item for the code.](https://assets.ai.engineer/sites/aiecode2025/sha256/2b1cb7a84479e35657fee18b05243429dcdf560b0ce1802a3f6f5bb4b42b8900.webp) A decision model can only point, so something has to say what it is pointing at. A map links decision models to LLMs: names, each with a description for the model and a thing for the code. ### Fast and cheap means we can speculate When an agent is unsure what to do, it can run several fast, inexpensive decision models in parallel. An arbiter chooses which result to act on so competing decisions don’t make conflicting changes to shared state. [32:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1925s) [Under the centered title, the subtitle reads “Run many decisions at once, and let the system decide which to trust.”](https://assets.ai.engineer/sites/aiecode2025/sha256/b2315429508fa52215ab746cce4f2f0daa886566896ff54dcd955457f367f4cb.webp) Fast and cheap means we can speculate Run many decisions at once, and let the system decide which to trust. ## Full transcript [0:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=0s) Hey, my name is Raymond Whitampamp and [0:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3s) Hey, my name is Raymond Whitampamp and [0:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3s) Hey, my name is Raymond Whitampamp and my talk today is originally titled [0:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=6s) my talk today is originally titled [0:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=6s) my talk today is originally titled recursive decision models. And what I [0:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=10s) recursive decision models. And what I [0:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=10s) recursive decision models. And what I think I've realized since submitting [0:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=13s) think I've realized since submitting [0:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=13s) think I've realized since submitting that title is what it's really about, [0:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=15s) that title is what it's really about, [0:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=15s) that title is what it's really about, what we're really going to be talking [0:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=16s) what we're really going to be talking [0:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=16s) what we're really going to be talking about today is operating [0:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=19s) about today is operating [0:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=19s) about today is operating system one. Okay. So maybe some quick [0:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=24s) system one. Okay. So maybe some quick [0:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=24s) system one. Okay. So maybe some quick definitions or what what does that mean? [0:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=27s) definitions or what what does that mean? [0:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=27s) definitions or what what does that mean? Uh so system one models the the um [0:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=31s) Uh so system one models the the um [0:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=31s) Uh so system one models the the um naming of system one models from type [0:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=34s) naming of system one models from type [0:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=34s) naming of system one models from type safe when they released Jev um not even [0:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=38s) safe when they released Jev um not even [0:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=38s) safe when they released Jev um not even two months ago now I think um came from [0:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=43s) two months ago now I think um came from [0:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=43s) two months ago now I think um came from uh this concept that Danny Conaman [0:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=44s) uh this concept that Danny Conaman [0:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=44s) uh this concept that Danny Conaman popularized uh in thinking fast and [0:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=47s) popularized uh in thinking fast and [0:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=47s) popularized uh in thinking fast and slow. this idea of like the system one [0:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=49s) slow. this idea of like the system one [0:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=50s) slow. this idea of like the system one part of our uh brains is making very [0:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=52s) part of our uh brains is making very [0:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=52s) part of our uh brains is making very fast and intuitive like judgment calls, [0:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=55s) fast and intuitive like judgment calls, [0:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=55s) fast and intuitive like judgment calls, gut reactions, reflexes, maybe maybe [0:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=58s) gut reactions, reflexes, maybe maybe [0:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=58s) gut reactions, reflexes, maybe maybe reflex is more uh purely neurological [1:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=61s) reflex is more uh purely neurological [1:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=61s) reflex is more uh purely neurological but um that idea of like a lot of the [1:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=64s) but um that idea of like a lot of the [1:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=64s) but um that idea of like a lot of the decisions that we make are not actually [1:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=66s) decisions that we make are not actually [1:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=66s) decisions that we make are not actually like purely rational. Um and so system [1:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=70s) like purely rational. Um and so system [1:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=70s) like purely rational. Um and so system one is fast, system two is slow, more [1:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=73s) one is fast, system two is slow, more [1:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=74s) one is fast, system two is slow, more deliberate, like logical reasoning, [1:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=76s) deliberate, like logical reasoning, [1:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=76s) deliberate, like logical reasoning, problem solving, etc. And um I think [1:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=79s) problem solving, etc. And um I think [1:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=79s) problem solving, etc. And um I think what was a little unclear when they um [1:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=82s) what was a little unclear when they um [1:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=82s) what was a little unclear when they um when they announced Jev was whether they [1:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=84s) when they announced Jev was whether they [1:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=84s) when they announced Jev was whether they intended system one to be like their [1:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=88s) intended system one to be like their [1:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=88s) intended system one to be like their proprietary name family of models or [1:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=91s) proprietary name family of models or [1:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=91s) proprietary name family of models or like that's they were expecting all the [1:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=93s) like that's they were expecting all the [1:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=93s) like that's they were expecting all the other people to call them system one [1:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=94s) other people to call them system one [1:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=94s) other people to call them system one models. I would say most um most people [1:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=98s) models. I would say most um most people [1:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=98s) models. I would say most um most people are calling them decision models and um [1:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=101s) are calling them decision models and um [1:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=101s) are calling them decision models and um and so I wasn't really willing to u to [1:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=104s) and so I wasn't really willing to u to [1:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=104s) and so I wasn't really willing to u to wait for the people to decide and so so [1:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=108s) wait for the people to decide and so so [1:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=108s) wait for the people to decide and so so when I came up with recursive decision [1:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=110s) when I came up with recursive decision [1:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=110s) when I came up with recursive decision models um you know it was just it was [1:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=113s) models um you know it was just it was [1:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=113s) models um you know it was just it was just kind of using that. So, um, a rose [1:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=117s) just kind of using that. So, um, a rose [1:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=117s) just kind of using that. So, um, a rose by any other name, tomato, tomato, we [1:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=119s) by any other name, tomato, tomato, we [1:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=119s) by any other name, tomato, tomato, we don't really need to get too hung up on [2:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=121s) don't really need to get too hung up on [2:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=121s) don't really need to get too hung up on this, uh, idea of what are they? Are [2:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=124s) this, uh, idea of what are they? Are [2:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=124s) this, uh, idea of what are they? Are they, um, decision models or not? Um, [2:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=127s) they, um, decision models or not? Um, [2:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=127s) they, um, decision models or not? Um, Ally here argues that Jeff makes the [2:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=129s) Ally here argues that Jeff makes the [2:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=129s) Ally here argues that Jeff makes the judgments, um, not the decisions. Your [2:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=132s) judgments, um, not the decisions. Your [2:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=132s) judgments, um, not the decisions. Your code makes the decision. I think that's [2:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=133s) code makes the decision. I think that's [2:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=134s) code makes the decision. I think that's that's absolutely correct. Um, and yet I [2:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=136s) that's absolutely correct. Um, and yet I [2:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=136s) that's absolutely correct. Um, and yet I think that that much of the zeitgeist [2:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=139s) think that that much of the zeitgeist [2:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=139s) think that that much of the zeitgeist and much of the folks working in this [2:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=141s) and much of the folks working in this [2:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=141s) and much of the folks working in this space outside of Typesafe are are kind [2:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=144s) space outside of Typesafe are are kind [2:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=144s) space outside of Typesafe are are kind of uh rallying around this idea of um [2:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=148s) of uh rallying around this idea of um [2:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=148s) of uh rallying around this idea of um decision model being the term. There is [2:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=150s) decision model being the term. There is [2:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=150s) decision model being the term. There is the hold out camp of um refusing to call [2:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=153s) the hold out camp of um refusing to call [2:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=153s) the hold out camp of um refusing to call them anything other than zero shot [2:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=154s) them anything other than zero shot [2:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=154s) them anything other than zero shot classifiers. Um anyways, I don't have a [2:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=158s) classifiers. Um anyways, I don't have a [2:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=158s) classifiers. Um anyways, I don't have a dog in the race. the naming just trying [2:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=160s) dog in the race. the naming just trying [2:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=160s) dog in the race. the naming just trying to explain what I am talking about and I [2:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=162s) to explain what I am talking about and I [2:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=162s) to explain what I am talking about and I think that one of the things that is now [2:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=166s) think that one of the things that is now [2:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=166s) think that one of the things that is now connected here to this idea of system [2:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=167s) connected here to this idea of system [2:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=168s) connected here to this idea of system one system two Danny con econom [2:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=171s) one system two Danny con econom [2:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=171s) one system two Danny con econom um maybe related to how we behave but I [2:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=174s) um maybe related to how we behave but I [2:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=174s) um maybe related to how we behave but I think much more interesting from this [2:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=176s) think much more interesting from this [2:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=176s) think much more interesting from this perspective of building AI systems is [2:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=178s) perspective of building AI systems is [2:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=178s) perspective of building AI systems is what if reasoning uh was actually many [3:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=181s) what if reasoning uh was actually many [3:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=181s) what if reasoning uh was actually many small decisions and um yeah maybe a [3:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=185s) small decisions and um yeah maybe a [3:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=185s) small decisions and um yeah maybe a different way of saying this like what [3:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=187s) different way of saying this like what [3:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=187s) different way of saying this like what what if um behaviors or tasks that we [3:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=189s) what if um behaviors or tasks that we [3:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=189s) what if um behaviors or tasks that we might think of as requiring [3:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=192s) might think of as requiring [3:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=192s) might think of as requiring um a lot of reasoning or a lot of um [3:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=197s) um a lot of reasoning or a lot of um [3:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=197s) um a lot of reasoning or a lot of um like deliberate cognitive processing um [3:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=201s) like deliberate cognitive processing um [3:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=201s) like deliberate cognitive processing um could actually be um solved in a [3:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=203s) could actually be um solved in a [3:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=203s) could actually be um solved in a different way with a lot of very tiny [3:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=206s) different way with a lot of very tiny [3:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=206s) different way with a lot of very tiny very very fast decisions. We're going to [3:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=208s) very very fast decisions. We're going to [3:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=208s) very very fast decisions. We're going to keep coming back to this question [3:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=210s) keep coming back to this question [3:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=210s) keep coming back to this question throughout the talk and I think it's [3:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=213s) throughout the talk and I think it's [3:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=213s) throughout the talk and I think it's it's really key in what's interesting [3:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=215s) it's really key in what's interesting [3:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=215s) it's really key in what's interesting here. So [3:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=218s) here. So [3:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=218s) here. So recursive decision models. So this is um [3:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=221s) recursive decision models. So this is um [3:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=221s) recursive decision models. So this is um my concept um that I've been developing. [3:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=225s) my concept um that I've been developing. [3:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=225s) my concept um that I've been developing. It's um at its simplest perhaps the [3:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=228s) It's um at its simplest perhaps the [3:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=228s) It's um at its simplest perhaps the marriage of two things that I got really [3:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=230s) marriage of two things that I got really [3:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=230s) marriage of two things that I got really excited about. One just over a year old. [3:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=232s) excited about. One just over a year old. [3:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=232s) excited about. One just over a year old. It's called recursive language models [3:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=233s) It's called recursive language models [3:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=234s) It's called recursive language models RLM. This is from MIT Alex Singh and [3:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=237s) RLM. This is from MIT Alex Singh and [3:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=237s) RLM. This is from MIT Alex Singh and Omar. Their original uh blog post is [3:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=239s) Omar. Their original uh blog post is [3:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=239s) Omar. Their original uh blog post is just over a year ago. Happy birthday to [4:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=242s) just over a year ago. Happy birthday to [4:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=242s) just over a year ago. Happy birthday to RLM. Belated birthday. Um, and then like [4:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=246s) RLM. Belated birthday. Um, and then like [4:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=246s) RLM. Belated birthday. Um, and then like many others, I was also chomping at the [4:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=248s) many others, I was also chomping at the [4:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=248s) many others, I was also chomping at the bit to um I think it's actually champing [4:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=251s) bit to um I think it's actually champing [4:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=251s) bit to um I think it's actually champing at the bit. Anyways, I was champing and [4:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=253s) at the bit. Anyways, I was champing and [4:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=254s) at the bit. Anyways, I was champing and chomping uh to get off the freaking Jev [4:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=258s) chomping uh to get off the freaking Jev [4:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=258s) chomping uh to get off the freaking Jev weight list, which thankfully got off [4:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=260s) weight list, which thankfully got off [4:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=260s) weight list, which thankfully got off real quick. Um, they did a good job with [4:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=263s) real quick. Um, they did a good job with [4:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=263s) real quick. Um, they did a good job with that. But I think that one of the things [4:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=265s) that. But I think that one of the things [4:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=265s) that. But I think that one of the things that I've been sharing with people is I [4:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=268s) that I've been sharing with people is I [4:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=268s) that I've been sharing with people is I think that part of what type safe did so [4:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=271s) think that part of what type safe did so [4:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=271s) think that part of what type safe did so well and what captured the attention of [4:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=274s) well and what captured the attention of [4:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=274s) well and what captured the attention of so many engineers a engineers is um this [4:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=279s) so many engineers a engineers is um this [4:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=279s) so many engineers a engineers is um this idea of Jev as convergence right like [4:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=284s) idea of Jev as convergence right like [4:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=284s) idea of Jev as convergence right like decisions converge [4:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=286s) decisions converge [4:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=286s) decisions converge and that is in direct contrast to [4:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=289s) and that is in direct contrast to [4:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=289s) and that is in direct contrast to generative AI which has been you know [4:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=292s) generative AI which has been you know [4:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=292s) generative AI which has been you know blowing up like crazy It's arguably like [4:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=294s) blowing up like crazy It's arguably like [4:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=294s) blowing up like crazy It's arguably like the fastest growing [4:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=296s) the fastest growing [4:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=296s) the fastest growing um you know market in the history of the [4:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=299s) um you know market in the history of the [4:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=299s) um you know market in the history of the world and um so much of the attention [5:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=303s) world and um so much of the attention [5:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=303s) world and um so much of the attention and money in in the space has been about [5:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=306s) and money in in the space has been about [5:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=306s) and money in in the space has been about generation right and that's awesome and [5:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=308s) generation right and that's awesome and [5:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=308s) generation right and that's awesome and there's amazing things that are possible [5:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=312s) there's amazing things that are possible [5:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=312s) there's amazing things that are possible with generation but if you really think [5:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=314s) with generation but if you really think [5:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=314s) with generation but if you really think about it like even if you if you have [5:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=316s) about it like even if you if you have [5:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=316s) about it like even if you if you have infinite tokens right you kind of have [5:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=319s) infinite tokens right you kind of have [5:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=319s) infinite tokens right you kind of have infinite divergence right like the [5:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=322s) infinite divergence right like the [5:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=322s) infinite divergence right like the generative stuff it's actually hard to [5:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=325s) generative stuff it's actually hard to [5:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=325s) generative stuff it's actually hard to um converge. It's like we still have [5:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=327s) um converge. It's like we still have [5:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=327s) um converge. It's like we still have slop is someone still needs to decide [5:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=330s) slop is someone still needs to decide [5:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=330s) slop is someone still needs to decide like what's going to make it into [5:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=331s) like what's going to make it into [5:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=331s) like what's going to make it into production and maybe you've got your [5:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=333s) production and maybe you've got your [5:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=333s) production and maybe you've got your like code factory set up there. All [5:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=335s) like code factory set up there. All [5:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=335s) like code factory set up there. All those like agents are doing those [5:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=337s) those like agents are doing those [5:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=337s) those like agents are doing those decisions for you. Um, but at the end of [5:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=340s) decisions for you. Um, but at the end of [5:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=340s) decisions for you. Um, but at the end of the day, like the the [5:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=344s) the day, like the the [5:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=344s) the day, like the the important thing that I got me excited [5:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=347s) important thing that I got me excited [5:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=347s) important thing that I got me excited and I think I think this is why many [5:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=349s) and I think I think this is why many [5:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=349s) and I think I think this is why many other people are so excited about about [5:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=351s) other people are so excited about about [5:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=351s) other people are so excited about about Jev, these system one models, these [5:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=353s) Jev, these system one models, these [5:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=353s) Jev, these system one models, these decision models is [snorts] this idea of [5:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=356s) decision models is [snorts] this idea of [5:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=356s) decision models is [snorts] this idea of yin and yang, right? It's a [5:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=357s) yin and yang, right? It's a [5:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=357s) yin and yang, right? It's a counterbalancing force. Um, and so the [6:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=360s) counterbalancing force. Um, and so the [6:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=360s) counterbalancing force. Um, and so the decision model cannot generate text. It [6:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=362s) decision model cannot generate text. It [6:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=362s) decision model cannot generate text. It can only uh rank and pick and choose [6:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=365s) can only uh rank and pick and choose [6:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=365s) can only uh rank and pick and choose amongst options. Um, and that is like in [6:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=368s) amongst options. Um, and that is like in [6:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=368s) amongst options. Um, and that is like in direct contrast to LMS which have this [6:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=371s) direct contrast to LMS which have this [6:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=371s) direct contrast to LMS which have this ability to just like spit out uh [6:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=373s) ability to just like spit out uh [6:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=373s) ability to just like spit out uh freewheeling open-ended um text and in [6:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=376s) freewheeling open-ended um text and in [6:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=376s) freewheeling open-ended um text and in the early days actually hard to get them [6:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=378s) the early days actually hard to get them [6:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=378s) the early days actually hard to get them to structure their output. So I I see [6:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=382s) to structure their output. So I I see [6:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=382s) to structure their output. So I I see this as as yin and yang. Um