# What Your Coding Agent Can't See in the Code - Tom Moor, Linear Tom Moor, Linear [Talk page](https://ai.engineer/talks/Xw9utONfhIY-what-your-coding-agent-cant-see) · [Video](https://www.youtube.com/watch?v=Xw9utONfhIY) ## What Your Coding Agent Can't See in the Code A pull request can bring back icons a team removed a year ago for good reason. Tom Moor explains how Linear gives coding agents access to those decisions, from the first investigation of an issue to the review of a change. ### Longer agent runs change who supplies the context [0:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1s) Linear started around 2019 because its founders wanted an issue tracker they enjoyed using. The tools they knew were slow, buggy and hard to use. Within the life of that product, the engineer picking up an issue went from accepting an AI-assisted tab completion in 2021 to steering an early coding agent in 2024. Tom Moor sees a consequential step after that one. When an engineer directs a local agent, they can explain a decision as the work unfolds. When several agents run automatically in the cloud, that explanation has to reach them some other way. How much cheaper might it get to hand work to an agent? Moor illustrates the trend with a chart comparing the cost of a level of model capability over three years, from $100 to five cents. He uses that comparison to imagine a future in which generating a feature’s worth of code takes about as long as snapping your fingers. The chart concerns model capability. The cost of getting a particular feature ready to ship is a different question. The other change is how long an agent can work between interventions. Moor describes runs lasting hours or days, though he questions how much code from the much-discussed weekend runs actually ships. If agents can work cheaply for that long, manually delegating, starting and stopping each run becomes an awkward way to coordinate them. More of the work moves into the *outer loop* — deciding what to hand an agent, what it needs to know and how to judge what comes back. ### The cost of one more setting [5:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=347s) Consider a settings page. Each request to add one more option can make sense on its own. Fulfill enough of them and the result starts to resemble Moor’s deliberately overstuffed illustration. Implementation has become cheap enough that the harder decision is which settings deserve to exist together. Consistency and restraint come from many small decisions that reinforce one another. Reading the implementation tells an agent how a feature works. The reason a customer needs it, the compromise that shaped its design and the way it should feel in use may live elsewhere. So may the information that another team is changing it, that a product lead owns the decision, or that someone reverted a similar change two weeks ago. Linear’s approach is to collect that context while investigating the original idea and carry it through planning, implementation and review. ### Triage leaves an investigation on the issue [8:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=536s) A support request, a chat message, a product idea from the CEO and a monitoring alert can all become incoming work. Linear collects them in triage. Its triage agent labels and assigns issues, groups them into themes, closes duplicates and routes work to the relevant team or person. When the report leaves a consequential question open, the agent flags the decision that needs more input. The agent also investigates and leaves its findings on the issue for the next person or agent. A terse crash report can acquire a stack trace and associated logs. Internal tools connected through MCP can supply account information. Customer history can reveal that a high-value customer has asked for the same thing three times in a month. Related issues can explain what the team decided the last time it touched this area. Each addition answers a different practical question, from where to start debugging to how much the request matters. Why did video captions keep turning on? Finding the setting would tell an engineer where the preference lived. Reproducing the bug revealed something more specific. The browser performed an action automatically, and the product interpreted that action as a choice the user had made. The triage agent left that distinction on the issue. Moor credits the diagnosis with making an automated fix possible. A second request raised a different question. Had the product lost behavior it used to support? The investigation found similar work that had already shipped and ruled out a regression. It identified the customers associated with the request, recognized its value and pointed to a likely starting place in the code. The handoff gave the implementer both a reason to pick up the work and somewhere to begin. ### Give the implementer the issue and the failures [12:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=736s) When triage judges a bug small enough, Linear starts the fix automatically. A person can also initiate the work, using the issue context locally through MCP or running the agent remotely. The choice of who starts the job is separate from where it runs. In either case, the coding agent begins with the investigation. Two things matter once the agent makes a change. It needs a way to verify its work, and failures from continuous integration (CI) need to reach the agent so it can act on them. The issue explains the intended behavior. The failing check gives the implementer fresh evidence about the proposed fix. Moor reports that this combination worked well for his team and points to reusable verification skills as one way to give an agent that capability. ### Use the diff to narrow the search for product history [13:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=819s) A code reviewer working from the repository can catch plenty of logical errors. Linear dispatches a review agent with a second question to answer. Given what the customer needed and what the team knows about this part of the product, is this the right change? The agent starts with the original issue, including the customer context and intent that triage recorded. What can review look up that triage could only anticipate? By the time a pull request exists, the reviewer knows which code actually changed. That narrows the search for earlier decisions, reverts, regressions and incidents. It also gives the agent a reason to revisit related issues. Minutes or weeks may have passed since intake, and another change may now overlap or conflict with this one. The reviewer can also read the full coding session. A one-shot implementation and a long exchange with a human can end in a similar diff, but the exchange may explain a critical choice. Ownership information answers a separate question about who still needs to weigh in. A product lead or the CEO may own the tradeoff even when someone else usually reviews the code. How does the reviewer find the history that matters to this change? The original request goes directly to review. The pull request also guides a fresh search, so an old incident or a newly overlapping change can inform the decision alongside the original intent. - Original issue — Customer, reported problem and findings from triage - Pull request — Changed code and the intent of the proposed change - History and related work — Decisions, incidents and reverts in the affected area, plus work that now overlaps - Coding session — Decisions made while implementing this change - Review of this change — Does the implementation fit the product, and whose judgment does it need? - Original issue → Review of this change (Supplies the customer's intent) - Pull request → History and related work (Identifies the affected area) - Pull request → Review of this change (Supplies the implementation to assess) - History and related work → Review of this change (Supplies precedents and possible conflicts) - Coding session → Review of this change (Explains choices behind the diff) ### What history caught in three pull requests [15:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=955s) - **An open database transaction.** The review agent found changes from six and seven months earlier that identified holding a database transaction open as an incident risk. A reviewer could recognize the risk from the code alone, but Codex missed it on this change. The earlier records gave Linear’s agent a specific reason to flag it. - **Icons the team had removed.** A year earlier, Linear dropped some iconography because it made the interface look busy. A new pull request proposed bringing back icons the team had rejected several times. The review agent recovered that design decision and put it in a review comment. The comment went straight back to the coding agent, giving it the reason those icons were missing in the first place. - **The order of two kinds of labels.** Should workspace labels appear below team labels? The proposed change appeared to put them in that order. Linear lists labels in many places, with no single piece of code defining the convention across the product. A previous revert and a related issue gave the reviewer grounds to flag the ordering. Reading either list in isolation would do little to explain why its order mattered. These findings also have a human destination. Linear built a review interface that guides a person through the changes and brings earlier decisions to their attention where a judgment call is needed. Knowing why the team removed an icon gives the reviewer something concrete to consider when deciding whether it belongs this time. ### Turn corrections that changed the code into guidance [18:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1096s) After a merge, the code contains the outcome. The issue and pull request still contain the discussion that explains it. Once a day, a Linear automation scans the day’s merged work for actionable lessons. It looks for places where a review comment led to a change or where a revert happened. That gives it a way to select feedback that affected the product and write the lesson into the repository’s agent instructions or the relevant documentation. Daily is a choice that fits Linear’s pull request volume. Another team could run the job more or less often, using its own automation tooling. The extraction prompt takes experimentation, too. A reviewer’s correction should be available to the next implementer, but continually appending every lesson would eventually leave an agent with an enormous instruction file to read. Choosing what deserves to stay is part of the work. Moor’s aim is fewer avoidable changes and less rework. At intake, the customer’s intent can redirect