I am not [6:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=385s) this as as yin and yang. Um I am not [6:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=385s) this as as yin and yang. Um I am not going to go super deep into um recursive [6:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=389s) going to go super deep into um recursive [6:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=389s) going to go super deep into um recursive language models. I would highly [6:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=390s) language models. I would highly [6:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=390s) language models. I would highly encourage you to check out my talk from [6:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=394s) encourage you to check out my talk from [6:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=394s) encourage you to check out my talk from the summer from AI uh engineer world's [6:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=398s) the summer from AI uh engineer world's [6:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=398s) the summer from AI uh engineer world's fair this summer recursive coding [6:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=400s) fair this summer recursive coding [6:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=400s) fair this summer recursive coding agents. It's on their YouTube channel. [6:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=401s) agents. It's on their YouTube channel. [6:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=401s) agents. It's on their YouTube channel. It's also at recursivecoding agents.com. [6:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=404s) It's also at recursivecoding agents.com. [6:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=404s) It's also at recursivecoding agents.com. Um my slides here are also a um a [6:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=408s) Um my slides here are also a um a [6:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=408s) Um my slides here are also a um a website and I'm going to make them [6:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=410s) website and I'm going to make them [6:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=410s) website and I'm going to make them available publicly after I record this. [6:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=413s) available publicly after I record this. [6:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=413s) available publicly after I record this. Um that is going to go a lot deeper into [6:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=416s) Um that is going to go a lot deeper into [6:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=416s) Um that is going to go a lot deeper into why I think RLMs are the new reasoning [6:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=418s) why I think RLMs are the new reasoning [6:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=418s) why I think RLMs are the new reasoning models. uh why this is a really [6:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=419s) models. uh why this is a really [6:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=419s) models. uh why this is a really important paradigm. So with that we do [7:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=423s) important paradigm. So with that we do [7:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=424s) important paradigm. So with that we do need to sort of just draw out the direct [7:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=426s) need to sort of just draw out the direct [7:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=426s) need to sort of just draw out the direct the direct analogy here. And so in the [7:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=430s) the direct analogy here. And so in the [7:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=430s) the direct analogy here. And so in the RLM idea, right, the root recursive [7:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=433s) RLM idea, right, the root recursive [7:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=433s) RLM idea, right, the root recursive language model has the ability to call [7:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=436s) language model has the ability to call [7:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=436s) language model has the ability to call sub recursive language models and those [7:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=439s) sub recursive language models and those [7:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=439s) sub recursive language models and those could call sub recursive language models [7:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=441s) could call sub recursive language models [7:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=441s) could call sub recursive language models and then you can you have to sort of [7:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=442s) and then you can you have to sort of [7:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=442s) and then you can you have to sort of like cap the depth of recursion to just [7:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=445s) like cap the depth of recursion to just [7:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=445s) like cap the depth of recursion to just not um if you ever want to like come [7:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=449s) not um if you ever want to like come [7:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=449s) not um if you ever want to like come come back from the deaths. Um and so [7:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=451s) come back from the deaths. Um and so [7:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=452s) come back from the deaths. Um and so here we say like okay at depth two then [7:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=454s) here we say like okay at depth two then [7:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=454s) here we say like okay at depth two then it's an LLM. So the the leaf node in [7:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=457s) it's an LLM. So the the leaf node in [7:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=457s) it's an LLM. So the the leaf node in this graph in this tree would would [7:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=459s) this graph in this tree would would [7:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=459s) this graph in this tree would would ultimately be an LLM call. We can make [7:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=461s) ultimately be an LLM call. We can make [7:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=461s) ultimately be an LLM call. We can make that exact analogy, that exact map onto [7:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=465s) that exact analogy, that exact map onto [7:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=465s) that exact analogy, that exact map onto system one models, onto decision models [7:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=467s) system one models, onto decision models [7:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=467s) system one models, onto decision models and say, okay, we've got a route um that [7:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=469s) and say, okay, we've got a route um that [7:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=469s) and say, okay, we've got a route um that can decide to call a sub uh RDM uh and [7:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=474s) can decide to call a sub uh RDM uh and [7:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=474s) can decide to call a sub uh RDM uh and then there's some max depth at which [7:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=476s) then there's some max depth at which [7:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=476s) then there's some max depth at which point we just have to make like a single [7:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=478s) point we just have to make like a single [7:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=478s) point we just have to make like a single call, a single system one call, a single [8:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=480s) call, a single system one call, a single [8:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=480s) call, a single system one call, a single decision call um but we also have to [8:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=485s) decision call um but we also have to [8:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=485s) decision call um but we also have to think about composing these things [8:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=487s) think about composing these things [8:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=487s) think about composing these things together, right? And the um I am going [8:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=492s) together, right? And the um I am going [8:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=492s) together, right? And the um I am going to show some examples as pure decisions [8:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=494s) to show some examples as pure decisions [8:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=494s) to show some examples as pure decisions all the way down. That's very important. [8:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=496s) all the way down. That's very important. [8:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=496s) all the way down. That's very important. Um this bottom left of what I'm I'm [8:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=499s) Um this bottom left of what I'm I'm [8:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=499s) Um this bottom left of what I'm I'm calling decision as a tool. Um so you [8:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=503s) calling decision as a tool. Um so you [8:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=503s) calling decision as a tool. Um so you could you could also this is what my [8:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=504s) could you could also this is what my [8:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=504s) could you could also this is what my recursive coding agent talk was about. [8:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=506s) recursive coding agent talk was about. [8:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=506s) recursive coding agent talk was about. It's like replace RLM with like agent, [8:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=509s) It's like replace RLM with like agent, [8:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=509s) It's like replace RLM with like agent, right? Coding agent. Okay, you got the [8:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=510s) right? Coding agent. Okay, you got the [8:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=510s) right? Coding agent. Okay, you got the root coding agent, you got the sub agent [8:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=512s) root coding agent, you got the sub agent [8:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=512s) root coding agent, you got the sub agent da da da. So I would say like this [8:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=515s) da da da. So I would say like this [8:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=515s) da da da. So I would say like this decisions as a tool or Jev as a tool was [8:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=518s) decisions as a tool or Jev as a tool was [8:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=518s) decisions as a tool or Jev as a tool was is like the obvious one and within you [8:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=522s) is like the obvious one and within you [8:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=522s) is like the obvious one and within you know 24 hours of people getting access [8:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=525s) know 24 hours of people getting access [8:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=525s) know 24 hours of people getting access to Jev there's tons and tons of examples [8:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=527s) to Jev there's tons and tons of examples [8:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=527s) to Jev there's tons and tons of examples here right people using Jev for [8:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=528s) here right people using Jev for [8:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=528s) here right people using Jev for compaction people using Jev as the thing [8:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=531s) compaction people using Jev as the thing [8:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=531s) compaction people using Jev as the thing that decides whether a tool call can [8:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=533s) that decides whether a tool call can [8:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=533s) that decides whether a tool call can execute or not people using Jev as a [8:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=536s) execute or not people using Jev as a [8:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=536s) execute or not people using Jev as a router to decide which um which model is [8:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=539s) router to decide which um which model is [8:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=539s) router to decide which um which model is going to get this particular request [9:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=540s) going to get this particular request [9:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=540s) going to get this particular request right so that stuff's Um, it's all very [9:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=544s) right so that stuff's Um, it's all very [9:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=544s) right so that stuff's Um, it's all very cool. There's some really interesting [9:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=546s) cool. There's some really interesting [9:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=546s) cool. There's some really interesting cases there and people just went buck [9:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=548s) cases there and people just went buck [9:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=548s) cases there and people just went buck wild going super fast um figuring that [9:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=550s) wild going super fast um figuring that [9:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=550s) wild going super fast um figuring that out. I think to me what's I'm more [9:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=553s) out. I think to me what's I'm more [9:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=553s) out. I think to me what's I'm more interested in is um is not as obvious [9:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=557s) interested in is um is not as obvious [9:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=557s) interested in is um is not as obvious how you would set it up. Uh and that's [9:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=560s) how you would set it up. Uh and that's [9:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=560s) how you would set it up. Uh and that's where [9:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=561s) where [9:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=561s) where the root is a decision model and you [9:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=564s) the root is a decision model and you [9:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=564s) the root is a decision model and you only can you could choose as a decision [9:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=568s) only can you could choose as a decision [9:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=568s) only can you could choose as a decision to call to a language model or an agent [9:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=571s) to call to a language model or an agent [9:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=571s) to call to a language model or an agent um as one as one possibility. So almost [9:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=575s) um as one as one possibility. So almost [9:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=575s) um as one as one possibility. So almost everything that I'm going to be talking [9:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=576s) everything that I'm going to be talking [9:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=576s) everything that I'm going to be talking about with RDM is going to um be in this [9:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=580s) about with RDM is going to um be in this [9:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=580s) about with RDM is going to um be in this paradigm where we're actually letting [9:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=582s) paradigm where we're actually letting [9:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=582s) paradigm where we're actually letting the root or the parent of this system be [9:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=586s) the root or the parent of this system be [9:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=586s) the root or the parent of this system be the system one model, the decision [9:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=588s) the system one model, the decision [9:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=588s) the system one model, the decision model. Okay. So now I want to show this [9:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=591s) model. Okay. So now I want to show this [9:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=591s) model. Okay. So now I want to show this for a couple reasons. This you might [9:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=594s) for a couple reasons. This you might [9:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=594s) for a couple reasons. This you might recognize this. This is the Jev plays [9:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=596s) recognize this. This is the Jev plays [9:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=596s) recognize this. This is the Jev plays Doom. It was part of like the launch [9:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=599s) Doom. It was part of like the launch [10:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=600s) Doom. It was part of like the launch tweet storm and got a lot of attention. [10:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=603s) tweet storm and got a lot of attention. [10:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=603s) tweet storm and got a lot of attention. And yet, despite that, I would say it's [10:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=606s) And yet, despite that, I would say it's [10:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=606s) And yet, despite that, I would say it's also one of the most um [10:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=610s) also one of the most um [10:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=610s) also one of the most um underappreciated [10:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=611s) underappreciated [10:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=611s) underappreciated or at least I would say underdeveloped [10:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=614s) or at least I would say underdeveloped [10:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=614s) or at least I would say underdeveloped um [10:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=616s) um [10:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=616s) um uh examples here. And so I'm just going [10:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=618s) uh examples here. And so I'm just going [10:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=618s) uh examples here. And so I'm just going to show a very short clip from this [10:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=619s) to show a very short clip from this [10:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=620s) to show a very short clip from this where [10:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=620s) where [10:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=620s) where we compose their answers into the [10:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=621s) we compose their answers into the [10:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=621s) we compose their answers into the player's next action and keep repeating [10:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=623s) player's next action and keep repeating [10:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=624s) player's next action and keep repeating that loop as the game changes. This is [10:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=625s) that loop as the game changes. This is [10:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=626s) that loop as the game changes. This is what we call a composition of AI [10:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=627s) what we call a composition of AI [10:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=628s) what we call a composition of AI primitives. [10:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=628s) primitives. [10:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=628s) primitives. Composition of AI primitives. [10:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=630s) Composition of AI primitives. [10:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=630s) Composition of AI primitives. Now, this composition handles the moment [10:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=633s) Now, this composition handles the moment [10:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=633s) Now, this composition handles the moment to moment moment to moment. So basically [10:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=635s) to moment moment to moment. So basically [10:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=635s) to moment moment to moment. So basically the first thing that they're showing is [10:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=637s) the first thing that they're showing is [10:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=637s) the first thing that they're showing is this harness that they built of [10:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=638s) this harness that they built of [10:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=638s) this harness that they built of decisions to handle like the first [10:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=640s) decisions to handle like the first [10:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=640s) decisions to handle like the first person shooting aspect. Is it gonna [10:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=643s) person shooting aspect. Is it gonna [10:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=643s) person shooting aspect. Is it gonna hold, carry on, walk, dodge, blah, blah, [10:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=647s) hold, carry on, walk, dodge, blah, blah, [10:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=647s) hold, carry on, walk, dodge, blah, blah, blah? It has these actions, right? This [10:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=650s) blah? It has these actions, right? This [10:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=650s) blah? It has these actions, right? This is where I I it's so simple and it's [10:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=654s) is where I I it's so simple and it's [10:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=654s) is where I I it's so simple and it's very much in line with this idea of [10:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=655s) very much in line with this idea of [10:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=655s) very much in line with this idea of composing systems together, recursive [10:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=659s) composing systems together, recursive [10:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=659s) composing systems together, recursive decision models, right? A graph of [11:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=662s) decision models, right? A graph of [11:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=662s) decision models, right? A graph of different types of decisions. Um, here [11:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=665s) different types of decisions. Um, here [11:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=665s) different types of decisions. Um, here we go. So this is the second composition [11:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=667s) we go. So this is the second composition [11:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=667s) we go. So this is the second composition composition that decides where to [11:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=669s) composition that decides where to [11:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=669s) composition that decides where to explore next and they can work together [11:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=671s) explore next and they can work together [11:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=671s) explore next and they can work together to make it through a level. [11:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=673s) to make it through a level. [11:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=673s) to make it through a level. Okay. So now you have two compositions. [11:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=675s) Okay. So now you have two compositions. [11:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=675s) Okay. So now you have two compositions. One is the first person perspective. One [11:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=678s) One is the first person perspective. One [11:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=678s) One is the first person perspective. One is this like bird's eye or like mapaware [11:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=681s) is this like bird's eye or like mapaware [11:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=681s) is this like bird's eye or like mapaware perspective. And by composing those two [11:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=683s) perspective. And by composing those two [11:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=683s) perspective. And by composing those two things together you have a a doom [11:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=686s) things together you have a a doom [11:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=686s) things together you have a a doom playing decision agent. And now [11:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=690s) playing decision agent. And now [11:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=690s) playing decision agent. And now to me I don't know maybe I missed [11:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=692s) to me I don't know maybe I missed [11:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=692s) to me I don't know maybe I missed something but I would say like um I have [11:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=696s) something but I would say like um I have [11:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=696s) something but I would say like um I have seen almost nothing very very very few [11:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=699s) seen almost nothing very very very few [11:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=700s) seen almost nothing very very very few um examples that are actually like [11:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=702s) um examples that are actually like [11:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=702s) um examples that are actually like taking this uh forward that are mo that [11:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=705s) taking this uh forward that are mo that [11:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=705s) taking this uh forward that are mo that are moving this is a super powerful idea [11:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=707s) are moving this is a super powerful idea [11:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=707s) are moving this is a super powerful idea of composing these two systems together. [11:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=710s) of composing these two systems together. [11:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=710s) of composing these two systems together. It's really insane that with like [11:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=713s) It's really insane that with like [11:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=713s) It's really insane that with like essentially two separate um but coupled [11:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=718s) essentially two separate um but coupled [11:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=718s) essentially two separate um but coupled um so two distinct but coupled decision [12:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=720s) um so two distinct but coupled decision [12:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=720s) um so two distinct but coupled decision agents, right? You now have with pure [12:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=724s) agents, right? You now have with pure [12:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=724s) agents, right? You now have with pure Jev calls um a Doom playing agent, [12:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=727s) Jev calls um a Doom playing agent, [12:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=727s) Jev calls um a Doom playing agent, right? Like that is really wild and I [12:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=729s) right? Like that is really wild and I [12:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=729s) right? Like that is really wild and I think really underappreciated despite [12:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=730s) think really underappreciated despite [12:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=730s) think really underappreciated despite the fact that probably like I don't know [12:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=733s) the fact that probably like I don't know [12:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=733s) the fact that probably like I don't know millions or tens of millions of people [12:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=734s) millions or tens of millions of people [12:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=734s) millions or tens of millions of people saw this video. And it comes back to [12:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=737s) saw this video. And it comes back to [12:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=737s) saw this video. And it comes back to this question, what if reasoning uh was [12:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=740s) this question, what if reasoning uh was [12:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=740s) this question, what if reasoning uh was actually many small decisions, right? [12:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=743s) actually many small decisions, right? [12:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=743s) actually many small decisions, right? And um or what are the types of [12:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=746s) And um or what are the types of [12:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=746s) And um or what are the types of activities or categories of things that [12:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=749s) activities or categories of things that [12:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=749s) activities or categories of things that we might naively assume would need [12:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=751s) we might naively assume would need [12:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=751s) we might naively assume would need reasoning when in fact what we need is a [12:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=753s) reasoning when in fact what we need is a [12:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=753s) reasoning when in fact what we need is a smart way of architecting small fast [12:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=756s) smart way of architecting small fast [12:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=756s) smart way of architecting small fast decisions. Okay, so now on to