an investigation before anyone builds the wrong fix. At review, an old decision or another team’s ongoing work can change whether a pull request should ship. A human who makes that call has already done the expensive part. Keeping the reason gives the next agent a chance to act on it before someone has to make the same correction twice. ## Key takeaways - Put the triage diagnosis on the issue. In the captions case, the next agent needed to know that the browser had acted automatically and the product had treated that action as user intent. - Refresh the history search at review time. The changed code narrows which past decisions matter, and the time since intake creates room for new conflicts. - Use changed code and reverts to identify review feedback worth keeping. The practical constraint is how much guidance a future agent must read. ## Slides ### Your coding agent has a context problem A human finger and dozens of hand-shaped cursors point toward each other. Tom Moor opens with a question about the work around coding: what context does an agent miss when it can only see the code? [0:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=12s) [Above the title, a grayscale hand points toward a cluster of small hand-shaped cursors.](https://assets.ai.engineer/sites/aiecode2025/sha256/4486cccb24649c4ef8aa6e6cba7f5e5b2914cbd64c5fc4efaeb38eb14ce253c0.webp) Your coding agent has a context problem Linear 1 Interface text: 1 / 23 ### What is Linear? Linear began as an issue tracker. Its inbox shows an assigned engineering issue alongside project updates, with the issue’s code details open on the right. Moor introduces this workspace as a product development system for teams and agents. [0:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=35s) [The slide describes Linear as “The product development system for teams and agents” above an angled view of its inbox and an open issue.](https://assets.ai.engineer/sites/aiecode2025/sha256/5448ec6f3e76fbebc056df3d0f08e1c53d609866e87d47fa0cf45a0eaa3fb23e.webp) What is Linear? The product development system for teams and agents Linear 2 Interface text: Linear Inbox 1 My issues Workspace Initiatives Projects Views Teams Favorites iOS app Q2 Initiatives Projects Docs Your teams Engineering Issues Active Inbox ENG-135 Refactor sonic crawler nan assigned you · 2h LLM Chatbot New project update by cristina · 8h ENG-159 Error uploading images via API SLA breached · 1d DES-498 Redesign users settings... karri mentioned you · 2d ENG-160 Holtzmann engine is broken You asked to be reminded about this issue · 4d Sign up flow experiments Added as a project member by edgar · 1w MKT-122 Design assets for marketing... erin marked as Duplicate · 1w Homepage v3 New project update by paco · 1w Engineering › Spice harvester › ENG-135 Clear notification Refactor sonic crawler Comment.documentContent is defined wrongly. It should be a LazyManyToOne relation[illegible] /** The document content that this comment is associated with. */ @ManyToOne(() => DocumentContent, "comments", { optional: true, [illegible] public documentContent?: DocumentContent; We would be accessing CachedPromise then, and [illegible] property touch. comment.documentContent.value would still be undefined until [illegible] + Add sub-issues Links #20319 bjor/eng-135 add source to insights file... Activity 2 / 23 ### Coding tools from 2021 to 2026 In 2021, Copilot helped developers complete code. By 2025, Moor saw Claude Code as a mature coding agent. For 2026, he sees the challenge shifting from guiding one local agent to coordinating several or letting them run in the cloud. [1:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=106s) [The timeline pairs GitHub Copilot with tab completion in 2021, ChatGPT with conversational debugging in 2022, early coding agents with 2024, Claude Code with maturity in 2025, and the software factory with orchestration in 2026.](https://assets.ai.engineer/sites/aiecode2025/sha256/90b165ae9ee83b5b01e82deea87340aa3a06e44517e055794efd66c4d9300295.webp) What the... 2021 - GitHub Copilot (tab completion) 2022 - ChatGPT (conversational debugging) 2024 - Coding agents (early) 2025 - Claude Code (mature) 2026 - Software factory (orchestration is frontier) Linear 3 Interface text: 3 / 23 ### Intelligence will continue to become cheaper and more abundant In Tom Moor’s example, the same level of intelligence that cost $100 costs just five cents three years later. He expects prices to keep falling in 2027 and beyond. [3:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=224s) [A descending curve shows $100 at Year 0, $7.7 at Year 1, $0.6 at Year 2 and $0.05 at Year 3.](https://assets.ai.engineer/sites/aiecode2025/sha256/0a9f1f2ec691e42a4b8d7464be916bae3c808063cf52513708907911eeae6a11.webp) 2027 and beyond Intelligence will continue to become cheaper and more abundant Y-axis unit: $ Y-axis ticks: 0, 20, 40, 60, 80, 100, 120 Timeline: Year 0, Year 1, Year 2, Year 3 Point labels: 100, 7.7, 0.6, 0.05 Linear 4 Interface text: 4 / 23 ### The task horizon is exponential METR measured agents taking on longer tasks. The purple curve projects how much longer those tasks could get by 2028 if the task horizon doubles every 4–7 months. [4:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=269s) [A chart spanning 2019 to 2028 shows METR measurements in gray and a steeply rising projection in purple. The legend gives a projected doubling time of four to seven months.](https://assets.ai.engineer/sites/aiecode2025/sha256/54f35759aea81008cdfce031f836d6a46bdcc8f632389c65bccb844ab855f8be.webp) 2027 and beyond The task horizon is exponential Legend: Measured (METR) · Projected (4–7 mo doubling) Y-axis ticks: 0, 1000, 2000, 3000, 4000 Timeline: 2019, 2020, 2021, 2022, 2023, 2024, 2025, 2026, 2027, 2028 Linear 5 Interface text: 5 / 23 ### What breaks Moor’s generated settings page has two dark mode switches in one panel, one on and one off. Adding a switch can make sense on its own and still leave users wondering which one controls the theme. [5:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=343s) [The slide states that thousands of locally correct changes create an incoherent product. Beside it, a grid shows several variations of controls for appearance, notifications, privacy and sync.](https://assets.ai.engineer/sites/aiecode2025/sha256/13c635f03c470fd098ea7f7c807338bea2417b7d0d03f125df275d91736490f0.webp) What breaks Thousands of locally correct changes create an incoherent product. Appearance Dark mode Dark mode Theme Display Auto brightness Adaptive lighting Theme System Interface Light Dark System Use system theme Color scheme Visual Theme Light Dark theme High contrast Reduce motion Notifications Push Email In-app Desktop Notification style Banners Alerts Enable notifications Email alerts Push alerts In app notifications Alert style None Notifications All Important only None Show previews Play sound Messages Inbox alerts Email notifications Push notifications Show in-app Notification style Compact Privacy Data sharing Share analytics Personalization Data Analytics Usage data Product improvement Share diagnostic data Tracking Allow tracking Analytics Ad personalization Usage statistics Privacy & data Send usage data Share diagnostics Personalize experience Tracking Limited Sync Sync data Auto sync Sync over mobile Synchronization Enable sync Sync automatically Background sync Sync frequency Hourly Auto sync Off On Wi-Fi only Sync interval Daily Syncing Sync Auto-sync Sync mode Manual Advanced Performance mode Battery saver Optimize automatically Performance Performance mode Balanced Low power mode Optimize for battery System Optimize performance Use less power Background activity Limited Power Energy saver Low power mode Restrict background activity Linear 6 Interface text: 6 / 23 ### What’s missing? A coding agent can read the code and still miss why a change matters, how it should fit the product, or who decides the tradeoffs. It might even miss that the team reverted a change in the same area two weeks ago. [6:53](https://www.youtube.com/watch?v=Xw9utONfhIY&t=413s) [Two columns pair intent with why a change matters, history with why it works this way, conventions with how it should fit, other work with what else is changing, ownership with who decides tradeoffs, and risks with what could break.](https://assets.ai.engineer/sites/aiecode2025/sha256/e7e36076adc7de1b1ff74f441e089ddf48d9a1739e378bda91c570b765f9831b.webp) What’s missing? Intent why this matters History why it works this way Conventions how it should fit Other work what else is changing Ownership who decides tradeoffs Risks what this could break Linear 7 Interface text: 7 / 23 ### Intake → Implement → Review → Distill Each stage passes context to the next. After review, the team distills what it learned and brings those lessons back to intake for the next round of work. [8:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=495s) [Arrows connect Intake, Implement, Review and Distill in sequence. A lower arrow leads from Distill back to Intake, with the caption “Each stage leaves context for the next.”](https://assets.ai.engineer/sites/aiecode2025/sha256/9996396008b13cca317d51b142e11913aee4c0fe3297d8c271e0bd2b677fdbe8.webp) Intake → Implement → Review → Distill Each stage leaves context for the next Linear 8 Interface text: 8 / 23 ### Intake Linear automated intake to bring product ideas and bug reports from Zendesk, Slack, Teams, Gong and other sources into one place for triage. [8:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=534s) [Lines from Zendesk, Slack, Teams, Gong and an ellipsis converge on the Linear logo, showing how ideas and bug reports arrive from multiple sources.](https://assets.ai.engineer/sites/aiecode2025/sha256/578604c6bb3d7d7da224c752611f207731558a4432ae21d7146100a2b6ac463b.webp) Intake Ideas and bugs ingested from Zendesk, Slack, Teams, Gong, more... Linear 9 Interface text: 9 / 23 ### Triage agent Linear’s triage agent adds customer details and related issues to an incoming report. The four branches show how triage can lead to different actions, from closing a duplicate to flagging a question that needs clarification. [9:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=567s) [The triage agent adds customer information, related issues, a likely assignee and a project. Lines from a box labeled “Incoming” branch toward four icons.](https://assets.ai.engineer/sites/aiecode2025/sha256/d340f05eece3ec4f1cccaaf0e850f2c362087f770713ffccf6fdf210392534c2.webp) Triage agent Enriches the issue context with customer info, related issues, likely assignee, project and more... Incoming Linear 10 Interface text: 10 / 23 ### Triage example — video caption preferences The triage agent reproduced the switch