recursive [12:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=760s) decisions. Okay, so now on to recursive [12:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=760s) decisions. Okay, so now on to recursive decision models. I think the the glib [12:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=763s) decision models. I think the the glib [12:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=764s) decision models. I think the the glib tagline for this would be like a [12:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=765s) tagline for this would be like a [12:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=765s) tagline for this would be like a decision that can decide to make another [12:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=767s) decision that can decide to make another [12:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=767s) decision that can decide to make another decision and [12:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=770s) decision and [12:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=770s) decision and um that's certainly an important element [12:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=773s) um that's certainly an important element [12:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=773s) um that's certainly an important element of it. Uh while I was developing these [12:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=777s) of it. Uh while I was developing these [12:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=777s) of it. Uh while I was developing these um this idea this concept I've I've [12:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=779s) um this idea this concept I've I've [12:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=779s) um this idea this concept I've I've given myself three rules. Okay, the [13:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=781s) given myself three rules. Okay, the [13:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=781s) given myself three rules. Okay, the three rules of RDMs are one that the [13:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=785s) three rules of RDMs are one that the [13:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=785s) three rules of RDMs are one that the context stays outside the model and so [13:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=788s) context stays outside the model and so [13:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=788s) context stays outside the model and so the code um can cut it into slices. So [13:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=792s) the code um can cut it into slices. So [13:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=792s) the code um can cut it into slices. So the the in the Doom example, the the [13:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=795s) the the in the Doom example, the the [13:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=795s) the the in the Doom example, the the full state of um the game uh is outside [13:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=801s) full state of um the game uh is outside [13:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=801s) full state of um the game uh is outside then those two little subh harnesses, [13:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=804s) then those two little subh harnesses, [13:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=804s) then those two little subh harnesses, the first person shooting harness has [13:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=807s) the first person shooting harness has [13:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=807s) the first person shooting harness has some specific subset of that. The um the [13:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=811s) some specific subset of that. The um the [13:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=811s) some specific subset of that. The um the map exploring bird's eye view decision [13:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=815s) map exploring bird's eye view decision [13:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=815s) map exploring bird's eye view decision model has another uh view of that. Uh [13:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=818s) model has another uh view of that. Uh [13:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=818s) model has another uh view of that. Uh the second thing is this uh and now this [13:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=821s) the second thing is this uh and now this [13:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=821s) the second thing is this uh and now this is where it's a little different from [13:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=822s) is where it's a little different from [13:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=822s) is where it's a little different from that demo is the decision can choose to [13:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=825s) that demo is the decision can choose to [13:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=825s) that demo is the decision can choose to call another decision and again up to [13:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=827s) call another decision and again up to [13:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=827s) call another decision and again up to this maximum depth we talked about this [13:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=829s) this maximum depth we talked about this [13:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=829s) this maximum depth we talked about this recursive tree [13:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=831s) recursive tree [13:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=832s) recursive tree now these two things are the most [13:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=834s) now these two things are the most [13:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=834s) now these two things are the most important fundamental aspects of RLM as [13:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=837s) important fundamental aspects of RLM as [13:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=837s) important fundamental aspects of RLM as well right the context is externalized [13:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=839s) well right the context is externalized [13:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=839s) well right the context is externalized the agent can make sub aents the RLM can [14:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=842s) the agent can make sub aents the RLM can [14:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=842s) the agent can make sub aents the RLM can call sub RLMs and sub LLM right um the [14:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=848s) call sub RLMs and sub LLM right um the [14:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=848s) call sub RLMs and sub LLM right um the so those two things are like copied [14:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=850s) so those two things are like copied [14:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=850s) so those two things are like copied straight from the RLM idea and it's [14:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=853s) straight from the RLM idea and it's [14:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=853s) straight from the RLM idea and it's applying that directly to to system one [14:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=855s) applying that directly to to system one [14:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=855s) applying that directly to to system one models decision models. Uh the third [14:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=858s) models decision models. Uh the third [14:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=858s) models decision models. Uh the third piece is first it must work with no [14:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=862s) piece is first it must work with no [14:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=862s) piece is first it must work with no generative model at all. And this was [14:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=864s) generative model at all. And this was [14:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=864s) generative model at all. And this was very important to me um because I again [14:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=868s) very important to me um because I again [14:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=868s) very important to me um because I again I'm like all this like Jev as a tool for [14:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=872s) I'm like all this like Jev as a tool for [14:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=872s) I'm like all this like Jev as a tool for a normal coding agent is is is cool and [14:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=875s) a normal coding agent is is is cool and [14:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=875s) a normal coding agent is is is cool and there's some really great examples but [14:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=876s) there's some really great examples but [14:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=876s) there's some really great examples but it's it's obvious to me. What was really [14:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=878s) it's it's obvious to me. What was really [14:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=878s) it's it's obvious to me. What was really interesting in terms of novel what might [14:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=880s) interesting in terms of novel what might [14:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=880s) interesting in terms of novel what might unlock some novel behaviors some novel [14:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=882s) unlock some novel behaviors some novel [14:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=882s) unlock some novel behaviors some novel capabilities [14:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=883s) capabilities [14:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=883s) capabilities um is what if there's no generative [14:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=885s) um is what if there's no generative [14:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=885s) um is what if there's no generative model at all. It also proves that the [14:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=887s) model at all. It also proves that the [14:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=887s) model at all. It also proves that the that the RDM architecture is actually uh [14:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=891s) that the RDM architecture is actually uh [14:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=891s) that the RDM architecture is actually uh correct like it should be able to work [14:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=893s) correct like it should be able to work [14:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=893s) correct like it should be able to work with no generation and at the same time [14:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=896s) with no generation and at the same time [14:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=896s) with no generation and at the same time it must compose with LMS and with agents [15:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=900s) it must compose with LMS and with agents [15:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=900s) it must compose with LMS and with agents right um so those are my three rules in [15:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=902s) right um so those are my three rules in [15:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=902s) right um so those are my three rules in developing this concept of recursive [15:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=904s) developing this concept of recursive [15:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=904s) developing this concept of recursive decision models. So first I want to show [15:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=907s) decision models. So first I want to show [15:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=907s) decision models. So first I want to show the the decisions all the way down [15:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=909s) the the decisions all the way down [15:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=909s) the the decisions all the way down example. Okay. So decisions all the way [15:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=913s) example. Okay. So decisions all the way [15:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=913s) example. Okay. So decisions all the way down. [15:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=915s) down. [15:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=915s) down. Um you might be familiar with this idea [15:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=917s) Um you might be familiar with this idea [15:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=917s) Um you might be familiar with this idea of a golton board. Um maybe you learned [15:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=920s) of a golton board. Um maybe you learned [15:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=920s) of a golton board. Um maybe you learned about this in the context of learning [15:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=921s) about this in the context of learning [15:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=921s) about this in the context of learning about um the the Gaussian distribution [15:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=925s) about um the the Gaussian distribution [15:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=925s) about um the the Gaussian distribution or or um uh what do they call it? Normal [15:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=928s) or or um uh what do they call it? Normal [15:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=928s) or or um uh what do they call it? Normal distribution. So like each ball has like [15:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=931s) distribution. So like each ball has like [15:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=931s) distribution. So like each ball has like a 50-50 chance at each pin of um falling [15:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=935s) a 50-50 chance at each pin of um falling [15:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=935s) a 50-50 chance at each pin of um falling and then you build this like bell curve [15:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=939s) and then you build this like bell curve [15:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=939s) and then you build this like bell curve um gaussian distribution. So the [15:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=941s) um gaussian distribution. So the [15:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=941s) um gaussian distribution. So the simplest way that I could think of to to [15:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=944s) simplest way that I could think of to to [15:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=944s) simplest way that I could think of to to visualize this idea of um like recursive [15:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=949s) visualize this idea of um like recursive [15:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=949s) visualize this idea of um like recursive layers of decisions is now what if the [15:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=952s) layers of decisions is now what if the [15:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=952s) layers of decisions is now what if the pins could decide? So you could think of [15:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=955s) pins could decide? So you could think of [15:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=955s) pins could decide? So you could think of this kind of like a binary search like [15:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=958s) this kind of like a binary search like [15:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=958s) this kind of like a binary search like so each um numbered ball just like oh is [16:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=961s) so each um numbered ball just like oh is [16:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=961s) so each um numbered ball just like oh is this one going to go higher or lower [16:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=964s) this one going to go higher or lower [16:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=964s) this one going to go higher or lower higher or lower? Um and so this pin is [16:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=967s) higher or lower? Um and so this pin is [16:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=968s) higher or lower? Um and so this pin is like splitting them in half. this pin [16:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=970s) like splitting them in half. this pin [16:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=970s) like splitting them in half. this pin spinning um in to its half etc. Um and [16:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=976s) spinning um in to its half etc. Um and [16:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=976s) spinning um in to its half etc. Um and again this is kind of like a a silly [16:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=978s) again this is kind of like a a silly [16:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=978s) again this is kind of like a a silly simple example. Uh but now when those [16:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=981s) simple example. Uh but now when those [16:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=981s) simple example. Uh but now when those are smart then we separate them [16:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=983s) are smart then we separate them [16:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=983s) are smart then we separate them perfectly. Now system one model like Jev [16:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=988s) perfectly. Now system one model like Jev [16:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=988s) perfectly. Now system one model like Jev decision model whatever we want to call [16:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=990s) decision model whatever we want to call [16:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=990s) decision model whatever we want to call them uh well we can ask about we can ask [16:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=993s) them uh well we can ask about we can ask [16:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=993s) them uh well we can ask about we can ask many questions at once. We don't have to [16:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=994s) many questions at once. We don't have to [16:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=994s) many questions at once. We don't have to just ask one on one axis. We could ask [16:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=997s) just ask one on one axis. We could ask [16:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=997s) just ask one on one axis. We could ask in multiple dimensions. So we can now [16:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1000s) in multiple dimensions. So we can now [16:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1000s) in multiple dimensions. So we can now visualize that in 3D as like the balls [16:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1003s) visualize that in 3D as like the balls [16:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1003s) visualize that in 3D as like the balls falling they're being sorted u by color [16:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1006s) falling they're being sorted u by color [16:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1006s) falling they're being sorted u by color in one axis by number in another. We [16:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1010s) in one axis by number in another. We [16:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1010s) in one axis by number in another. We could put the cross-sectional views here [16:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1012s) could put the cross-sectional views here [16:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1012s) could put the cross-sectional views here and see that in the case that this is a [16:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1014s) and see that in the case that this is a [16:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1014s) and see that in the case that this is a perfect um classifier right um it's [16:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1018s) perfect um classifier right um it's [16:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1018s) perfect um classifier right um it's perfectly classifying the numbered [17:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1020s) perfectly classifying the numbered [17:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1020s) perfectly classifying the numbered bucket that it should go in in the end [17:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1022s) bucket that it should go in in the end [17:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1022s) bucket that it should go in in the end the uh colored bucket. So again, this is [17:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1026s) the uh colored bucket. So again, this is [17:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1026s) the uh colored bucket. So again, this is very much a toy example in the sense [17:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1027s) very much a toy example in the sense [17:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1028s) very much a toy example in the sense that like actually what you you could do [17:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1030s) that like actually what you you could do [17:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1030s) that like actually what you you could do all of this in one classifier call. You [17:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1032s) all of this in one classifier call. You [17:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1032s) all of this in one classifier call. You could do all of this in one decision [17:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1034s) could do all of this in one decision [17:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1034s) could do all of this in one decision call. You don't need um this like [17:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1036s) call. You don't need um this like [17:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1036s) call. You don't need um this like recursive tree or like iterative tree to [17:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1040s) recursive tree or like iterative tree to [17:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1040s) recursive tree or like iterative tree to do it. But I'm um I want you to keep the [17:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1043s) do it. But I'm um I want you to keep the [17:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1043s) do it. But I'm um I want you to keep the visualization of this because it's [17:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1045s) visualization of this because it's [17:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1045s) visualization of this because it's exactly what's going to happen in some [17:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1046s) exactly what's going to happen in some [17:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1046s) exactly what's going to happen in some of these more specific examples of [17:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1049s) of these more specific examples of [17:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1049s) of these more specific examples of recursive decision models. Okay. So, [17:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1052s) recursive decision models. Okay. So, [17:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1052s) recursive decision models. Okay. So, hierarchical search was one of the [17:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1057s) hierarchical search was one of the [17:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1057s) hierarchical search was one of the launch day uh cookbooks of Typesafe and [17:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1061s) launch day uh cookbooks of Typesafe and [17:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1061s) launch day uh cookbooks of Typesafe and they what they did is they took um this [17:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1064s) they what they did is they took um this [17:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1064s) they what they did is they took um this Shopify uh database of 12,000 product [17:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1067s) Shopify uh database of 12,000 product [17:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1067s) Shopify uh database of 12,000 product categories and they implemented like a [17:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1071s) categories and they implemented like a [17:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1071s) categories and they implemented like a beam search algorithm with hierarchical [17:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1073s) beam search algorithm with hierarchical [17:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1073s) beam search algorithm with hierarchical calls to um Jev to figure out from from [17:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1078s) calls to um Jev to figure out from from [17:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1078s) calls to um Jev to figure out from from a user query which product category [18:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1081s) a user query which product category [18:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1081s) a user query which product category um the user's query was describing. [18:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1085s) um the user's query was describing. [18:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1085s) um the user's query was describing. And what is unique as a simple [18:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1089s) And what is unique as a simple [18:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1089s) And what is unique as a simple demonstration of what's possible with [18:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1090s) demonstration of what's possible with [18:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1090s) demonstration of what's possible with recursive decision models is you can do [18:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1093s) recursive decision models is you can do [18:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1093s) recursive decision models is you can do the beam search without the beam. So [18:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1097s) the beam search without the beam. So [18:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1097s) the beam search without the beam. So essentially at each layer here, we're [18:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1099s) essentially at each layer here, we're [18:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1100s) essentially at each layer here, we're simply asking well um of the next sub [18:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1104s) simply asking well um of the next sub [18:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1104s) simply asking well um of the next sub layer of products. What is the next one [18:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1107s) layer of products. What is the next one [18:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1107s) layer of products. What is the next one that we're interested [snorts] um in [18:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1109s) that we're interested [snorts] um in [18:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1110s) that we're interested [snorts] um in exploring? And so uh in this in this [18:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1113s) exploring? And so uh in this in this [18:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1113s) exploring? And so uh in this in this example, we're exploring we're not [18:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1115s) example, we're exploring we're not [18:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1115s) example, we're exploring we're not exploring all 12,000 categories at once. [18:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1117s) exploring all 12,000 categories at once. [18:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1118s) exploring all 12,000 categories at once. That would like overload the context [18:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1119s) That would like overload the context [18:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1119s) That would like overload the context window. We're taking the parent [18:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1121s) window. We're taking the parent [18:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1121s) window. We're taking the parent category. We're deciding okay the user [18:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1124s) category. We're deciding okay the user [18:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1124s) category. We're deciding okay the user said laptop backpack okay which category [18:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1128s) said laptop backpack okay which category [18:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1128s) said laptop backpack okay which category am I going to explore next and and [18:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1131s) am I going to explore next and and [18:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1131s) am I going to explore next and and iterating along [18:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1133s) iterating along [18:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1133s) iterating along uh from there um and we could decide in [18:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1137s) uh from there um and we could decide in [18:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1137s) uh from there um and we could decide in the code how we implement it. Do we do [18:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1139s) the code how we implement it. Do we do [18:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1139s) the code how we implement it. Do we do greedy? Do we pick like only the top [19:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1142s) greedy? Do we pick like only the top [19:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1142s) greedy? Do we pick like only the top one? Right? Do we explore [sighs] [19:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1145s) one? Right? Do we explore [sighs] [19:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1145s) one? Right? Do we explore [sighs] top three um that are rated uh with the [19:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1149s) top three um that are rated uh with the [19:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1149s) top three um that are rated uh with the different probabilities etc? But you [19:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1151s) different probabilities etc? But you [19:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1151s) different probabilities etc? But you could kind of again it's a like similar [19:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1153s) could kind of again it's a like similar [19:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1153s) could kind of again it's a like similar to a binary search algorithm in theme [19:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1158s) to a binary search algorithm in theme [19:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1158s) to a binary search algorithm in theme except that like the compositional unit [19:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1161s) except that like the compositional unit [19:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1161s) except that like the compositional unit of it is a classifier is a judgment is a [19:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1165s) of it is a classifier is a judgment is a [19:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1165s) of it is a classifier is a judgment is a decision model is a system one model [19:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1167s) decision