from off to on. The browser enabled a default caption track, and the player saved that as the user’s choice. The agent couldn’t verify the reporter’s original preference or browser settings. [10:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=659s) [Overlapping cards show a report that captions turn on despite a saved off preference and a triage response. In testing, loading a captioned video changed the saved preference without user interaction. The reporter’s original settings remained unverified.](https://assets.ai.engineer/sites/aiecode2025/sha256/541f443ebcefe58aa31a26129f507a1a5cc0571d93f294639a55e300a957390d.webp) Triage agent Some examples Video captions turn on despite a saved off preference Video captions appear automatically in project comments. Investigation reproduced loading a captioned video changing a saved off preference to on without user interaction. Linear 5d ago via Slack I reproduced a path that turns captions on despite a saved “off” preference: • Caption tracks are always marked default, so the browser can enable them automatically • The player then treats that as a preference change and saves “on” to user settings • Project comments use the same player, so this is not specific to them In testing, loading a captioned video changed saved “off” to “on” without any interaction. This likely explains the report, though I couldn’t verify the original saved preference or browser settings. No code changes made Linear 11 Interface text: 11 / 23 ### Implementation context and verification The coding agent starts with the triage findings and can look up organizational context through MCP. It uses a verification skill to check its changes, then uses CI failures to guide the next attempt. [12:14](https://www.youtube.com/watch?v=Xw9utONfhIY&t=734s) [The Implement slide lists enriched context, MCP access to organizational context, a verification skill, and CI failures. A repeat symbol appears below the list.](https://assets.ai.engineer/sites/aiecode2025/sha256/51d39654a9548c0a74a2a137dc9249dab01e50c3882edfab3448016c6592c503.webp) Implement Provide the enriched context MCP with access to org context Verification skill CI failures 🔁 Linear 13 Interface text: 13 / 23 ### Inputs to pull request review Moor asks what a review agent can see. If its context comes from the code changes, repo history, and pull request description, he expects it to focus mainly on whether the code’s logic is correct. [13:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=817s) [Under Review, three lines list the code changes, the repository history, and the pull request description.](https://assets.ai.engineer/sites/aiecode2025/sha256/73ad90375cbf26a24081e1c758ec563b1b4edb3edca0903b7ed298e0c3b389b6.webp) Review The code changes The repo history The PR description Linear 14 Interface text: 14 / 23 ### Review example: team and workspace labels The review flags a change that again puts workspace labels below team labels by default. It cites an issue and a previous revert to explain why the labels belong together. A workspace label can be the more relevant choice. [17:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1029s) [A review comment flags workspace labels appearing below team labels. Two excerpts explain why the labels should appear together by default.](https://assets.ai.engineer/sites/aiecode2025/sha256/52d521c35607b388ed22ed5fba505173c4fdae0e98c51cc72ba9f2c25af8358f.webp) Review Example Workspace labels again sit below team labels Workspace labels again sit below team labels when they are not selected, suggested or frequent. Produc[illegible] feedback rejected this split because workspace labels can be more relevant. Keep team and workspace labels together by default. Code Sources 2 “and often the more relevant labels are workspace level.” Andreas Eldh in LIN-45649 · 11 months ago “Separately, there was product feedback to revert the team labels split.” Ema Milojkovic in Revert spliting our team & selected... · 10 months ago Interface text: Linear 1 of 1 Follow ups Keep[illegible] Leave a[illegible] 18 18 / 23 ### Review guide for a notification refactor The guide groups Android and iOS files under one design decision: notification categories now own their default text. Beside that explanation, the Kotlin diff shows the category field changing from Category to NotificationCategory. [17:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1067s) [The guide explains a shared notification model and lists changed Android and iOS files. A code diff on the right highlights a change to the notification category type.](https://assets.ai.engineer/sites/aiecode2025/sha256/ec3b9244c73df31cb6364abe7a408625b5639f64d1e8f021fb06d1d9aa267843.webp) Review Guide Refactor notification model and capture parse failures eldh · rideshare#1557 · main ← eldh/driver/notification-parse-failures 01 / 03 Shared Notification Model Separating category, route, and parse-failure concepts keeps driver notifications consistent across platforms. Categories now own their default copy, while notifications carry optional parse failure details for analytics reporting. NotificationCategory.kt driver/android/app/src/main/kotli... +57 DriverNotification.kt driver/android/app/src/main/kotlin/c... +2 -52 NotificationCategory.swift driver/ios/DriverApp/Services/... +42 DriverNotification.swift driver/ios/DriverApp/Services/No... +8 -36 NotificationCategory.kt driver/android/app/src/main/kotlin... +57 DriverNotification.kt driver/android/app/src/main/kotlin/co... +2 -52 1 package com.ridershare.driver.common.notifications 2 import java.time.Instant 3 4 data class DriverNotification( 5 val id: String, - val category: Category, 6 val category: NotificationCategory, 7 val title: String, 8 val body: String, 9 val route: DriverNotificationRoute, 10 val receivedAt: Instant, 11 val source: Source, 12 val raw: Map, 13 ) { - enum class Category(val rawValue: String) { - TripRequest("trip_request"), - TripUpdate("trip_update"), - Earnings("earnings"), Interface text: Linear DRIVE-882 › Refactor notification model and capture parse failures +64 -41 Activity Guide Diff Su[illegible] 19 19 / 23 ### Distill the rationale for a decision When a reviewer rejects an approach, record why alongside the code change. The next agent can use that reasoning to avoid repeating the same mistake. [18:14](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1094s) [Under “Distill,” three lines call for capturing the rejected approach and why, recording the decision with the change, and making the rationale available to the next agent.](https://assets.ai.engineer/sites/aiecode2025/sha256/b5a83fe6afe3b93a3e1f7dd9eaa33b310e076543846a9d71c208800db5a2e6ce.webp) Distill Rejected approach and why Decision recorded with the change Rationale available to the next agent Linear 20 Interface text: 20 / 23 ### Distill loop: learning from merged pull request reviews Linear’s daily automation checks pull requests merged in the past 24 hours. It traces reviewer feedback to the code change it prompted, then checks AGENTS.md before recording a new lesson with a link to its source. [18:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1126s) [An enabled automation card overlaps an instruction panel. The instructions link review feedback to the resulting code changes and compare those lessons with existing agent guidance.](https://assets.ai.engineer/sites/aiecode2025/sha256/103237d5b728174501531d3ef5252b7ad64cb6b52cc5ed7a20fc9b02c66c552a.webp) Distill Loop Learn from merged PR reviews Capture review-driven lessons and maintain concise, evidence-backed agent guidance Instructions Review pull requests in linear/linear-app merged during the preceding 24 hours, using merge timestamps rather than update timestamps. Paginate results so all matching PRs are considered. For each PR, inspect review threads, submitted reviews, and commit/file diffs before and after review. Extract durable, actionable lessons specifically supported by changes made in response to review. Link reviewer feedback to the resulting change; do not treat the entire merged diff, unrelated development, unimplemented suggestions, or timing alone as evidence of a review-driven lesson. Skip uncertain cases and one-off details that do not generalize. Read the current AGENTS.md on the repository's default branch and compare candidate lessons against existing guidance. Record concise lessons that help future work avoid the same review corrections, preserving relevant conditions and scope. Include compact links to the source PR or review discussion. Avoid duplicate guidance or restating lessons already covered. Compress existing learnings only with high confidence: consolidate redundant guidance without losing meaningful constraints; resolve conflicts only when review evidence and current code establish which guidance applies; remove obsolete learnings only when current code or explicit accepted review decisions demonstrate they are no longer relevant. Age or absence from recent reviews is not sufficient evidence. Leave uncertain or differently scoped guidance unchanged. Preserve unrelated instructions and document structure. Linear 21 Interface text: Run now Enabled Owned by Tom Moor · Last update just now 21 / 23 ### What this buys us Clarifying intent at intake helps a coding agent tackle the right problem. Carrying past decisions into implementation cuts rework, and seeing related work helps reviewers judge whether a change fits the product. [20:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1201s) [Five benefits are listed: fewer changes that solve the wrong problem, less rework from missed decisions, more changes that fit the product, faster iteration, and higher quality.](https://assets.ai.engineer/sites/aiecode2025/sha256/34bb2c3e618be98d31091b5304cc3e46befcbd63f2bd76da945ae41d06dcbd15.webp) What this buys us Fewer changes that solve the wrong problem Less rework from missed decisions More changes that fit the product Faster iteration Higher quality Linear 22 Interface text: 22 / 23 ## Full transcript [0:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1s) [music] [0:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=12s) Thanks for turning up to a talk about [0:14](https://www.youtube.com/watch?v=Xw9utONfhIY&t=14s) Thanks for turning up to a talk about [0:14](https://www.youtube.com/watch?v=Xw9utONfhIY&t=14s) Thanks for turning up to a talk about agents and context. I, you know, I [0:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=17s) agents and context. I, you know, I [0:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=17s) agents and context. I, you know, I didn't think these were keywords that [0:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=18s) didn't think these were keywords that [0:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=18s) didn't think these were keywords that would really hit today. Uh, I'm sure it [0:21](https://www.youtube.com/watch?v=Xw9utONfhIY&t=21s) would really hit today. Uh, I'm sure it [0:21](https://www.youtube.com/watch?v=Xw9utONfhIY&t=21s) would really hit today. Uh, I'm sure it won't be the the last the last time you [0:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=22s) won't be the the last the last time you [0:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=22s) won't be the the last the last time you hear about it, but I'm going to try and [0:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=24s) hear about it, but I'm going to try and [0:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=24s) hear about it, but I'm going to try and kind of get out of the the inner loop a [0:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=27s) kind of get out of the the inner loop a [0:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=27s) kind of get out of the the inner loop a little bit uh and think more about the [0:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=29s) little bit uh and think more about the [0:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=29s) little bit uh and think more about the outer loop of of agents. So, um, excuse [0:31](https://www.youtube.com/watch?v=Xw9utONfhIY&t=31s) outer loop of of agents. So, um, excuse [0:31](https://www.youtube.com/watch?v=Xw9utONfhIY&t=31s) outer loop of of agents. So, um, excuse me if it feels a little off topic at [0:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=33s) me if it feels a little off topic at [0:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=33s) me if it feels a little off topic at times, but we we'll get there. [0:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=37s) times, but we we'll get there. [0:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=37s) times, but we we'll get there. So, yeah, I I work at a company called [0:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=39s) So, yeah, I I work at a company called [0:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=39s) So, yeah, I I work at a company called Linear. Um, I'll start with a just a [0:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=42s) Linear. Um, I'll start with a just a [0:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=42s) Linear. Um, I'll start with a just a quick intro for those in the room that [0:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=44s) quick intro for those in the room that [0:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=44s) quick intro for those in the room that might not be familiar. Uh, Lineia [0:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=46s) might not be familiar. Uh, Lineia [0:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=46s) might not be familiar. Uh, Lineia started life as an issue tracker. Um, it [0:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=50s) started life as an issue tracker. Um, it [0:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=50s) started life as an issue tracker. Um, it was born out of the frustrations with [0:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=52s) was born out of the frustrations with [0:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=52s) was born out of the frustrations with the existing products in the space. Uh, [0:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=55s) the existing products in the space. Uh, [0:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=55s) the existing products in the space. Uh, how poor quality they were, kind of [0:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=57s) how poor quality they were, kind of [0:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=57s) how poor quality they were, kind of slow, hard to use, buggy, slow. I won't [1:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=61s) slow, hard to use, buggy, slow. I won't [1:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=61s) slow, hard to use, buggy, slow. I won't name any specific products, but you may [1:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=63s) name any specific products, but you may [1:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=63s) name any specific products, but you may have used other products in the space. [1:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=65s) have used other products in the space. [1:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=65s) have used other products in the space. Um, so you know, the founders very [1:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=68s) Um, so you know, the founders very [1:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=68s) Um, so you know, the founders very selfishly uh wanted to work with great [1:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=70s) selfishly uh wanted to work with great [1:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=70s) selfishly uh wanted to work with great tools. you the gra the the the best [1:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=73s) tools. you the gra the the the best [1:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=73s) tools. you the gra the the the best crafts people [1:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=75s) crafts people [1:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=75s) crafts people always invest in the highest quality [1:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=77s) always invest in the highest quality [1:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=77s) always invest in the highest quality tools that they can afford. Uh and we [1:19](https://www.youtube.com/watch?v=Xw9utONfhIY&t=79s) tools that they can afford. Uh and we [1:19](https://www.youtube.com/watch?v=Xw9utONfhIY&t=79s) tools that they can afford. Uh and we felt that at the time you know engineers [1:21](https://www.youtube.com/watch?v=Xw9utONfhIY&t=81s) felt that at the time you know engineers [1:21](https://www.youtube.com/watch?v=Xw9utONfhIY&t=81s) felt that at the time you know engineers had a distinct lack of good tools. Uh [1:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=84s) had a distinct lack of good tools. Uh [1:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=84s) had a distinct lack of good tools. Uh now this this was around 2019. This is [1:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=86s) now this this was around 2019. This is [1:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=86s) now this this was around 2019. This is when uh linear was founded. Um and you [1:31](https://www.youtube.com/watch?v=Xw9utONfhIY&t=91s) when uh linear was founded. Um and you [1:31](https://www.youtube.com/watch?v=Xw9utONfhIY&t=91s) when uh linear was founded. Um and you know you you might say that the world [1:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=92s) know you you might say that the world [1:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=92s) know you you might say that the world has changed a lot since then. Um some [1:36](https://www.youtube.com/watch?v=Xw9utONfhIY&t=96s) has changed a lot since then. Um some [1:36](https://www.youtube.com/watch?v=Xw9utONfhIY&t=96s) has changed a lot since then. Um some some things have happened since 2019 [1:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=98s) some things have happened since 2019 [1:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=98s) some things have happened since 2019 since 2020. Uh, but as you know, as a [1:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=100s) since 2020. Uh, but as you know, as a [1:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=100s) since 2020. Uh, but as you know, as a company, we're we're still striving to [1:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=101s) company, we're we're still striving to [1:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=101s) company, we're we're still striving to to build build high quality software and [1:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=104s) to build build high quality software and [1:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=104s) to build build high quality software and and build systems that enable that for [1:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=106s) and build systems that enable that for [1:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=106s) and build systems that enable that for everybody else. [1:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=108s) everybody else. [1:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=108s) everybody else. So, what what changed? I hope you guys [1:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=111s) So, what what changed? I hope you guys [1:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=111s) So, what what changed? I hope you guys will just humor me for a second here. [1:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=114s) will just humor me for a second here. [1:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=114s) will just humor me for a second here. Um, this is not new information to you, [1:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=117s) Um, this is not new information to you, [1:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=117s) Um, this is not new information to you, but I I I want to just like bask in in [2:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=120s) but I I I want to just like bask in in [2:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=120s) but I I I want to just like bask in in how fast this stuff moves for a split [2:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=122s) how fast this stuff moves for a split [2:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=122s) how fast this stuff moves for a split second. like we've we've had a complete [2:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=125s) second. like we've we've had a complete [2:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=125s) second. like we've we've had a complete reinvention of software engineering in [2:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=127s) reinvention of software engineering in [2:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=127s) reinvention of software engineering in the time that that our company exists [2:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=129s) the time that that our company exists [2:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=129s) the time that that our company exists and it's only half a decade old. Um it's [2:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=132s) and