model is a system one model [19:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1167s) decision model is a system one model again too many names I don't want to [19:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1169s) again too many names I don't want to [19:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1169s) again too many names I don't want to fight about the names okay visually um [19:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1173s) fight about the names okay visually um [19:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1173s) fight about the names okay visually um we can show this um as well and so the [19:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1177s) we can show this um as well and so the [19:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1178s) we can show this um as well and so the uh Jev quickly became uh multiot total [19:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1181s) uh Jev quickly became uh multiot total [19:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1181s) uh Jev quickly became uh multiot total uh just a few weeks ago, about a month [19:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1184s) uh just a few weeks ago, about a month [19:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1184s) uh just a few weeks ago, about a month after the initial uh launch. And um Lily [19:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1188s) after the initial uh launch. And um Lily [19:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1188s) after the initial uh launch. And um Lily over here uh posted like the first [19:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1191s) over here uh posted like the first [19:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1191s) over here uh posted like the first variation of this, which is like [19:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1193s) variation of this, which is like [19:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1193s) variation of this, which is like wherever you are getting a hit, you are [19:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1196s) wherever you are getting a hit, you are [19:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1196s) wherever you are getting a hit, you are um now then kind of like zooming in to [20:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1200s) um now then kind of like zooming in to [20:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1200s) um now then kind of like zooming in to have like a feedback loop and like [20:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1202s) have like a feedback loop and like [20:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1202s) have like a feedback loop and like figure out, you know, ask again on a on [20:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1204s) figure out, you know, ask again on a on [20:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1204s) figure out, you know, ask again on a on a smaller box. Okay. Is this really uh [20:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1207s) a smaller box. Okay. Is this really uh [20:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1207s) a smaller box. Okay. Is this really uh in her her case looking for u beetles, [20:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1210s) in her her case looking for u beetles, [20:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1210s) in her her case looking for u beetles, right? Um and then Nathan from the type [20:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1213s) right? Um and then Nathan from the type [20:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1213s) right? Um and then Nathan from the type safe team kind of followed up on this [20:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1215s) safe team kind of followed up on this [20:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1215s) safe team kind of followed up on this and um what's interesting about his [20:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1217s) and um what's interesting about his [20:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1217s) and um what's interesting about his implic implementation is it's much more [20:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1220s) implic implementation is it's much more [20:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1220s) implic implementation is it's much more like um this um binary search in the [20:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1224s) like um this um binary search in the [20:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1224s) like um this um binary search in the sense it's like a recursive having of [20:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1227s) sense it's like a recursive having of [20:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1227s) sense it's like a recursive having of the of the image, right? Like a like a [20:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1230s) the of the image, right? Like a like a [20:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1230s) the of the image, right? Like a like a golden um golden spiral style like [20:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1233s) golden um golden spiral style like [20:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1233s) golden um golden spiral style like having, right? And then within those uh [20:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1236s) having, right? And then within those uh [20:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1236s) having, right? And then within those uh blocks within those buckets um uh like [20:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1240s) blocks within those buckets um uh like [20:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1240s) blocks within those buckets um uh like what what is it that we're um seeing? So [20:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1244s) what what is it that we're um seeing? So [20:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1244s) what what is it that we're um seeing? So again, this is this is the um these are [20:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1247s) again, this is this is the um these are [20:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1247s) again, this is this is the um these are simple simple examples that you you [20:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1250s) simple simple examples that you you [20:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1250s) simple simple examples that you you might not need recursion here like a you [20:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1253s) might not need recursion here like a you [20:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1253s) might not need recursion here like a you know a good vision model could probably [20:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1255s) know a good vision model could probably [20:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1255s) know a good vision model could probably answer the question in one shot and I [20:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1257s) answer the question in one shot and I [20:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1257s) answer the question in one shot and I think some people have have pointed that [20:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1259s) think some people have have pointed that [20:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1259s) think some people have have pointed that out clearly but I'm I'm trying to um [21:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1263s) out clearly but I'm I'm trying to um [21:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1263s) out clearly but I'm I'm trying to um give a number of different visual [21:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1264s) give a number of different visual [21:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1264s) give a number of different visual examples of of how a recursive decision [21:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1266s) examples of of how a recursive decision [21:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1266s) examples of of how a recursive decision model could work and where um um [21:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1271s) model could work and where um um [21:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1271s) model could work and where um um different ways of you seeing what that [21:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1273s) different ways of you seeing what that [21:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1273s) different ways of you seeing what that might look like um in terms of the [21:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1275s) might look like um in terms of the [21:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1275s) might look like um in terms of the recursion. So now let's get back to to [21:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1277s) recursion. So now let's get back to to [21:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1277s) recursion. So now let's get back to to playing video games. And so this is this [21:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1279s) playing video games. And so this is this [21:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1279s) playing video games. And so this is this is um me playing Factorio [21:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1282s) is um me playing Factorio [21:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1282s) is um me playing Factorio with a recursive decision model. So [21:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1286s) with a recursive decision model. So [21:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1286s) with a recursive decision model. So we've got all of these um these lines [21:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1289s) we've got all of these um these lines [21:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1289s) we've got all of these um these lines that are making our steel or whatever. [21:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1292s) that are making our steel or whatever. [21:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1292s) that are making our steel or whatever. I'm I'm not very good at factorial. Um [21:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1294s) I'm I'm not very good at factorial. Um [21:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1294s) I'm I'm not very good at factorial. Um so apologize. Uh, I actually would like [21:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1296s) so apologize. Uh, I actually would like [21:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1296s) so apologize. Uh, I actually would like to um my goal actually with this is to [21:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1299s) to um my goal actually with this is to [21:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1299s) to um my goal actually with this is to have my recursive decision model teach [21:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1301s) have my recursive decision model teach [21:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1301s) have my recursive decision model teach me how to play Factorio because I' i'd [21:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1303s) me how to play Factorio because I' i'd [21:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1303s) me how to play Factorio because I' i'd never really played before, but it's a [21:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1305s) never really played before, but it's a [21:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1305s) never really played before, but it's a it's a pretty cool like very open-ended [21:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1308s) it's a pretty cool like very open-ended [21:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1308s) it's a pretty cool like very open-ended um way of evaluating these systems and [21:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1310s) um way of evaluating these systems and [21:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1310s) um way of evaluating these systems and benchmarking systems. So, you can see [21:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1312s) benchmarking systems. So, you can see [21:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1312s) benchmarking systems. So, you can see here it's got all these lines. It's [21:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1313s) here it's got all these lines. It's [21:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1313s) here it's got all these lines. It's figuring, okay, this group is the the [21:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1316s) figuring, okay, this group is the the [21:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1316s) figuring, okay, this group is the the one that's um got a problem. Uh this [21:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1319s) one that's um got a problem. Uh this [22:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1320s) one that's um got a problem. Uh this line is the one that's got a problem. [22:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1323s) line is the one that's got a problem. [22:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1323s) line is the one that's got a problem. It's this element, the inserter that I [22:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1326s) It's this element, the inserter that I [22:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1326s) It's this element, the inserter that I think I need to go uh fix and I need to [22:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1330s) think I need to go uh fix and I need to [22:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1330s) think I need to go uh fix and I need to replace it. So now it's going to run [22:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1333s) replace it. So now it's going to run [22:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1333s) replace it. So now it's going to run over and go do that. And now the thing [22:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1336s) over and go do that. And now the thing [22:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1336s) over and go do that. And now the thing that I I want to point out here is there [22:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1340s) that I I want to point out here is there [22:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1340s) that I I want to point out here is there is no reasoning here. There's no large [22:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1343s) is no reasoning here. There's no large [22:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1343s) is no reasoning here. There's no large language model here. Um there's nothing [22:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1346s) language model here. Um there's nothing [22:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1346s) language model here. Um there's nothing generative here, right? So, it's kind it [22:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1349s) generative here, right? So, it's kind it [22:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1349s) generative here, right? So, it's kind it I don't know, maybe it's a little bit of [22:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1350s) I don't know, maybe it's a little bit of [22:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1350s) I don't know, maybe it's a little bit of a toy example in that it's literally [22:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1352s) a toy example in that it's literally [22:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1352s) a toy example in that it's literally like a matrix of these things and [22:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1355s) like a matrix of these things and [22:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1355s) like a matrix of these things and they're all the same. Um, but then it [22:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1358s) they're all the same. Um, but then it [22:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1358s) they're all the same. Um, but then it moves on to the next one. You'll see [22:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1360s) moves on to the next one. You'll see [22:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1360s) moves on to the next one. You'll see actually I should have maybe added some [22:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1363s) actually I should have maybe added some [22:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1363s) actually I should have maybe added some path finding composition of recursion [22:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1367s) path finding composition of recursion [22:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1367s) path finding composition of recursion because it ran like the long way around [22:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1370s) because it ran like the long way around [22:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1370s) because it ran like the long way around to go the next one. So, it's not totally [22:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1372s) to go the next one. So, it's not totally [22:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1372s) to go the next one. So, it's not totally perfect, right? But this is just an [22:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1374s) perfect, right? But this is just an [22:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1374s) perfect, right? But this is just an example of what what if you just looked [22:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1377s) example of what what if you just looked [22:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1377s) example of what what if you just looked at this without knowing it. I think you [22:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1379s) at this without knowing it. I think you [22:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1379s) at this without knowing it. I think you would I don't know at least I would [23:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1381s) would I don't know at least I would [23:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1381s) would I don't know at least I would assume that you need some kind of [23:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1384s) assume that you need some kind of [23:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1384s) assume that you need some kind of reasoning model here. You need um and um [23:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1388s) reasoning model here. You need um and um [23:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1388s) reasoning model here. You need um and um yeah you like this is like I don't know [23:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1390s) yeah you like this is like I don't know [23:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1390s) yeah you like this is like I don't know like Claude plays Pokemon level uh thing [23:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1393s) like Claude plays Pokemon level uh thing [23:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1393s) like Claude plays Pokemon level uh thing and it to me it's just absolutely [23:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1395s) and it to me it's just absolutely [23:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1395s) and it to me it's just absolutely fascinating that this can be done uh [23:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1397s) fascinating that this can be done uh [23:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1397s) fascinating that this can be done uh without any uh generative language model [23:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1399s) without any uh generative language model [23:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1399s) without any uh generative language model in there at all. [23:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1401s) in there at all. [23:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1401s) in there at all. So again we come back to this idea. What [23:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1403s) So again we come back to this idea. What [23:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1403s) So again we come back to this idea. What if reasoning or tasks that we might have [23:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1406s) if reasoning or tasks that we might have [23:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1406s) if reasoning or tasks that we might have assumed naively required reasoning can [23:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1408s) assumed naively required reasoning can [23:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1408s) assumed naively required reasoning can actually be done with these many many [23:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1410s) actually be done with these many many [23:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1410s) actually be done with these many many many small decisions. [23:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1413s) many small decisions. [23:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1413s) many small decisions. All right. So the the weakness or the [23:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1416s) All right. So the the weakness or the [23:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1416s) All right. So the the weakness or the downside um the limitation of decisions [23:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1420s) downside um the limitation of decisions [23:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1420s) downside um the limitation of decisions all the way down, right? Is that [23:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1423s) all the way down, right? Is that [23:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1423s) all the way down, right? Is that something has to have made those [23:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1425s) something has to have made those [23:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1425s) something has to have made those decisions like what what um are the [23:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1428s) decisions like what what um are the [23:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1428s) decisions like what what um are the allowed decisions and um for any of you [23:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1432s) allowed decisions and um for any of you [23:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1432s) allowed decisions and um for any of you who have a type safe account, you might [23:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1435s) who have a type safe account, you might [23:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1435s) who have a type safe account, you might remember the the quiz. You have to pass [23:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1437s) remember the the quiz. You have to pass [23:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1437s) remember the the quiz. You have to pass a quiz to get an API key. And the the [24:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1440s) a quiz to get an API key. And the the [24:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1440s) a quiz to get an API key. And the the question is can Jeff generate text or [24:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1443s) question is can Jeff generate text or [24:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1443s) question is can Jeff generate text or not? And the answer is no. And if you [24:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1445s) not? And the answer is no. And if you [24:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1445s) not? And the answer is no. And if you put yes, it won't give you an API key. [24:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1448s) put yes, it won't give you an API key. [24:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1448s) put yes, it won't give you an API key. So, so the um you h something like [24:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1452s) So, so the um you h something like [24:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1452s) So, so the um you h something like programmatically um whatever like a [24:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1456s) programmatically um whatever like a [24:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1456s) programmatically um whatever like a script has to do it, another agent has [24:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1458s) script has to do it, another agent has [24:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1458s) script has to do it, another agent has to do it. The application itself has to [24:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1461s) to do it. The application itself has to [24:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1461s) to do it. The application itself has to have pre-enumerated all of it, right? Um [24:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1464s) have pre-enumerated all of it, right? Um [24:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1464s) have pre-enumerated all of it, right? Um if you can only have decisions, then [24:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1466s) if you can only have decisions, then [24:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1466s) if you can only have decisions, then there's no generative aspect. Um, and [24:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1469s) there's no generative aspect. Um, and [24:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1469s) there's no generative aspect. Um, and what's crazy to me is that for games, [24:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1474s) what's crazy to me is that for games, [24:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1474s) what's crazy to me is that for games, for browsers, and I think even for a lot [24:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1476s) for browsers, and I think even for a lot [24:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1476s) for browsers, and I think even for a lot of elements of computer use, that [24:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1478s) of elements of computer use, that [24:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1478s) of elements of computer use, that actually might be enough, right? And I [24:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1481s) actually might be enough, right? And I [24:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1481s) actually might be enough, right? And I think if you kind of visualize now back [24:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1483s) think if you kind of visualize now back [24:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1483s) think if you kind of visualize now back all these examples that we just showed, [24:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1485s) all these examples that we just showed, [24:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1485s) all these examples that we just showed, uh, you hopefully can understand why [24:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1487s) uh, you hopefully can understand why [24:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1487s) uh, you hopefully can understand why with a tree of many many many small [24:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1489s) with a tree of many many many small [24:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1489s) with a tree of many many many small decisions, we can actually um find [24:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1492s) decisions, we can actually um find [24:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1492s) decisions, we can actually um find things. So you imagine a game the number [24:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1495s) things. So you imagine a game the number [24:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1495s) things. So you imagine a game the number the total possible um actions available [24:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1498s) the total possible um actions available [24:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1498s) the total possible um actions available to a user the total um possible state of [25:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1502s) to a user the total um possible state of [25:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1502s) to a user the total um possible state of the game um can be like enumerated [25:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1506s) the game um can be like enumerated [25:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1506s) the game um can be like enumerated programmatically and recursively [25:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1509s) programmatically and recursively [25:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1509s) programmatically and recursively addressed to make judgments make [25:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1511s) addressed to make judgments make [25:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1511s) addressed to make judgments make decisions about what what the next move [25:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1513s) decisions about what what the next move [25:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1513s) decisions about what what the next move is um in a browser. So these slides are [25:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1517s) is um in a browser. So these slides are [25:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1517s) is um in a browser. So these slides are is actually a website. Um in this [25:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1520s) is actually a website. Um in this [25:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1520s) is actually a website. Um in this particular website you can go up, you [25:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1521s) particular website you can go up, you [25:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1521s) particular website you can go up, you can go down, you can start a video. So [25:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1523s) can go down, you can start a video. So [25:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1523s) can go down, you can start a video. So there's like three interactive elements [25:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1525s) there's like three interactive elements [25:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1525s) there's like three interactive elements in the DOM. Um the the you think just [25:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1530s) in the DOM. Um the the you think just [25:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1530s) in the DOM. Um the the you think just think about it on your computer. Like [25:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1531s) think about it on your computer. Like [25:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1531s) think about it on your computer. Like really if you're on your computer right [25:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1533s) really if you're on your computer right [25:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1533s) really if you're on your computer right now, like how many um how many actions [25:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1537s) now, like how many um how many actions [25:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1537s) now, like how many um how many actions are available to you on the current [25:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1539s) are available to you on the current [25:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1539s) are available to you on the current screen uh or with like the hot keys? [25:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1542s) screen uh or with like the hot keys? [25:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1542s) screen uh or with like the hot keys? Like it's it's a lot. And if and [25:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1546s) Like it's it's a lot. And if and [25:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1546s) Like it's it's a lot. And if and typically computer use has been this [25:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1548s) typically computer use has been this [25:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1548s) typically computer use has been this benchmark where you think of oh only the [25:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1550s) benchmark where you think of oh only the [25:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1550s) benchmark where you think of oh only the smartest frontier reasoning models can [25:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1553s) smartest frontier reasoning models can [25:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1553s) smartest frontier reasoning models can actually figure this out. Um well maybe [25:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1556s) actually figure this out. Um well maybe [25:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1556s) actually figure this out. Um well maybe not like maybe actually if you could [25:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1558s) not like maybe actually if you could [25:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1558s) not like maybe actually if you could just enumerate like all of the possible [26:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1561s) just enumerate like all of the possible [26:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1561s) just enumerate like all of the possible options to you um you could navigate it [26:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1565s) options to you um you could navigate it [26:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1565s) options to you um you could navigate it with uh very quickly um and and we're [26:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1568s) with uh very quickly um and and we're [26:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1568s) with uh very quickly um and and we're going to get towards real time machine [26:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1569s) going to get towards real time machine [26:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1570s) going to get towards real time machine learning examples in a little bit. Okay. [26:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1572s) learning examples in a little bit. Okay. [26:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1572s) learning examples in a little bit. Okay. So it also must compose generatively. [26:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1575s) So it also must compose generatively. [26:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1576s) So it also must compose generatively. Okay. So there is a weakness um with [26:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1580s) Okay. So there is a weakness um with [26:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1580s) Okay. So there is a weakness um with decisions only. And one way that you [26:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1583s) decisions only. And one way that you [26:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1584s) decisions only. And one way that you might think about this is [26:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1586s) might think about this is [26:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1586s) might think about this is um like the the symmetry. So so um not [26:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1592s) um like the the symmetry. So so um not [26:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1592s) um like the the symmetry. So so um not just the symmetry but like the the [26:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1594s) just the symmetry but like the the [26:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1594s) just the symmetry but like the the predefin [26:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1595s) predefin [26:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1595s) predefin of the tree. So like in these examples [26:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1598s) of the tree. So like in these examples [26:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1598s) of the tree. So like in these examples of like the having search or this like [26:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1600s) of like the having search or this like [26:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1600s) of like the having search or this like hierarchical search. So like here I'm [26:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1602s) hierarchical search. So like here I'm [26:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1602s) hierarchical search. So like here I'm calling it outline walk. It's a similar [26:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1605s) calling it outline walk. It's a similar [26:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1605s) calling it outline walk. It's a similar example. You find the title, you find [26:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1606s) example. You find the title, you find [26:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1606s) example. You find the title, you find the um the H1 heading, you find the [26:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1609s) the um the H1 heading, you find the [26:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1609s) the um the H1 heading, you find the subsection heading, you find the [26:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1611s) subsection heading, you find the [26:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1611s) subsection heading, you find the paragraph, you find the sentence, you [26:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1613s) paragraph, you find the sentence, you [26:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1614s) paragraph, you find the sentence, you find the word, right? Um so at each one [26:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1617s) find the word, right? Um so at each one [26:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1617s) find the word, right? Um so at each one of those those layers of abstraction, [26:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1619s) of those those layers of abstraction, [26:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1619s) of those those layers of abstraction, you can call another one. So that's not [27:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1622s) you can call another one. So that's not [27:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1622s) you can call another one. So that's not like super symmetric like this like pure [27:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1624s) like super symmetric like this like pure [27:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1624s) like super symmetric like this like pure binary search um having one but the [27:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1628s) binary search um having one but the [27:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1628s) binary search um having one but the paths are are essentially like [27:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1631s) paths are are essentially like [27:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1631s) paths are are essentially like predefined um or need to be [27:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1634s) predefined um or need to be [27:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1634s) predefined um or need to be programmatically defined. Um and what's [27:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1637s) programmatically defined. Um and what's [27:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1637s) programmatically defined. Um and what's very interesting about then when you can [27:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1640s) very interesting about then when you can [27:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1640s) very interesting about then when you can um now have a generative element which [27:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1643s) um now have a generative element which [27:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1643s) um now have a generative element which I'm showing as like the square here. So [27:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1645s) I'm showing as like the square here. So [27:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1645s) I'm showing as like the square here. So like in this previous one it was like [27:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1647s) like in this previous one it was like [27:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1647s) like in this previous one it was like the generative ones diverge the um the [27:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1651s) the generative ones diverge the um the [27:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1651s) the generative ones diverge the um the decision models converge generative ones [27:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1654s) decision models converge generative ones [27:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1654s) decision models converge generative ones diverge decision models converge it's [27:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1656s) diverge decision models converge it's [27:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1656s) diverge decision models converge it's kind of an animation similar idea here [27:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1658s) kind of an animation similar idea here [27:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1658s) kind of an animation similar idea here now you can start to have so these are [27:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1660s) now you can start to have so these are [27:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1660s) now you can start to have so these are actual real call trees from the demo I'm [27:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1663s) actual real call trees from the demo I'm [27:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1663s) actual real call trees from the demo I'm about to show you of the smart house you [27:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1665s) about to show you of the smart house you [27:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1665s) about to show you of the smart house you have these much more organic shapes and [27:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1667s) have these much more organic shapes and [27:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1667s) have these much more organic shapes and these much more like emergent graphs and [27:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1668s) these much more like emergent graphs and [27:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1668s) these much more like emergent graphs and I think there's this um I have this idea [27:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1672s) I think there's this um I have this idea [27:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1672s) I think there's this um I have this idea that like that pairing the yin and yang [27:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1674s) that like that pairing the yin and yang [27:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1674s) that like that pairing the yin and yang of the decision model and the generative [27:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1676s) of the decision model and the generative [27:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1676s) of the decision model and the generative model itself is kind of this very [27:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1679s) model itself is kind of this very [27:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1679s) model itself is kind of this very powerful automaton, right? That can kind [28:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1682s) powerful automaton, right? That can kind [28:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1682s) powerful automaton, right? That can kind of ratchet through the system and and [28:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1686s) of ratchet through the system and and [28:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1686s) of ratchet through the system and and generate emergent behaviormer complex [28:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1689s) generate emergent behaviormer complex [28:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1689s) generate emergent behaviormer complex systems uh from the pairing of these [28:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1690s) systems uh from the pairing of these [28:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1690s) systems uh from the pairing of these things. And you can get us some sense of [28:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1692s) things. And you can get us some sense of [28:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1692s) things. And you can get us some sense of that from from the the different [28:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1695s) that from from the the different [28:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1695s) that from from the the different geometries of these graphs that you're [28:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1697s) geometries of these graphs that you're [28:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1697s) geometries of these graphs that you're looking at without seeing the actual [28:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1699s) looking at without seeing the actual [28:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1699s) looking at without seeing the actual behavior. Um, Typesafe had a really cool [28:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1703s) behavior. Um, Typesafe had a really cool [28:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1703s) behavior. Um, Typesafe had a really cool uh cookbook uh that they also launched [28:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1705s) uh cookbook uh that they also launched [28:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1706s) uh cookbook uh that they also launched uh with or maybe it was called a um I [28:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1708s) uh with or maybe it was called a um I [28:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1708s) uh with or maybe it was called a um I don't know demo or whatever um of a of a [28:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1711s) don't know demo or whatever um of a of a [28:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1711s) don't know demo or whatever um of a of a smart home. And so the the key thing [28:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1715s) smart home. And so the the key thing [28:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1715s) smart home. And so the the key thing from the smart home demo was that that [28:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1718s) from the smart home demo was that that [28:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1718s) from the smart home demo was that that they showed was when the problem cannot [28:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1721s) they showed was when the problem cannot [28:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1721s) they showed was when the problem cannot be solved purely with a decision then it [28:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1724s) be solved purely with a decision then it [28:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1724s) be solved purely with a decision then it can go to an LM to do generative [28:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1727s) can go to an LM to do generative [28:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1727s) can go to an LM to do generative element. So what I've done is adapt the [28:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1730s) element. So what I've done is adapt the [28:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1730s) element. So what I've done is adapt the smart home demo, make it uh much many [28:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1733s) smart home demo, make it uh much many [28:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1733s) smart home demo, make it uh much many more rooms in the home, many more [28:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1735s) more rooms in the home, many more [28:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1735s) more rooms in the home, many more appliances and things like that to show [28:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1737s) appliances and things like that to show [28:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1737s) appliances and things like that to show how the recursive decision model can um [29:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1741s) how the recursive decision model can um [29:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1741s) how the recursive decision model can um can interact with the home. You can [29:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1743s) can interact with the home. You can [29:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1743s) can interact with the home. You can express a very high level thing like [29:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1744s) express a very high level thing like [29:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1744s) express a very high level thing like we're off to bed and then you can ask [29:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1747s) we're off to bed and then you can ask [29:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1747s) we're off to bed and then you can ask that and it will do all of the substeps [29:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1749s) that and it will do all of the substeps [29:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1750s) that and it will do all of the substeps very very quickly to get you to door [29:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1752s) very very quickly to get you to door [29:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1752s) very very quickly to get you to door locked and lamped on uh lamp on [29:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1754s) locked and lamped on uh lamp on [29:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1754s) locked and lamped on uh lamp on everything off. And then you can have um [29:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1758s) everything off. And then you can have um [29:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1758s) everything off. And then you can have um these more like oblique type things that [29:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1760s) these more like oblique type things that [29:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1760s) these more like oblique type things that it's learning how to do over time. Let's [29:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1762s) it's learning how to do over time. Let's [29:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1762s) it's learning how to do over time. Let's watch a film. Um and then we will ask [29:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1767s) watch a film. Um and then we will ask [29:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1767s) watch a film. Um and then we will ask now it's going to uh like work through [29:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1769s) now it's going to uh like work through [29:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1769s) now it's going to uh like work through all those things. In order for the [29:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1773s) all those things. In order for the [29:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1773s) all those things. In order for the decision model and the language model to [29:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1775s) decision model and the language model to [29:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1775s) decision model and the language model to talk to each other, we need some kind of [29:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1777s) talk to each other, we need some kind of [29:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1778s) talk to each other, we need some kind of shared interface. And so this is um what [29:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1782s) shared interface. And so this is um what [29:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1782s) shared interface. And so this is um what I'm calling the map. And so again, the [29:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1786s) I'm calling the map. And so again, the [29:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1786s) I'm calling the map. And so again, the decision model can only point um it can [29:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1788s) decision model can only point um it can [29:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1788s) decision model can only point um it can judge various things that were given to [29:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1791s) judge various things that were given to [29:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1791s) judge various things that were given to it. And LMS in their sort of most raw [29:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1795s) it. And LMS in their sort of most raw [29:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1795s) it. And LMS in their sort of most raw form have no um structure to their [29:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1797s) form have no um structure to their [29:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1798s) form have no um structure to their output. It's perfectly open-ended stream [30:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1800s) output. It's perfectly open-ended stream [30:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1800s) output. It's perfectly open-ended stream of tokens, right? And so what we need is [30:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1803s) of tokens, right? And so what we need is [30:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1803s) of tokens, right? And so what we need is to kind of have a shared interface. And [30:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1806s) to kind of have a shared interface. And [30:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1806s) to kind of have a shared interface. And so I I call this the map. Um, and the um [30:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1811s) so I I call this the map. Um, and the um [30:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1811s) so I I call this the map. Um, and the um the map is um you could think of it as a [30:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1817s) the map is um you could think of it as a [30:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1817s) the map is um you could think of it as a um a a set of like pointers in the sense [30:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1821s) um a a set of like pointers in the sense [30:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1821s) um a a set of like pointers in the sense that it u could literally have like a [30:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1823s) that it u could literally have like a [30:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1823s) that it u could literally have like a name like TV on that has a very simple [30:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1826s) name like TV on that has a very simple [30:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1826s) name like TV on that has a very simple description and then some function um or [30:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1829s) description and then some function um or [30:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1829s) description and then some function um or in this case it's like flip flipping [30:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1830s) in this case it's like flip flipping [30:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1830s) in this case it's like flip flipping this boolean to be on, right? it um it [30:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1834s) this boolean to be on, right? it um it [30:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1834s) this boolean to be on, right? it um it might have a much more complicated [30:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1837s) might have a much more complicated [30:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1837s) might have a much more complicated um function underneath it like um this [30:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1841s) um function underneath it like um this [30:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1841s) um function underneath it like um this sort of like uh loop here. Um but the [30:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1846s) sort of like uh loop here. Um but the [30:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1846s) sort of like uh loop here. Um but the key idea is that the the um decision [30:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1851s) key idea is that the the um decision [30:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1851s) key idea is that the the um decision model can decide from the map which [30:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1855s) model can decide from the map which [30:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1855s) model can decide from the map which things are the right thing. It doesn't [30:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1857s) things are the right thing. It doesn't [30:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1857s) things are the right thing. It doesn't act it so it can decide lights on the [31:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1860s) act it so it can decide lights on the [31:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1860s) act it so it can decide lights on the code underneath it can implement the [31:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1863s) code underneath it can implement the [31:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1864s) code underneath it can implement the light can implement the lambda function [31:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1867s) light can implement the lambda function [31:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1867s) light can implement the lambda function lights on. Okay. The LLM now has a [31:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1871s) lights on. Okay. The LLM now has a [31:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1871s) lights on. Okay. The LLM now has a structured interface to add and edit and [31:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1874s) structured interface to add and edit and [31:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1874s) structured interface to add and edit and modify the map. So the LLM can basically [31:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1877s) modify the map. So the LLM can basically [31:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1877s) modify the map. So the LLM can basically um be learning from new user behavior [31:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1880s) um be learning from new user behavior [31:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1880s) um be learning from new user behavior that okay this movie time is now the [31:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1883s) that okay this movie time is now the [31:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1883s) that okay this movie time is now the combination of turning on the TV and [31:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1886s) combination of turning on the TV and [31:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1886s) combination of turning on the TV and turning off the lights. Um and so now [31:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1889s) turning off the lights. Um and so now [31:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1889s) turning off the lights. Um and so now there is this shared interface shared [31:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1892s) there is this shared interface shared [31:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1892s) there is this shared interface shared substrate um that both can talk to. I'm [31:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1896s) substrate um that both can talk to. I'm [31:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1896s) substrate um that both can talk to. I'm not going to go into it here because it [31:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1897s) not going to go into it here because it [31:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1897s) not going to go into it here because it gets a little bit more complicated. I [31:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1899s) gets a little bit more complicated. I [31:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1899s) gets a little bit more complicated. I also want to point out that the map [31:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1901s) also want to point out that the map [31:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1901s) also want to point out that the map itself can be uh multi-dimensional. So [31:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1905s) itself can be uh multi-dimensional. So [31:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1905s) itself can be uh multi-dimensional. So it can essentially be representing uh [31:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1908s) it can essentially be representing uh [31:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1908s) it can essentially be representing uh embeddings and and for example um like [31:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1910s) embeddings and and for example um like [31:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1910s) embeddings and and for example um like Kieran Clawson from every had a really [31:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1913s) Kieran Clawson from every had a really [31:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1913s) Kieran Clawson from every had a really interesting idea of using um using Jev [31:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1916s) interesting idea of using um using Jev [31:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1916s) interesting idea of using um using Jev to create lowdimensional but highly [31:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1919s) to create lowdimensional but highly [31:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1919s) to create lowdimensional but highly semantically meaningful embeddings. So [32:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1921s) semantically meaningful embeddings. So [32:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1921s) semantically meaningful embeddings. So so this map that I built into the RDM [32:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1923s) so this map that I built into the RDM [32:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1924s) so this map that I built into the RDM codebase can can handle that as well. [32:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1927s) codebase can can handle that as well. [32:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1928s) codebase can can handle that as well. The last uh property or the last um [32:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1930s) The last uh property or the last um [32:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1930s) The last uh property or the last um primitive or behavior that I have [32:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1933s) primitive or behavior that I have [32:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1933s) primitive or behavior that I have included in my recursive decision models [32:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1936s) included in my recursive decision models [32:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1936s) included in my recursive decision models is uh what I'm calling the arbiter. And [32:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1940s) is uh what I'm calling the arbiter. And [32:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1940s) is uh what I'm calling the arbiter. And here's why this is important, which is [32:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1943s) here's why this is important, which is [32:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1943s) here's why this is important, which is Jev is fast and cheap. There's decision [32:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1945s) Jev is fast and cheap. There's decision [32:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1945s) Jev is fast and cheap. There's decision models that can run on your phone, on a [32:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1949s) models that can run on your phone, on a [32:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1949s) models that can run on your phone, on a crappy 10-year-old laptop, on um [32:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1951s) crappy 10-year-old laptop, on um [32:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1951s) crappy 10-year-old laptop, on um obviously you can get a better one on [32:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1953s) obviously you can get a better one on [32:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1953s) obviously you can get a better one on your new um your new MacBook Pro, [32:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1956s) your new um your new MacBook Pro, [32:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1956s) your new um your new MacBook Pro, whatever. Um, these things are very [32:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1959s) whatever. Um, these things are very [32:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1959s) whatever. Um, these things are very fast. They're very inexpensive either [32:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1961s) fast. They're very inexpensive either [32:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1961s) fast. They're very inexpensive either from a compute perspective or a dollars [32:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1962s) from a compute perspective or a dollars [32:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1962s) from a compute perspective or a dollars perspective. [32:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1964s) perspective. [32:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1964s) perspective. And that means that we can run many at [32:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1966s) And that means that we can run many at [32:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1966s) And that means that we can run many at once. We can we can actually [32:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1967s) once. We can we can actually [32:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1967s) once. We can we can actually speculatively [32:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1969s) speculatively [32:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1969s) speculatively run a bunch of them when we're uncertain [32:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1971s) run a bunch of them when we're uncertain [32:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1971s) run a bunch of them when we're uncertain all in parallel. And now the thing is [32:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1976s) all in parallel. And now the thing is [32:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1976s) all in parallel. And now the thing is that we need to if we just let all of [32:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1979s) that we need to if we just let all of [32:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1979s) that we need to if we just let all of those act on the state [33:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1982s) those act on the state [33:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1982s) those act on the state um we're going to have a lot of [33:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1983s) um we're going to have a lot of [33:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1984s) um we're going to have a lot of collisions. We're going to have a lot of [33:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1985s) collisions. We're going to have a lot of [33:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1985s) collisions. We're going to have a lot of challenge. So um an example from how [33:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1989s) challenge. So um an example from how [33:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1989s) challenge. So um an example from how your CPU is working would be like that [33:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1993s) your CPU is working would be like that [33:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1993s) your CPU is working would be like that there can be many things reading in [33:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1995s) there can be many things reading in [33:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1995s) there can be many things reading in parallel concurrently [33:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1997s) parallel concurrently [33:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1997s) parallel concurrently um but now we need some kind of lock [33:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1999s) um but now we need some kind of lock [33:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=1999s) um but now we need some kind of lock when we write because we don't want to [33:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2003s) when we write because we don't want to [33:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2003s) when we write because we don't want to uh collide have colliding uh writers. [33:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2006s) uh collide have colliding uh writers. [33:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2006s) uh collide have colliding uh writers. There's also an interesting idea also [33:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2008s) There's also an interesting idea also [33:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2008s) There's also an interesting idea also from Alex Zang um at MIT in Omar's [33:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2011s) from Alex Zang um at MIT in Omar's [33:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2011s) from Alex Zang um at MIT in Omar's group, the inventor of RLMs about um [33:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2015s) group, the inventor of RLMs about um [33:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2015s) group, the inventor of RLMs about um speculative tool calling, speculative [33:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2017s) speculative tool calling, speculative [33:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2017s) speculative tool calling, speculative programmatic tool calling for agents. Uh [33:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2019s) programmatic tool calling for agents. Uh [33:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2019s) programmatic tool calling for agents. Uh and this is um you know informed my [33:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2022s) and this is um you know informed my [33:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2022s) and this is um you know informed my thinking here. But here's here's this [33:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2024s) thinking here. But here's here's this [33:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2024s) thinking here. But here's here's this idea applied to audio which is like we [33:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2027s) idea applied to audio which is like we [33:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2027s) idea applied to audio which is like we have uh I'm speaking uh these sentences. [33:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2031s) have uh I'm speaking uh these sentences. [33:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2031s) have uh I'm speaking uh these sentences. we're having we're making decisions [33:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2033s) we're having we're making decisions [33:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2033s) we're having we're making decisions where I haven't even actually like [33:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2035s) where I haven't even actually like [33:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2035s) where I haven't even actually like finished the sentence. And so we [33:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2038s) finished the sentence. And so we [33:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2038s) finished the sentence. And so we essentially have um at various times the [34:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2041s) essentially have um at various times the [34:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2041s) essentially have um at various times the arbiter is like deciding okay with the [34:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2044s) arbiter is like deciding okay with the [34:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2044s) arbiter is like deciding okay with the given information that I have across all [34:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2047s) given information that I have across all [34:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2047s) given information that I have across all of these different channels what is the [34:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2049s) of these different channels what is the [34:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2049s) of these different channels what is the decision that I'm going to allow [34:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2051s) decision that I'm going to allow [34:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2051s) decision that I'm going to allow through. So you could you could imagine [34:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2052s) through. So you could you could imagine [34:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2052s) through. So you could you could imagine that like each one of these uh lanes [34:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2056s) that like each one of these uh lanes [34:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2056s) that like each one of these uh lanes might like want to um might be voting [34:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2059s) might like want to um might be voting [34:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2059s) might like want to um might be voting for like a certain action and the [34:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2061s) for like a certain action and the [34:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2061s) for like a certain action and the arbiter's job is to figure out like [34:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2063s) arbiter's job is to figure out like [34:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2064s) arbiter's job is to figure out like where it has the confidence to actually [34:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2066s) where it has the confidence to actually [34:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2066s) where it has the confidence to actually like allow that action to change the [34:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2068s) like allow that action to change the [34:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2068s) like allow that action to change the state because again if we let them all [34:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2071s) state because again if we let them all [34:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2071s) state because again if we let them all uh act we could have a lot of [34:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2072s) uh act we could have a lot of [34:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2072s) uh act we could have a lot of collisions. Let's say just a simple [34:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2074s) collisions. Let's say just a simple [34:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2074s) collisions. Let's say just a simple example, right? There's two there's two [34:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2076s) example, right? There's two there's two [34:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2076s) example, right? There's two there's two that are making uh that they're [34:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2078s) that are making uh that they're [34:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2078s) that are making uh that they're different models. They're making the [34:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2080s) different models. They're making the [34:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2080s) different models. They're making the same kind of decision um and one of them [34:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2084s) same kind of decision um and one of them [34:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2084s) same kind of decision um and one of them uh goes a little bit faster, changes the [34:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2087s) uh goes a little bit faster, changes the [34:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2087s) uh goes a little bit faster, changes the state um before the second one makes a [34:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2090s) state um before the second one makes a [34:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2090s) state um before the second one makes a different decision that would have [34:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2091s) different decision that would have [34:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2091s) different decision that would have changed the state the other way. You can [34:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2093s) changed the state the other way. You can [34:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2093s) changed the state the other way. You can imagine all kinds of problems or [34:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2094s) imagine all kinds of problems or [34:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2094s) imagine all kinds of problems or collisions that could happen when we do [34:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2096s) collisions that could happen when we do [34:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2096s) collisions that could happen when we do that. So this this pattern of the [34:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2098s) that. So this this pattern of the [34:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2098s) that. So this this pattern of the arbiter is what I've developed um as a [35:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2100s) arbiter is what I've developed um as a [35:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2100s) arbiter is what I've developed um as a way of like [35:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2102s) way of like [35:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2102s) way of like um um it's like a gate for what can [35:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2107s) um um it's like a gate for what can [35:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2107s) um um it's like a gate for what can actually now um act on the system change [35:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2109s) actually now um act on the system change [35:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2109s) actually now um act on the system change the state of the system and it's very [35:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2112s) the state of the system and it's very [35:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2112s) the state of the system and it's very useful when you want to run many of [35:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2114s) useful when you want to run many of [35:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2114s) useful when you want to run many of these things in parallel. Okay. [35:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2117s) these things in parallel. Okay. [35:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2117s) these things in parallel. Okay. So I'd love you to try it. Um my very [35:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2120s) So I'd love you to try it. Um my very [35:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2120s) So I'd love you to try it. Um my very first implementation of RDM was as a [35:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2123s) first implementation of RDM was as a [35:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2123s) first implementation of RDM was as a DSPI module. Um this is in part because [35:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2127s) DSPI module. Um this is in part because [35:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2127s) DSPI module. Um this is in part because the DSPI uh team in particular the the [35:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2130s) the DSPI uh team in particular the the [35:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2130s) the DSPI uh team in particular the the folks at Compound AI uh were very very [35:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2133s) folks at Compound AI uh were very very [35:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2133s) folks at Compound AI uh were very very quick to add support for system one API [35:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2136s) quick to add support for system one API [35:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2136s) quick to add support for system one API calls within um DSPI. Um and secondly [35:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2141s) calls within um DSPI. Um and secondly [35:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2142s) calls within um DSPI. Um and secondly DSPI is a very um opinionated way of [35:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2145s) DSPI is a very um opinionated way of [35:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2145s) DSPI is a very um opinionated way of working. It's very structured way of [35:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2147s) working. It's very structured way of [35:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2147s) working. It's very structured way of working and I found this to be a very [35:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2149s) working and I found this to be a very [35:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2149s) working and I found this to be a very useful playground for um essentially [35:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2153s) useful playground for um essentially [35:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2153s) useful playground for um essentially ensuring that the ideas of recursive [35:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2157s) ensuring that the ideas of recursive [35:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2157s) ensuring that the ideas of recursive decision models um were structured in [35:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2159s) decision models um were structured in [35:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2159s) decision models um were structured in the right way that um the modules were [36:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2162s) the right way that um the modules were [36:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2162s) the right way that um the modules were behaving uh in the right way. Uh I have [36:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2165s) behaving uh in the right way. Uh I have [36:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2165s) behaving uh in the right way. Uh I have also uh made a standalone rust package [36:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2169s) also uh made a standalone rust package [36:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2169s) also uh made a standalone rust package um RDM which is great when uh you want [36:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2172s) um RDM which is great when uh you want [36:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2172s) um RDM which is great when uh you want to bring uh very very fast and portable [36:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2176s) to bring uh very very fast and portable [36:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2176s) to bring uh very very fast and portable um RDMs to all kinds of systems. [36:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2180s) um RDMs to all kinds of systems. [36:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2180s) um RDMs to all kinds of systems. So yeah, check it out um in either my [36:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2183s) So yeah, check it out um in either my [36:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2183s) So yeah, check it out um in either my DSp fork uh if you're a DSP nerd or um [36:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2185s) DSp fork uh if you're a DSP nerd or um [36:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2185s) DSp fork uh if you're a DSP nerd or um the RDM package if you want to play [36:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2188s) the RDM package if you want to play [36:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2188s) the RDM package if you want to play around with Rust. [36:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2190s) around with Rust. [36:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2190s) around with Rust. Okay, so this is where we move into [36:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2195s) Okay, so this is where we move into [36:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2195s) Okay, so this is where we move into operating system one. And [36:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2199s) operating system one. And [36:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2199s) operating system one. And I mean this uh with both of these [36:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2203s) I mean this uh with both of these [36:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2203s) I mean this uh with both of these phrasings, right? Um this is an [36:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2206s) phrasings, right? Um this is an [36:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2206s) phrasings, right? Um this is an operating system one. It's operating [36:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2208s) operating system one. It's operating [36:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2208s) operating system one. It's operating system one. And we're also learning how [36:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2211s) system one. And we're also learning how [36:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2211s) system one. And we're also learning how to operate system one models, right? So [36:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2214s) to operate system one models, right? So [36:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2214s) to operate system one models, right? So that's sort of the play on words here [36:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2217s) that's sort of the play on words here [36:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2217s) that's sort of the play on words here with operating system one. And [37:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2222s) what's really important to me um [37:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2225s) what's really important to me um [37:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2225s) what's really important to me um personally [37:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2227s) personally [37:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2227s) personally is this physical imshment that we have [37:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2232s) is this physical imshment that we have [37:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2232s) is this physical imshment that we have with with our devices, right? Um I've [37:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2236s) with with our devices, right? Um I've [37:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2236s) with with our devices, right? Um I've got like tech neck, right? Like my [37:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2239s) got like tech neck, right? Like my [37:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2239s) got like tech neck, right? Like my head's always down. I've got this like [37:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2240s) head's always down. I've got this like [37:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2240s) head's always down. I've got this like crazy like shoulder [37:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2243s) crazy like shoulder [37:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2243s) crazy like shoulder problem and tension from mousing all the [37:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2245s) problem and tension from mousing all the [37:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2245s) problem and tension from mousing all the time. When I was in grad school with [37:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2248s) time. When I was in grad school with [37:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2248s) time. When I was in grad school with really stressful deadlines, I was having [37:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2250s) really stressful deadlines, I was having [37:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2250s) really stressful deadlines, I was having all kinds of um repetitive stress issues [37:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2254s) all kinds of um repetitive stress issues [37:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2254s) all kinds of um repetitive stress issues and I've got every crazy vertical mouse [37:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2257s) and I've got every crazy vertical mouse [37:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2257s) and I've got every crazy vertical mouse and track ball and ergonomic keyboard [37:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2260s) and track ball and ergonomic keyboard [37:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2260s) and track ball and ergonomic keyboard that you could possibly imagine. But the [37:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2266s) that you could possibly imagine. But the [37:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2266s) that you could possibly imagine. But the like there is great um power and [37:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2270s) like there is great um power and [37:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2270s) like there is great um power and economic benefit in working with [37:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2272s) economic benefit in working with [37:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2272s) economic benefit in working with computers, in working with devices, in [37:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2274s) computers, in working with devices, in [37:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2274s) computers, in working with devices, in working with um our phones and our [37:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2276s) working with um our phones and our [37:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2276s) working with um our phones and our tablets in terms of being interconnected [37:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2278s) tablets in terms of being interconnected [37:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2278s) tablets in terms of being interconnected and be in terms of being able to just do [38:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2281s) and be in terms of being able to just do [38:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2281s) and be in terms of being able to just do our normal jobs. And obviously anyone [38:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2283s) our normal jobs. And obviously anyone [38:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2283s) our normal jobs. And obviously anyone who is an AI engineer is going to need a [38:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2285s) who is an AI engineer is going to need a [38:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2285s) who is an AI engineer is going to need a computer to do their job. And at the [38:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2288s) computer to do their job. And at the [38:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2288s) computer to do their job. And at the same time like there is just all of this [38:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2292s) same time like there is just all of this [38:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2292s) same time like there is just all of this um I would say unnecessary suffering [38:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2296s) um I would say unnecessary suffering [38:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2296s) um I would say unnecessary suffering that is caused by the coupling of our [38:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2299s) that is caused by the coupling of our [38:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2299s) that is caused by the coupling of our physicality like the coupling of our [38:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2302s) physicality like the coupling of our [38:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2302s) physicality like the coupling of our physical bodies to the physical computer [38:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2305s) physical bodies to the physical computer [38:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2305s) physical bodies to the physical computer like through the traditional interfaces [38:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2309s) like through the traditional interfaces [38:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2309s) like through the traditional interfaces that have made their way in as the [38:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2312s) that have made their way in as the [38:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2312s) that have made their way in as the popular ones of society the cordy [38:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2314s) popular ones of society the cordy [38:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2314s) popular ones of society the cordy keyboard the mouse the touchcreen etc. [38:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2317s) keyboard the mouse the touchcreen etc. [38:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2318s) keyboard the mouse the touchcreen etc. Okay, so some of you might rec recognize [38:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2321s) Okay, so some of you might rec recognize [38:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2321s) Okay, so some of you might rec recognize this um this sequence. Uh I used it in [38:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2324s) this um this sequence. Uh I used it in [38:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2324s) this um this sequence. Uh I used it in my recursive coding agents talk and it's [38:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2327s) my recursive coding agents talk and it's [38:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2327s) my recursive coding agents talk and it's also the back of a t-shirt uh from the [38:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2330s) also the back of a t-shirt uh from the [38:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2330s) also the back of a t-shirt uh from the AI engineer code event in New York a [38:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2332s) AI engineer code event in New York a [38:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2332s) AI engineer code event in New York a couple years ago. And so so here we go. [38:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2334s) couple years ago. And so so here we go. [38:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2334s) couple years ago. And so so here we go. We've got the evolution of of the AI [38:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2336s) We've got the evolution of of the AI [38:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2336s) We've got the evolution of of the AI engineer uh you know typing on the [38:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2338s) engineer uh you know typing on the [38:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2338s) engineer uh you know typing on the laptop. Not great posture, you know. Now [39:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2342s) laptop. Not great posture, you know. Now [39:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2342s) laptop. Not great posture, you know. Now here it's got like the AI uh sparkle [39:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2344s) here it's got like the AI uh sparkle [39:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2344s) here it's got like the AI uh sparkle features. So we're we're vibe code in [39:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2347s) features. So we're we're vibe code in [39:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2347s) features. So we're we're vibe code in here maybe. Um you know in the third one [39:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2350s) here maybe. Um you know in the third one [39:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2350s) here maybe. Um you know in the third one you can see there's no more computer [39:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2352s) you