it's only half a decade old. Um it's [2:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=132s) and it's only half a decade old. Um it's not that long ago that tab completion [2:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=135s) not that long ago that tab completion [2:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=135s) not that long ago that tab completion came out. Um at least AI assisted tab [2:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=138s) came out. Um at least AI assisted tab [2:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=138s) came out. Um at least AI assisted tab completion with like GitHub copilot in [2:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=140s) completion with like GitHub copilot in [2:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=140s) completion with like GitHub copilot in 2021. Um [2:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=143s) 2021. Um [2:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=143s) 2021. Um coding agents did not exist in any [2:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=146s) coding agents did not exist in any [2:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=146s) coding agents did not exist in any functional sense until 2 years ago. Um, [2:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=150s) functional sense until 2 years ago. Um, [2:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=150s) functional sense until 2 years ago. Um, the first coding agents came out in [2:31](https://www.youtube.com/watch?v=Xw9utONfhIY&t=151s) the first coding agents came out in [2:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=152s) the first coding agents came out in 2024. [2:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=153s) 2024. [2:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=153s) 2024. Um, these were really early. They were [2:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=155s) Um, these were really early. They were [2:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=155s) Um, these were really early. They were not that great and you definitely had to [2:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=157s) not that great and you definitely had to [2:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=157s) not that great and you definitely had to be kind of cutting edge to be using [2:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=158s) be kind of cutting edge to be using [2:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=158s) be kind of cutting edge to be using them. [2:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=160s) them. [2:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=160s) them. Um, so coding agents themselves are both [2:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=162s) Um, so coding agents themselves are both [2:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=162s) Um, so coding agents themselves are both like long in the tooth and like super [2:45](https://www.youtube.com/watch?v=Xw9utONfhIY&t=165s) like long in the tooth and like super [2:45](https://www.youtube.com/watch?v=Xw9utONfhIY&t=165s) like long in the tooth and like super early right now. Um, I'm assuming that [2:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=169s) early right now. Um, I'm assuming that [2:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=169s) early right now. Um, I'm assuming that if you're in this room, you're probably [2:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=171s) if you're in this room, you're probably [2:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=171s) if you're in this room, you're probably using a coding agent as the primary way [2:53](https://www.youtube.com/watch?v=Xw9utONfhIY&t=173s) using a coding agent as the primary way [2:53](https://www.youtube.com/watch?v=Xw9utONfhIY&t=173s) using a coding agent as the primary way to work right now. Um, that is the [2:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=177s) to work right now. Um, that is the [2:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=177s) to work right now. Um, that is the assumption of this talk. If not, come to [2:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=179s) assumption of this talk. If not, come to [2:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=179s) assumption of this talk. If not, come to me afterwards. I'm intrigued what you're [3:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=181s) me afterwards. I'm intrigued what you're [3:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=181s) me afterwards. I'm intrigued what you're doing. Um, you may have you may have [3:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=183s) doing. Um, you may have you may have [3:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=184s) doing. Um, you may have you may have moved to to a world beyond this and [3:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=187s) moved to to a world beyond this and [3:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=187s) moved to to a world beyond this and definitely come and talk to me then. Um, [3:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=189s) definitely come and talk to me then. Um, [3:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=189s) definitely come and talk to me then. Um, so you know, we hear a lot of talk about [3:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=191s) so you know, we hear a lot of talk about [3:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=191s) so you know, we hear a lot of talk about software factory. Now, I don't know if [3:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=192s) software factory. Now, I don't know if [3:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=192s) software factory. Now, I don't know if that's the right terminology. You're [3:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=193s) that's the right terminology. You're [3:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=193s) that's the right terminology. You're also going to see our co-founder did a [3:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=195s) also going to see our co-founder did a [3:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=195s) also going to see our co-founder did a talk a couple of weeks ago about how we [3:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=197s) talk a couple of weeks ago about how we [3:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=197s) talk a couple of weeks ago about how we shouldn't call it a factory. I'm not [3:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=198s) shouldn't call it a factory. I'm not [3:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=198s) shouldn't call it a factory. I'm not going to harp on that particular piece. [3:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=200s) going to harp on that particular piece. [3:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=200s) going to harp on that particular piece. Um, but I do think that we are, you [3:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=203s) Um, but I do think that we are, you [3:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=203s) Um, but I do think that we are, you know, one of the important pieces here [3:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=205s) know, one of the important pieces here [3:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=205s) know, one of the important pieces here is that we're kind of quickly moving [3:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=207s) is that we're kind of quickly moving [3:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=207s) is that we're kind of quickly moving from a world where you're handdriving an [3:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=209s) from a world where you're handdriving an [3:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=209s) from a world where you're handdriving an agent on your local device um to [3:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=212s) agent on your local device um to [3:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=212s) agent on your local device um to orchestrating more agents or uh having [3:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=215s) orchestrating more agents or uh having [3:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=215s) orchestrating more agents or uh having them run in the cloud automatically and [3:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=217s) them run in the cloud automatically and [3:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=217s) them run in the cloud automatically and things like that. And so that that does [3:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=219s) things like that. And so that that does [3:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=219s) things like that. And so that that does uh change how it does change how you use [3:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=223s) uh change how it does change how you use [3:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=223s) uh change how it does change how you use the context, what context you need. [3:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=226s) the context, what context you need. [3:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=226s) the context, what context you need. Um, yeah, if we look forward like this [3:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=229s) Um, yeah, if we look forward like this [3:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=229s) Um, yeah, if we look forward like this this this is just the beginning, right? [3:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=231s) this this is just the beginning, right? [3:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=231s) this this is just the beginning, right? Like that that speed that we've seen, [3:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=234s) Like that that speed that we've seen, [3:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=234s) Like that that speed that we've seen, it's we're still at the very very [3:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=237s) it's we're still at the very very [3:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=237s) it's we're still at the very very beginning. There's a lot of folks on [3:58](https://www.youtube.com/watch?v=Xw9utONfhIY&t=238s) beginning. There's a lot of folks on [3:58](https://www.youtube.com/watch?v=Xw9utONfhIY&t=238s) beginning. There's a lot of folks on Twitter that are like naysayers saying [4:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=240s) Twitter