can see there's no more computer [39:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2352s) you can see there's no more computer anymore. So now we're just like talking [39:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2354s) anymore. So now we're just like talking [39:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2354s) anymore. So now we're just like talking to this like orb. Um that's pretty cool. [39:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2358s) to this like orb. Um that's pretty cool. [39:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2358s) to this like orb. Um that's pretty cool. And uh then in the this ultimate like [39:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2360s) And uh then in the this ultimate like [39:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2360s) And uh then in the this ultimate like state of enlightenment, right? We just [39:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2362s) state of enlightenment, right? We just [39:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2362s) state of enlightenment, right? We just like think and and u things are [39:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2366s) like think and and u things are [39:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2366s) like think and and u things are manifested uh in in our world or I I [39:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2370s) manifested uh in in our world or I I [39:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2370s) manifested uh in in our world or I I don't even know what kind of uh [39:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2372s) don't even know what kind of uh [39:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2372s) don't even know what kind of uh wavelengths are coming out of this guy's [39:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2374s) wavelengths are coming out of this guy's [39:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2374s) wavelengths are coming out of this guy's head. So, I'm going here. I'd like to [39:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2377s) head. So, I'm going here. I'd like to [39:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2377s) head. So, I'm going here. I'd like to get to the the right um making some [39:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2380s) get to the the right um making some [39:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2380s) get to the the right um making some progress, but I think right now with uh [39:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2383s) progress, but I think right now with uh [39:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2384s) progress, but I think right now with uh with operating system one and my [39:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2387s) with operating system one and my [39:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2387s) with operating system one and my recursive decision models, I'm I'm much [39:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2389s) recursive decision models, I'm I'm much [39:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2389s) recursive decision models, I'm I'm much closer to to the audio piece. Um and so [39:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2393s) closer to to the audio piece. Um and so [39:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2393s) closer to to the audio piece. Um and so I want to um I want to sort of take a um [39:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2397s) I want to um I want to sort of take a um [39:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2397s) I want to um I want to sort of take a um to go back in time a little bit. Um, [40:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2400s) to go back in time a little bit. Um, [40:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2400s) to go back in time a little bit. Um, this these videos are um I'm getting [40:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2404s) this these videos are um I'm getting [40:04](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2404s) this these videos are um I'm getting old. They're they're coming up on uh not [40:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2407s) old. They're they're coming up on uh not [40:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2407s) old. They're they're coming up on uh not quite 20 years old. It's like 17 18 year [40:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2410s) quite 20 years old. It's like 17 18 year [40:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2410s) quite 20 years old. It's like 17 18 year old uh videos. And I want to show you [40:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2413s) old uh videos. And I want to show you [40:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2413s) old uh videos. And I want to show you some experiments um in what I call [40:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2416s) some experiments um in what I call [40:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2416s) some experiments um in what I call radical human computer interfacing [40:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2420s) radical human computer interfacing [40:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2420s) radical human computer interfacing um from some of my work in music. I'm [40:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2422s) um from some of my work in music. I'm [40:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2422s) um from some of my work in music. I'm just show these really quick and I'll [40:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2423s) just show these really quick and I'll [40:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2423s) just show these really quick and I'll explain a little bit what's going on. [40:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2424s) explain a little bit what's going on. [40:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2424s) explain a little bit what's going on. So, so over here. [40:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2431s) [music] [40:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2433s) So, so just watch here. I'm really [40:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2435s) So, so just watch here. I'm really [40:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2435s) So, so just watch here. I'm really setting this up. [40:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2437s) setting this up. [40:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2437s) setting this up. All right. So, we've got like um hitting [40:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2441s) All right. So, we've got like um hitting [40:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2441s) All right. So, we've got like um hitting these four different locations on the [40:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2443s) these four different locations on the [40:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2443s) these four different locations on the laptop and that is uh being classified [40:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2448s) laptop and that is uh being classified [40:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2448s) laptop and that is uh being classified uh to [40:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2450s) uh to [40:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2450s) uh to make a drum kit, right? Drum roll on by [40:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2452s) make a drum kit, right? Drum roll on by [40:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2452s) make a drum kit, right? Drum roll on by hitting our laptop. So, the laptops used [40:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2454s) hitting our laptop. So, the laptops used [40:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2454s) hitting our laptop. So, the laptops used to have um accelerometers built into [40:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2457s) to have um accelerometers built into [40:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2457s) to have um accelerometers built into them back when we're still have disc [40:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2459s) them back when we're still have disc [40:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2459s) them back when we're still have disc spinning disc hard drives. Um, this one [41:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2462s) spinning disc hard drives. Um, this one [41:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2462s) spinning disc hard drives. Um, this one I'm not going to play um so much of, but [41:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2465s) I'm not going to play um so much of, but [41:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2465s) I'm not going to play um so much of, but you can um uh that left one I'm like [41:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2468s) you can um uh that left one I'm like [41:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2468s) you can um uh that left one I'm like triggering the drum kit here. You can [41:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2471s) triggering the drum kit here. You can [41:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2471s) triggering the drum kit here. You can use the tilt of the um laptop to control [41:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2475s) use the tilt of the um laptop to control [41:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2475s) use the tilt of the um laptop to control different effects. I'm using the hit of [41:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2477s) different effects. I'm using the hit of [41:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2477s) different effects. I'm using the hit of this thing as a side chain compressor. [41:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2480s) this thing as a side chain compressor. [41:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2480s) this thing as a side chain compressor. Um, so it's like a ducking compressor [41:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2483s) Um, so it's like a ducking compressor [41:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2483s) Um, so it's like a ducking compressor every time I hit it. Uh, and this is a [41:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2485s) every time I hit it. Uh, and this is a [41:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2485s) every time I hit it. Uh, and this is a much longer video, so I'm not going to [41:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2487s) much longer video, so I'm not going to [41:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2487s) much longer video, so I'm not going to go into that, but this is um I want to [41:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2489s) go into that, but this is um I want to [41:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2489s) go into that, but this is um I want to bring this up um because it's personally [41:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2492s) bring this up um because it's personally [41:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2492s) bring this up um because it's personally meaningful to me. It's also the way that [41:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2494s) meaningful to me. It's also the way that [41:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2494s) meaningful to me. It's also the way that I actually learned um how to program and [41:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2497s) I actually learned um how to program and [41:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2497s) I actually learned um how to program and u the way that I learned machine [41:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2498s) u the way that I learned machine [41:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2498s) u the way that I learned machine learning. And this is through the [41:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2499s) learning. And this is through the [41:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2499s) learning. And this is through the Princeton Laptop Orchestra. Um it just [41:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2503s) Princeton Laptop Orchestra. Um it just [41:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2503s) Princeton Laptop Orchestra. Um it just celebrated its 20th anniversary uh this [41:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2506s) celebrated its 20th anniversary uh this [41:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2506s) celebrated its 20th anniversary uh this spring. [41:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2507s) spring. [41:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2507s) spring. And really the idea of the Princeton [41:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2510s) And really the idea of the Princeton [41:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2510s) And really the idea of the Princeton Laptop Orchestra was what would it what [41:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2512s) Laptop Orchestra was what would it what [41:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2512s) Laptop Orchestra was what would it what does it mean um to have the laptop [41:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2516s) does it mean um to have the laptop [41:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2516s) does it mean um to have the laptop itself be the instrument? So there were [41:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2519s) itself be the instrument? So there were [41:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2519s) itself be the instrument? So there were pieces we were performing where the [42:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2520s) pieces we were performing where the [42:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2520s) pieces we were performing where the trackpad was the instrument. There were [42:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2522s) trackpad was the instrument. There were [42:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2522s) trackpad was the instrument. There were pieces where the keyboard was the [42:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2523s) pieces where the keyboard was the [42:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2523s) pieces where the keyboard was the instrument. There were pieces where all [42:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2526s) instrument. There were pieces where all [42:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2526s) instrument. There were pieces where all the computers wereworked together on um [42:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2529s) the computers wereworked together on um [42:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2529s) the computers wereworked together on um Wi-Fi and like syncing super super um [42:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2533s) Wi-Fi and like syncing super super um [42:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2533s) Wi-Fi and like syncing super super um quick like low latency clock times to [42:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2536s) quick like low latency clock times to [42:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2536s) quick like low latency clock times to send messages back and forth to each [42:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2538s) send messages back and forth to each [42:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2538s) send messages back and forth to each other. And um I was able to work with [42:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2543s) other. And um I was able to work with [42:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2543s) other. And um I was able to work with this uh brand new cutting edge real-time [42:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2546s) this uh brand new cutting edge real-time [42:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2546s) this uh brand new cutting edge real-time machine learning framework called the [42:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2547s) machine learning framework called the [42:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2547s) machine learning framework called the Wenator developed by Rebecca Feebrink um [42:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2550s) Wenator developed by Rebecca Feebrink um [42:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2550s) Wenator developed by Rebecca Feebrink um to actually develop um and again this is [42:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2553s) to actually develop um and again this is [42:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2553s) to actually develop um and again this is like this is like 19 years ago 18 years [42:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2556s) like this is like 19 years ago 18 years [42:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2556s) like this is like 19 years ago 18 years ago um these um very very small um you [42:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2561s) ago um these um very very small um you [42:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2561s) ago um these um very very small um you know k nearest neighbors um like single [42:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2564s) know k nearest neighbors um like single [42:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2564s) know k nearest neighbors um like single layer neural nets um you know things [42:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2567s) layer neural nets um you know things [42:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2567s) layer neural nets um you know things that look much more like principal [42:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2569s) that look much more like principal [42:49](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2569s) that look much more like principal component analysis than like what we [42:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2571s) component analysis than like what we [42:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2571s) component analysis than like what we think of as like transformers, large [42:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2573s) think of as like transformers, large [42:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2573s) think of as like transformers, large language models. And and um and you can [42:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2577s) language models. And and um and you can [42:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2577s) language models. And and um and you can actually um do uh what you might call [43:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2580s) actually um do uh what you might call [43:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2580s) actually um do uh what you might call like on the-ly um machine learning to to [43:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2585s) like on the-ly um machine learning to to [43:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2585s) like on the-ly um machine learning to to um classify and map and do nonlinear [43:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2589s) um classify and map and do nonlinear [43:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2590s) um classify and map and do nonlinear interpolations. Um it was just s so much [43:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2593s) interpolations. Um it was just s so much [43:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2593s) interpolations. Um it was just s so much fun. So, I've been having a lot of fun [43:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2595s) fun. So, I've been having a lot of fun [43:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2595s) fun. So, I've been having a lot of fun with coming back to um that a little bit [43:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2598s) with coming back to um that a little bit [43:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2598s) with coming back to um that a little bit and kind of merging um my excitement [43:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2602s) and kind of merging um my excitement [43:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2602s) and kind of merging um my excitement about RLMs um my excitement about Jev, [43:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2605s) about RLMs um my excitement about Jev, [43:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2605s) about RLMs um my excitement about Jev, these decision models, this this [43:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2607s) these decision models, this this [43:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2607s) these decision models, this this invention of recursive decision models [43:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2610s) invention of recursive decision models [43:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2610s) invention of recursive decision models and now towards um this kind of personal [43:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2613s) and now towards um this kind of personal [43:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2613s) and now towards um this kind of personal goal of just like like I want to be able [43:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2615s) goal of just like like I want to be able [43:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2615s) goal of just like like I want to be able to get things done through my computer [43:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2619s) to get things done through my computer [43:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2619s) to get things done through my computer and uh work in this field without [43:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2623s) and uh work in this field without [43:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2624s) and uh work in this field without hurting, without like touching a [43:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2626s) hurting, without like touching a [43:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2626s) hurting, without like touching a keyboard all day, without touching a [43:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2627s) keyboard all day, without touching a [43:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2627s) keyboard all day, without touching a mouse all day. Um, and so, [43:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2631s) mouse all day. Um, and so, [43:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2631s) mouse all day. Um, and so, uh, I'm going to show this example. This [43:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2633s) uh, I'm going to show this example. This [43:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2633s) uh, I'm going to show this example. This is sort of a precursor to to OSO that I [43:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2636s) is sort of a precursor to to OSO that I [43:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2636s) is sort of a precursor to to OSO that I called earshot. And I want to set this [43:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2638s) called earshot. And I want to set this [43:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2638s) called earshot. And I want to set this up and explain it because it's actually [43:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2639s) up and explain it because it's actually [44:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2640s) up and explain it because it's actually a little bit confusing video to watch. [44:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2642s) a little bit confusing video to watch. [44:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2642s) a little bit confusing video to watch. Um, so you can think of this as this is [44:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2646s) Um, so you can think of this as this is [44:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2646s) Um, so you can think of this as this is a a recursive decision model. we have uh [44:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2651s) a a recursive decision model. we have uh [44:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2651s) a a recursive decision model. we have uh one layer of the system that is [44:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2652s) one layer of the system that is [44:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2652s) one layer of the system that is transcribing the audio and then choosing [44:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2657s) transcribing the audio and then choosing [44:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2657s) transcribing the audio and then choosing an action to take based on that [44:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2659s) an action to take based on that [44:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2659s) an action to take based on that transcribed word. And so the um here [44:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2662s) transcribed word. And so the um here [44:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2662s) transcribed word. And so the um here you're just going to see there's it's [44:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2664s) you're just going to see there's it's [44:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2664s) you're just going to see there's it's going to be cycling through my different [44:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2666s) going to be cycling through my different [44:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2666s) going to be cycling through my different workspaces one um through 8. Um and then [44:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2672s) workspaces one um through 8. Um and then [44:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2672s) workspaces one um through 8. Um and then uh while that's happening there is a [44:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2674s) uh while that's happening there is a [44:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2674s) uh while that's happening there is a very small um uh based in Chuck there's [44:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2678s) very small um uh based in Chuck there's [44:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2678s) very small um uh based in Chuck there's a new thing called Chai that's like the [44:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2679s) a new thing called Chai that's like the [44:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2679s) a new thing called Chai that's like the Chuck AI that's sort of the successor to [44:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2682s) Chuck AI that's sort of the successor to [44:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2682s) Chuck AI that's sort of the successor to the Wenator that I mentioned before very [44:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2684s) the Wenator that I mentioned before very [44:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2684s) the Wenator that I mentioned before very very small model model operating on a [44:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2687s) very small model model operating on a [44:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2687s) very small model model operating on a very very small slice of audio 40 [44:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2690s) very very small slice of audio 40 [44:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2690s) very very small slice of audio 40 milliseconds 80 milliseconds 100 [44:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2691s) milliseconds 80 milliseconds 100 [44:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2692s) milliseconds 80 milliseconds 100 milliseconds right and essentially [44:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2695s) milliseconds right and essentially [44:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2695s) milliseconds right and essentially trying to just map from some very crude [44:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2698s) trying to just map from some very crude [44:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2698s) trying to just map from some very crude bin [44:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2699s) bin [44:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2699s) bin of so the features are just like some [45:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2701s) of so the features are just like some [45:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2701s) of so the features are just like some bins of audio frequency bins of audio um [45:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2705s) bins of audio frequency bins of audio um [45:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2705s) bins of audio frequency bins of audio um which is the right one now this is [45:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2707s) which is the right one now this is [45:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2707s) which is the right one now this is supervised right so it need it's it [45:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2709s) supervised right so it need it's it [45:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2709s) supervised right so it need it's it knows nothing in the beginning and we [45:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2711s) knows nothing in the beginning and we [45:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2711s) knows nothing in the beginning and we can't trust it in the beginning so there [45:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2714s) can't trust it in the beginning so there [45:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2714s) can't trust it in the beginning so there is one layer on which there's like a [45:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2716s) is one layer on which there's like a [45:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2716s) is one layer on which there's like a teacher model that's teaching this small [45:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2718s) teacher model that's teaching this small [45:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2718s) teacher model that's teaching this small real-time model right and then there's a [45:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2721s) real-time model right and then there's a [45:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2721s) real-time model right and then there's a second layer on which the system is [45:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2724s) second layer on which the system is [45:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2724s) second layer on which the system is independently figuring out when it [45:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2726s) independently figuring out when it [45:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2726s) independently figuring out when it actually trusts the small model enough [45:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2730s) actually trusts the small model enough [45:30](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2730s) actually trusts the small model enough to follow it. So, we're going to watch [45:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2731s) to follow it. So, we're going to watch [45:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2732s) to follow it. So, we're going to watch it now. And what you're going to see is [45:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2734s) it now. And what you're going to see is [45:34](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2734s) it now. And what you're going to see is in the beginning there's a big delay [45:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2736s) in the beginning there's a big delay [45:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2736s) in the beginning there's a big delay between the word and this comes from [45:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2737s) between the word and this comes from [45:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2738s) between the word and this comes from Philip Glass's um uh knee play one part [45:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2741s) Philip Glass's um uh knee play one part [45:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2741s) Philip Glass's um uh knee play one part of uh Einstein on the Beach. And um so [45:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2745s) of uh Einstein on the Beach. And um so [45:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2745s) of uh Einstein on the Beach. And um so you're going to see that [music] it's [45:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2747s) you're going to see that [music] it's [45:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2747s) you're going to see that [music] it's going look up here for the 1 2 3 4. So [45:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2750s) going look up here for the 1 2 3 4. So [45:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2750s) going look up here for the 1 2 3 4. So there's a big delay right now. [45:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2754s) there's a big delay right now. [45:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2754s) there's a big delay right now. [music] [45:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2756s) Okay, now it's it's supervised learning. [46:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2760s) Okay, now it's it's supervised learning. [46:00](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2760s) Okay, now it's it's supervised learning. So this [music] early numbers where it's [46:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2763s) So this [music] early numbers where it's [46:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2763s) So this [music] early numbers where it's hearing them more 1 2 3 4 are going to [46:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2766s) hearing them more 1 2 3 4 are going to [46:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2766s) hearing them more 1 2 3 4 are going to learn a little faster. So now see how [46:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2768s) learn a little faster. So now see how [46:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2768s) learn a little faster. So now see how it's [music] speeding up a little bit [46:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2770s) it's [music] speeding up a little bit [46:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2770s) it's [music] speeding up a little bit for the early ones, but there's a lack [46:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2773s) for the early ones, but there's a lack [46:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2773s) for the early ones, but there's a lack in the late. [46:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2775s) in the late. [46:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2775s) in the late. [music] [46:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2775s) [music] [46:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2776s) [music] By the end of this thing, you're going [46:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2777s) By the end of this thing, you're going [46:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2777s) By the end of this thing, you're going to see that we got down to essentially [46:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2780s) to see that we got down to essentially [46:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2780s) to see that we got down to essentially 45 millisecond um [music] time. 6. [46:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2785s) 45 millisecond um [music] time. 6. [46:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2786s) 45 millisecond um [music] time. 6. Here we go. 1 2 3 4 1 2 3 4 5 6. [music] [46:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2791s) Here we go. 1 2 3 4 1 2 3 4 5 6. [music] [46:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2791s) Here we go. 1 2 3 4 1 2 3 4 5 6. [music] So now for those, we're really locked [46:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2793s) So now for those, we're really locked [46:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2793s) So now for those, we're really locked in, right? [46:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2796s) in, right? [46:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2796s) in, right? Oh, and there's a little blip, but it's [46:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2798s) Oh, and there's a little blip, but it's [46:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2798s) Oh, and there's a little blip, but it's learning and it's going. [music] So this [46:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2799s) learning and it's going. [music] So this [46:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2800s) learning and it's going. [music] So this is recursive decision model deciding [46:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2803s) is recursive decision model deciding [46:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2803s) is recursive decision model deciding um when to trust this smaller supervised [46:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2806s) um when to trust this smaller supervised [46:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2806s) um when to trust this smaller supervised [music] model that is learning on the [46:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2808s) [music] model that is learning on the [46:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2808s) [music] model that is learning on the fly and operating on tens of [46:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2811s) fly and operating on tens of [46:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2811s) fly and operating on tens of milliseconds of information [music] on [46:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2813s) milliseconds of information [music] on [46:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2813s) milliseconds of information [music] on the pure audio right. [46:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2816s) the pure audio right. [46:56](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2816s) the pure audio right. So here we go by the end [46:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2819s) So here we go by the end [46:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2819s) So here we go by the end it's really [music] going to get [47:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2821s) it's really [music] going to get [47:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2821s) it's really [music] going to get 2. [47:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2826s) Anyways to me that's super exciting. um [47:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2829s) Anyways to me that's super exciting. um [47:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2829s) Anyways to me that's super exciting. um really really fun to like come back to [47:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2832s) really really fun to like come back to [47:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2832s) really really fun to like come back to these um these real- time on the-fly [47:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2835s) these um these real- time on the-fly [47:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2835s) these um these real- time on the-fly machine learning algorithms which is [47:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2837s) machine learning algorithms which is [47:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2837s) machine learning algorithms which is actually how I first learned um about [47:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2839s) actually how I first learned um about [47:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2839s) actually how I first learned um about machine learning and and I first learned [47:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2841s) machine learning and and I first learned [47:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2841s) machine learning and and I first learned programming through through music. So um [47:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2845s) programming through through music. So um [47:25](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2845s) programming through through music. So um I also want to so that's the audio [47:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2847s) I also want to so that's the audio [47:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2847s) I also want to so that's the audio example. I also want to give a shout out [47:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2849s) example. I also want to give a shout out [47:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2849s) example. I also want to give a shout out to this project called Hunch uh Matt [47:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2851s) to this project called Hunch uh Matt [47:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2851s) to this project called Hunch uh Matt Mastrai. I'm not sure exactly how to [47:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2853s) Mastrai. I'm not sure exactly how to [47:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2853s) Mastrai. I'm not sure exactly how to pronounce it. He's done something really [47:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2855s) pronounce it. He's done something really [47:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2855s) pronounce it. He's done something really crazy visually messing with just [47:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2858s) crazy visually messing with just [47:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2858s) crazy visually messing with just embedding Gemma 2 um and essentially [47:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2861s) embedding Gemma 2 um and essentially [47:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2861s) embedding Gemma 2 um and essentially having this thing play Pictionary. So he [47:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2863s) having this thing play Pictionary. So he [47:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2863s) having this thing play Pictionary. So he has this thing classifying your doodles [47:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2867s) has this thing classifying your doodles [47:47](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2867s) has this thing classifying your doodles um on your machine in WebGPU in like [47:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2871s) um on your machine in WebGPU in like [47:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2871s) um on your machine in WebGPU in like what look at these numbers like 150 [47:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2873s) what look at these numbers like 150 [47:53](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2873s) what look at these numbers like 150 milliseconds. So again I want I I just [47:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2875s) milliseconds. So again I want I I just [47:55](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2875s) milliseconds. So again I want I I just want to give this a shout out as like [47:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2877s) want to give this a shout out as like [47:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2877s) want to give this a shout out as like I'm not the only one working in this [47:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2878s) I'm not the only one working in this [47:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2878s) I'm not the only one working in this space. audio isn't the only um way in, [48:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2882s) space. audio isn't the only um way in, [48:02](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2882s) space. audio isn't the only um way in, but these these very very very small um [48:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2885s) but these these very very very small um [48:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2885s) but these these very very very small um embedding models, encoder models, [48:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2887s) embedding models, encoder models, [48:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2887s) embedding models, encoder models, classifier models, right? Like there's a [48:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2889s) classifier models, right? Like there's a [48:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2889s) classifier models, right? Like there's a whole resurgence of this and if you can [48:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2892s) whole resurgence of this and if you can [48:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2892s) whole resurgence of this and if you can uh structure them and architect them [48:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2894s) uh structure them and architect them [48:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2894s) uh structure them and architect them into a bigger space, you can do really [48:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2896s) into a bigger space, you can do really [48:16](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2896s) into a bigger space, you can do really really powerful things. Okay, so with [48:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2899s) really powerful things. Okay, so with [48:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2899s) really powerful things. Okay, so with this, this is kind of where I'm going. [48:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2901s) this, this is kind of where I'm going. [48:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2901s) this, this is kind of where I'm going. It's like how can we decouple our bodies [48:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2904s) It's like how can we decouple our bodies [48:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2904s) It's like how can we decouple our bodies from our devices? [48:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2906s) from our devices? [48:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2906s) from our devices? What would it mean to work on a computer [48:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2908s) What would it mean to work on a computer [48:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2908s) What would it mean to work on a computer without touching it? What would it mean [48:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2911s) without touching it? What would it mean [48:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2911s) without touching it? What would it mean to work on a computer without being able [48:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2912s) to work on a computer without being able [48:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2912s) to work on a computer without being able to see it? Right? And that's really what [48:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2916s) to see it? Right? And that's really what [48:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2916s) to see it? Right? And that's really what um I'm building up towards with this [48:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2919s) um I'm building up towards with this [48:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2919s) um I'm building up towards with this project that I have called OSO [48:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2921s) project that I have called OSO [48:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2921s) project that I have called OSO OSO.computer. [48:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2923s) OSO.computer. [48:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2923s) OSO.computer. It's short for operating system 1. It's [48:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2925s) It's short for operating system 1. It's [48:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2925s) It's short for operating system 1. It's not quite ready. Um, but that earshot [48:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2928s) not quite ready. Um, but that earshot [48:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2928s) not quite ready. Um, but that earshot demo that I showed you where we're very [48:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2931s) demo that I showed you where we're very [48:51](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2931s) demo that I showed you where we're very quickly from a single uh word uh even [48:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2934s) quickly from a single uh word uh even [48:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2934s) quickly from a single uh word uh even before uh even at a speed that would be [48:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2937s) before uh even at a speed that would be [48:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2937s) before uh even at a speed that would be faster than we can actually transcribe [48:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2939s) faster than we can actually transcribe [48:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2939s) faster than we can actually transcribe what we're saying, we're able to react [49:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2941s) what we're saying, we're able to react [49:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2941s) what we're saying, we're able to react and respond um to to our voice. So, [49:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2945s) and respond um to to our voice. So, [49:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2945s) and respond um to to our voice. So, we're starting starting with voice as [49:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2947s) we're starting starting with voice as [49:07](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2947s) we're starting starting with voice as the first input. Um, and yeah, just like [49:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2950s) the first input. Um, and yeah, just like [49:10](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2950s) the first input. Um, and yeah, just like a little bit about OSO as like a teaser [49:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2952s) a little bit about OSO as like a teaser [49:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2952s) a little bit about OSO as like a teaser and I hope you'll join the wait list and [49:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2954s) and I hope you'll join the wait list and [49:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2954s) and I hope you'll join the wait list and try it out when it's ready is like it is [49:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2958s) try it out when it's ready is like it is [49:18](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2958s) try it out when it's ready is like it is a computer use agent but it's built on [49:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2960s) a computer use agent but it's built on [49:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2960s) a computer use agent but it's built on recursive decision models. So yes, there [49:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2962s) recursive decision models. So yes, there [49:22](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2962s) recursive decision models. So yes, there are uh regular agents and RLM agents um [49:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2967s) are uh regular agents and RLM agents um [49:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2967s) are uh regular agents and RLM agents um and you can actually um uh you can um [49:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2971s) and you can actually um uh you can um [49:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2972s) and you can actually um uh you can um sort of farm out jobs or um uh give them [49:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2975s) sort of farm out jobs or um uh give them [49:35](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2975s) sort of farm out jobs or um uh give them to like your normal coding agents like [49:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2977s) to like your normal coding agents like [49:37](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2977s) to like your normal coding agents like cloud code and codeex. Um but it's it's [49:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2979s) cloud code and codeex. Um but it's it's [49:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2979s) cloud code and codeex. Um but it's it's decision first, right? And so you can [49:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2981s) decision first, right? And so you can [49:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2982s) decision first, right? And so you can also do things that are super fast uh [49:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2984s) also do things that are super fast uh [49:44](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2984s) also do things that are super fast uh without touching the keyboard, without [49:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2986s) without touching the keyboard, without [49:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2986s) without touching the keyboard, without touching the mouse. Um it's recursively [49:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2988s) touching the mouse. Um it's recursively [49:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2988s) touching the mouse. Um it's recursively improving. It's constantly trying to [49:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2990s) improving. It's constantly trying to [49:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2990s) improving. It's constantly trying to learn um what the preferences are um of [49:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2994s) learn um what the preferences are um of [49:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2994s) learn um what the preferences are um of the user and it's also trying to [49:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2998s) the user and it's also trying to [49:58](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=2998s) the user and it's also trying to essentially always like that um earshot [50:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3001s) essentially always like that um earshot [50:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3001s) essentially always like that um earshot demo the the Philip Glass demo. It's [50:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3003s) demo the the Philip Glass demo. It's [50:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3003s) demo the the Philip Glass demo. It's also trying to like farm out the [50:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3005s) also trying to like farm out the [50:05](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3005s) also trying to like farm out the decision to the the fastest smallest um [50:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3009s) decision to the the fastest smallest um [50:09](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3009s) decision to the the fastest smallest um available model once it can build trust [50:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3011s) available model once it can build trust [50:12](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3012s) available model once it can build trust in that. So that's this idea of the [50:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3014s) in that. So that's this idea of the [50:14](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3014s) in that. So that's this idea of the arbiter, right? It's like figuring out, [50:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3015s) arbiter, right? It's like figuring out, [50:15](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3015s) arbiter, right? It's like figuring out, okay, where have we seen this thing do [50:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3017s) okay, where have we seen this thing do [50:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3017s) okay, where have we seen this thing do well before and then now how much can I [50:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3020s) well before and then now how much can I [50:20](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3020s) well before and then now how much can I trust it over time? So, it's a it's an [50:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3021s) trust it over time? So, it's a it's an [50:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3021s) trust it over time? So, it's a it's an evolving system. It is a tool to control [50:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3024s) evolving system. It is a tool to control [50:24](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3024s) evolving system. It is a tool to control your computer without touching it. [50:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3026s) your computer without touching it. [50:26](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3026s) your computer without touching it. Again, we're starting with voice. It's [50:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3028s) Again, we're starting with voice. It's [50:28](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3028s) Again, we're starting with voice. It's an experiment in what I consider radical [50:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3031s) an experiment in what I consider radical [50:31](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3031s) an experiment in what I consider radical human computer interfacing. Let's like [50:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3033s) human computer interfacing. Let's like [50:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3033s) human computer interfacing. Let's like rethink everything. I think there's this [50:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3036s) rethink everything. I think there's this [50:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3036s) rethink everything. I think there's this really interesting thing about everyone [50:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3038s) really interesting thing about everyone [50:38](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3038s) really interesting thing about everyone going into the terminal um with cloud [50:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3040s) going into the terminal um with cloud [50:40](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3040s) going into the terminal um with cloud code and I don't think it's actually [50:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3042s) code and I don't think it's actually [50:42](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3042s) code and I don't think it's actually anything [50:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3043s) anything [50:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3043s) anything um special about the terminal that's why [50:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3046s) um special about the terminal that's why [50:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3046s) um special about the terminal that's why um that happened. The terminal is [50:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3048s) um that happened. The terminal is [50:48](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3048s) um that happened. The terminal is clearly not like the best interface for [50:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3050s) clearly not like the best interface for [50:50](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3050s) clearly not like the best interface for it. I think a big part of what happened [50:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3052s) it. I think a big part of what happened [50:52](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3052s) it. I think a big part of what happened over the last couple years of us all [50:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3054s) over the last couple years of us all [50:54](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3054s) over the last couple years of us all going into the terminal was really about [50:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3057s) going into the terminal was really about [50:57](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3057s) going into the terminal was really about uh just realizing the limitations of [50:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3059s) uh just realizing the limitations of [50:59](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3059s) uh just realizing the limitations of these default interfaces that we've been [51:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3061s) these default interfaces that we've been [51:01](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3061s) these default interfaces that we've been given, these default application [51:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3063s) given, these default application [51:03](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3063s) given, these default application interfaces, these default visual um [51:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3066s) interfaces, these default visual um [51:06](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3066s) interfaces, these default visual um interfaces, touchscreen interfaces, [51:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3068s) interfaces, touchscreen interfaces, [51:08](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3068s) interfaces, touchscreen interfaces, mice, keyboards, all this stuff. Um, and [51:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3071s) mice, keyboards, all this stuff. Um, and [51:11](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3071s) mice, keyboards, all this stuff. Um, and then yeah, lastly, it's just like um I'm [51:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3073s) then yeah, lastly, it's just like um I'm [51:13](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3073s) then yeah, lastly, it's just like um I'm having so much fun uh coming back to uh [51:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3077s) having so much fun uh coming back to uh [51:17](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3077s) having so much fun uh coming back to uh what I consider my roots in in real time [51:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3079s) what I consider my roots in in real time [51:19](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3079s) what I consider my roots in in real time machine learning and and it's really fun [51:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3081s) machine learning and and it's really fun [51:21](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3081s) machine learning and and it's really fun to be building this this computer use [51:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3083s) to be building this this computer use [51:23](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3083s) to be building this this computer use agent also. So anyways, I hope you will [51:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3087s) agent also. So anyways, I hope you will [51:27](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3087s) agent also. So anyways, I hope you will check it out. I'm really grateful for [51:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3089s) check it out. I'm really grateful for [51:29](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3089s) check it out. I'm really grateful for your attention. This is a long talk. I'm [51:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3092s) your attention. This is a long talk. I'm [51:32](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3092s) your attention. This is a long talk. I'm using the opportunity to do the uh [51:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3093s) using the opportunity to do the uh [51:33](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3093s) using the opportunity to do the uh virtual talk to to really go deep. Um, [51:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3096s) virtual talk to to really go deep. Um, [51:36](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3096s) virtual talk to to really go deep. Um, and yeah, I hope you'll check out the [51:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3099s) and yeah, I hope you'll check out the [51:39](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3099s) and yeah, I hope you'll check out the recursive decision model repositories, [51:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3101s) recursive decision model repositories, [51:41](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3101s) recursive decision model repositories, and I hope you'll sign up uh for this [51:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3103s) and I hope you'll sign up uh for this [51:43](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3103s) and I hope you'll sign up uh for this weight list when OSO is ready at [51:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3105s) weight list when OSO is ready at [51:45](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3105s) weight list when OSO is ready at oso.computer. [51:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3106s) oso.computer. [51:46](https://www.youtube.com/watch?v=hFNbPmzzcEs&t=3106s) oso.computer. Thank you so much.