that are like naysayers saying [4:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=240s) Twitter that are like naysayers saying that, oh, the free ride's over. It's all [4:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=242s) that, oh, the free ride's over. It's all [4:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=242s) that, oh, the free ride's over. It's all going to go up from here. Like they were [4:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=243s) going to go up from here. Like they were [4:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=243s) going to go up from here. Like they were just subsidizing. I I don't believe in [4:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=245s) just subsidizing. I I don't believe in [4:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=245s) just subsidizing. I I don't believe in that at all. Um, the cost is going to [4:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=248s) that at all. Um, the cost is going to [4:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=248s) that at all. Um, the cost is going to keep dropping. like this this graph just [4:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=250s) keep dropping. like this this graph just [4:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=250s) keep dropping. like this this graph just is just an example of you know three [4:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=253s) is just an example of you know three [4:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=253s) is just an example of you know three years ago what the intelligence that [4:14](https://www.youtube.com/watch?v=Xw9utONfhIY&t=254s) years ago what the intelligence that [4:14](https://www.youtube.com/watch?v=Xw9utONfhIY&t=254s) years ago what the intelligence that would have cost you $100 now costs you [4:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=256s) would have cost you $100 now costs you [4:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=256s) would have cost you $100 now costs you five cents. Um so it's it's you can [4:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=260s) five cents. Um so it's it's you can [4:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=260s) five cents. Um so it's it's you can extrapolate that into the future and [4:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=262s) extrapolate that into the future and [4:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=262s) extrapolate that into the future and kind of get very interesting results. Uh [4:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=264s) kind of get very interesting results. Uh [4:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=264s) kind of get very interesting results. Uh in a few years like ultraast will be [4:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=266s) in a few years like ultraast will be [4:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=266s) in a few years like ultraast will be common. You'll you'll snap your fingers [4:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=267s) common. You'll you'll snap your fingers [4:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=267s) common. You'll you'll snap your fingers and a features worth of code will [4:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=269s) and a features worth of code will [4:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=269s) and a features worth of code will appear. [4:31](https://www.youtube.com/watch?v=Xw9utONfhIY&t=271s) appear. [4:31](https://www.youtube.com/watch?v=Xw9utONfhIY&t=271s) appear. Um and then on top of that like task [4:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=273s) Um and then on top of that like task [4:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=273s) Um and then on top of that like task horizon is exponential too. So excuse [4:36](https://www.youtube.com/watch?v=Xw9utONfhIY&t=276s) horizon is exponential too. So excuse [4:36](https://www.youtube.com/watch?v=Xw9utONfhIY&t=276s) horizon is exponential too. So excuse the graph. pitch. Pitch is great [4:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=278s) the graph. pitch. Pitch is great [4:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=278s) the graph. pitch. Pitch is great software, but it does not let you um it [4:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=280s) software, but it does not let you um it [4:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=280s) software, but it does not let you um it does not let you do an exponential [4:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=282s) does not let you do an exponential [4:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=282s) does not let you do an exponential access, but it makes it look more [4:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=283s) access, but it makes it look more [4:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=283s) access, but it makes it look more impressive. Um so, you know, the [4:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=286s) impressive. Um so, you know, the [4:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=286s) impressive. Um so, you know, the important thing here is we we now [4:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=288s) important thing here is we we now [4:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=288s) important thing here is we we now leaving these coding agents running [4:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=289s) leaving these coding agents running [4:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=289s) leaving these coding agents running longer and longer. So, think back to [4:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=291s) longer and longer. So, think back to [4:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=291s) longer and longer. So, think back to 2024, you had a coding agent, but you [4:53](https://www.youtube.com/watch?v=Xw9utONfhIY&t=293s) 2024, you had a coding agent, but you [4:53](https://www.youtube.com/watch?v=Xw9utONfhIY&t=293s) 2024, you had a coding agent, but you were definitely kind of very in the [4:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=295s) were definitely kind of very in the [4:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=295s) were definitely kind of very in the loop, tight loop there, right? Um [4:58](https://www.youtube.com/watch?v=Xw9utONfhIY&t=298s) loop, tight loop there, right? Um [4:58](https://www.youtube.com/watch?v=Xw9utONfhIY&t=298s) loop, tight loop there, right? Um and now these things can run for hours, [5:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=301s) and now these things can run for hours, [5:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=301s) and now these things can run for hours, days. You hear you hear folks leaving [5:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=303s) days. You hear you hear folks leaving [5:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=303s) days. You hear you hear folks leaving things running over the weekend, having [5:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=305s) things running over the weekend, having [5:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=305s) things running over the weekend, having success with it. I don't know how much [5:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=306s) success with it. I don't know how much [5:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=306s) success with it. I don't know how much truth there is in that, how much of that [5:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=307s) truth there is in that, how much of that [5:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=308s) truth there is in that, how much of that code actually ships. Um, but no doubt [5:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=311s) code actually ships. Um, but no doubt [5:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=311s) code actually ships. Um, but no doubt it's increasing and will continue to do [5:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=313s) it's increasing and will continue to do [5:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=313s) it's increasing and will continue to do so. So why am I mentioning all this? [5:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=315s) so. So why am I mentioning all this? [5:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=315s) so. So why am I mentioning all this? Like what what's this got to do with [5:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=317s) Like what what's this got to do with [5:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=317s) Like what what's this got to do with agent context? Um, I I think one of the [5:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=320s) agent context? Um, I I think one of the [5:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=320s) agent context? Um, I I think one of the main reasons is that the context kind of [5:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=322s) main reasons is that the context kind of [5:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=322s) main reasons is that the context kind of moves out of that inner loop and largely [5:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=325s) moves out of that inner loop and largely [5:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=325s) moves out of that inner loop and largely into the outer loop. [5:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=327s) into the outer loop. [5:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=327s) into the outer loop. If we have like super fast systems [5:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=329s) If we have like super fast systems [5:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=329s) If we have like super fast systems completing tasks for sense and running [5:31](https://www.youtube.com/watch?v=Xw9utONfhIY&t=331s) completing tasks for sense and running [5:31](https://www.youtube.com/watch?v=Xw9utONfhIY&t=331s) completing tasks for sense and running for hours or days, you're not going to [5:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=333s) for hours or days, you're not going to [5:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=333s) for hours or days, you're not going to have humans manually delegating and [5:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=335s) have humans manually delegating and [5:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=335s) have humans manually delegating and starting and stopping them. Um, so we [5:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=338s) starting and stopping them. Um, so we [5:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=338s) starting and stopping them. Um, so we need to start building systems to manage [5:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=340s) need to start building systems to manage [5:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=340s) need to start building systems to manage this chaos and manage the context. [5:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=347s) So yeah, that this sounds great, right? [5:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=348s) So yeah, that this sounds great, right? [5:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=348s) So yeah, that this sounds great, right? Like we're going to have tons of code [5:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=351s) Like we're going to have tons of code [5:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=351s) Like we're going to have tons of code produced. It's going to be really cheap. [5:53](https://www.youtube.com/watch?v=Xw9utONfhIY&t=353s) produced. It's going to be really cheap. [5:53](https://www.youtube.com/watch?v=Xw9utONfhIY&t=353s) produced. It's going to be really cheap. We're going to build software factories. [5:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=354s) We're going to build software factories. [5:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=354s) We're going to build software factories. we're going to pump it out like features [5:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=356s) we're going to pump it out like features [5:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=356s) we're going to pump it out like features coming out left, right, and center. Um, [5:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=359s) coming out left, right, and center. Um, [5:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=359s) coming out left, right, and center. Um, and I think like to to be a little trit [6:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=361s) and I think like to to be a little trit [6:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=361s) and I think like to to be a little trit like you know code code is not the [6:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=363s) like you know code code is not the [6:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=363s) like you know code code is not the bottleneck now. Implementation is not [6:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=365s) bottleneck now. Implementation is not [6:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=365s) bottleneck now. Implementation is not the bottleneck now. Um, the failure mode [6:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=368s) the bottleneck now. Um, the failure mode [6:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=368s) the bottleneck now. Um, the failure mode here isn't an inability to build. Um, [6:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=371s) here isn't an inability to build. Um, [6:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=371s) here isn't an inability to build. Um, the failure mode is like an illability [6:14](https://www.youtube.com/watch?v=Xw9utONfhIY&t=374s) the failure mode is like an illability [6:14](https://www.youtube.com/watch?v=Xw9utONfhIY&t=374s) the failure mode is like an illability inability to maintain coherence. Um, to [6:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=376s) inability to maintain coherence. Um, to [6:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=376s) inability to maintain coherence. Um, to have a product that makes sense as a [6:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=378s) have a product that makes sense as a [6:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=378s) have a product that makes sense as a whole. Um what makes a product great is [6:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=382s) whole. Um what makes a product great is [6:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=382s) whole. Um what makes a product great is something something we care about a lot [6:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=383s) something something we care about a lot [6:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=383s) something something we care about a lot and we something that we are trying to [6:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=385s) and we something that we are trying to [6:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=385s) and we something that we are trying to build a product that helps you build [6:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=388s) build a product that helps you build [6:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=388s) build a product that helps you build great product too. So it's very meta. Um [6:31](https://www.youtube.com/watch?v=Xw9utONfhIY&t=391s) great product too. So it's very meta. Um [6:31](https://www.youtube.com/watch?v=Xw9utONfhIY&t=391s) great product too. So it's very meta. Um but what makes a product great is [6:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=392s) but what makes a product great is [6:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=392s) but what makes a product great is consistency like restraint uh tradeoffs [6:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=395s) consistency like restraint uh tradeoffs [6:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=395s) consistency like restraint uh tradeoffs and like lots of tiny decisions that [6:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=397s) and like lots of tiny decisions that [6:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=397s) and like lots of tiny decisions that reinforce one another. This was a codeex [6:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=400s) reinforce one another. This was a codeex [6:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=400s) reinforce one another. This was a codeex generated image of a [laughter] a [6:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=403s) generated image of a [laughter] a [6:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=403s) generated image of a [laughter] a settings page that's gone a bit crazy. [6:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=404s) settings page that's gone a bit crazy. [6:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=404s) settings page that's gone a bit crazy. Uh don't look too closely at it. Um, [6:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=408s) Uh don't look too closely at it. Um, [6:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=408s) Uh don't look too closely at it. Um, but it's a good example, right? Like [6:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=409s) but it's a good example, right? Like [6:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=409s) but it's a good example, right? Like just add another setting. You know, what [6:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=411s) just add another setting. You know, what [6:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=411s) just add another setting. You know, what kind of harm can it do? Um, [6:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=417s) so you know, a coding a coding agent [7:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=420s) so you know, a coding a coding agent [7:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=420s) so you know, a coding a coding agent today will will read the code, but it [7:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=422s) today will will read the code, but it [7:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=422s) today will will read the code, but it can still and that's great like there is [7:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=424s) can still and that's great like there is [7:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=424s) can still and that's great like there is a lot of value in the code obviously. [7:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=426s) a lot of value in the code obviously. [7:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=426s) a lot of value in the code obviously. Um, but it can still miss a lot of the [7:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=429s) Um, but it can still miss a lot of the [7:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=429s) Um, but it can still miss a lot of the context behind it. Um so some examples [7:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=433s) context behind it. Um so some examples [7:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=433s) context behind it. Um so some examples would be like um the intent the intent [7:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=436s) would be like um the intent the intent [7:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=436s) would be like um the intent the intent what what problem are we solving here [7:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=438s) what what problem are we solving here [7:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=438s) what what problem are we solving here why does this matter uh why does this [7:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=440s) why does this matter uh why does this [7:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=440s) why does this matter uh why does this matter to the end user um which [7:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=442s) matter to the end user um which [7:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=442s) matter to the end user um which decisions were made previously what were [7:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=445s) decisions were made previously what were [7:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=445s) decisions were made previously what were the tradeoffs that resulted in this [7:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=446s) the tradeoffs that resulted in this [7:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=446s) the tradeoffs that resulted in this design or in this feature you know um [7:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=450s) design or in this feature you know um [7:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=450s) design or in this feature you know um how should this look like visually [7:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=452s) how should this look like visually [7:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=452s) how should this look like visually outside of the the code uh how should [7:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=455s) outside of the the code uh how should [7:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=455s) outside of the the code uh how should this feel like in the product um maybe [7:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=457s) this feel like in the product um maybe [7:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=458s) this feel like in the product um maybe things like is there other work underway [7:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=460s) things like is there other work underway [7:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=460s) things like is there other work underway in other parts of the organization. Um, [7:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=462s) in other parts of the organization. Um, [7:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=462s) in other parts of the organization. Um, and like who who is the decision maker [7:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=464s) and like who who is the decision maker [7:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=464s) and like who who is the decision maker on this? Who's leading this project? [7:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=466s) on this? Who's leading this project? [7:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=466s) on this? Who's leading this project? Who's ultimately responsible for the [7:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=468s) Who's ultimately responsible for the [7:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=468s) Who's ultimately responsible for the decisions that happen here? And then are [7:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=470s) decisions that happen here? And then are [7:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=470s) decisions that happen here? And then are there any like additional risks? Like [7:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=472s) there any like additional risks? Like [7:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=472s) there any like additional risks? Like you can see some things from the code, [7:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=474s) you can see some things from the code, [7:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=474s) you can see some things from the code, right? But you can't always see oh um [7:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=477s) right? But you can't always see oh um [7:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=477s) right? But you can't always see oh um like a revert happened two weeks ago in [7:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=479s) like a revert happened two weeks ago in [7:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=479s) like a revert happened two weeks ago in this area of the codebase may not be [8:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=481s) this area of the codebase may not be [8:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=481s) this area of the codebase may not be immediately apparent. [8:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=483s) immediately apparent. [8:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=483s) immediately apparent. So the idea here is that we carry this [8:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=486s) So the idea here is that we carry this [8:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=486s) So the idea here is that we carry this uh we carry this context through um from [8:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=490s) uh we carry this context through um from [8:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=490s) uh we carry this context through um from the initial idea um through planning, [8:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=493s) the initial idea um through planning, [8:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=493s) the initial idea um through planning, implementation and review. [8:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=496s) implementation and review. [8:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=496s) implementation and review. So um I'll I'll talk a little bit about [8:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=498s) So um I'll I'll talk a little bit about [8:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=498s) So um I'll I'll talk a little bit about our system for doing that. Um [8:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=502s) our system for doing that. Um [8:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=502s) our system for doing that. Um most of this is in production in some [8:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=503s) most of this is in production in some [8:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=503s) most of this is in production in some form and most of it's in production in a [8:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=505s) form and most of it's in production in a [8:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=505s) form and most of it's in production in a way that other people can use too. Um [8:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=508s) way that other people can use too. Um [8:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=508s) way that other people can use too. Um this this isn't a linear ad. You you can [8:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=510s) this this isn't a linear ad. You you can [8:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=510s) this this isn't a linear ad. You you can use linear for this of course, but the [8:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=512s) use linear for this of course, but the [8:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=512s) use linear for this of course, but the principles apply broadly and you can you [8:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=514s) principles apply broadly and you can you [8:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=514s) principles apply broadly and you can you can build this into any system that [8:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=515s) can build this into any system that [8:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=515s) can build this into any system that you're using. And it's it's agents all [8:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=518s) you're using. And it's it's agents all [8:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=518s) you're using. And it's it's agents all the way down, I'm afraid. Um this is one [8:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=521s) the way down, I'm afraid. Um this is one [8:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=521s) the way down, I'm afraid. Um this is one of the reasons I'm I'm bullish on AI as [8:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=523s) of the reasons I'm I'm bullish on AI as [8:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=523s) of the reasons I'm I'm bullish on AI as a whole. Like every time you deeply dive [8:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=526s) a whole. Like every time you deeply dive [8:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=526s) a whole. Like every time you deeply dive into a problem, the solution is more AI. [8:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=529s) into a problem, the solution is more AI. [8:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=529s) into a problem, the solution is more AI. Um and it's definitely the case here. [8:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=536s) Um so the first thing we did is automate [9:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=540s) Um so the first thing we did is automate [9:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=540s) Um so the first thing we did is automate is automate intake. Um, you have, you [9:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=544s) is automate intake. Um, you have, you [9:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=544s) is automate intake. Um, you have, you know, as if you're developing a product, [9:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=546s) know, as if you're developing a product, [9:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=546s) know, as if you're developing a product, you have ideas coming in from Zenesk, [9:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=549s) you have ideas coming in from Zenesk, [9:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=549s) you have ideas coming in from Zenesk, Slack, Teams, Gong, Twitter, your [9:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=552s) Slack, Teams, Gong, Twitter, your [9:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=552s) Slack, Teams, Gong, Twitter, your internal product team, your CEO. All of [9:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=556s) internal product team, your CEO. All of [9:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=556s) internal product team, your CEO. All of these ideas kind of need to be [9:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=557s) these ideas kind of need to be [9:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=557s) these ideas kind of need to be collected. Um, we feed it into this [9:19](https://www.youtube.com/watch?v=Xw9utONfhIY&t=559s) collected. Um, we feed it into this [9:19](https://www.youtube.com/watch?v=Xw9utONfhIY&t=559s) collected. Um, we feed it into this concept called triage. Um, data dog [9:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=562s) concept called triage. Um, data dog [9:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=562s) concept called triage. Um, data dog alerts, like this, there's so many [9:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=564s) alerts, like this, there's so many [9:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=564s) alerts, like this, there's so many things. I this is like one fifth of it. [9:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=568s) things. I this is like one fifth of it. [9:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=568s) things. I this is like one fifth of it. Um, then we have something called a [9:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=569s) Um, then we have something called a [9:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=569s) Um, then we have something called a triage agent. Um, it it labels, it [9:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=573s) triage agent. Um, it it labels, it [9:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=574s) triage agent. Um, it it labels, it assigns and [snorts] groups the incoming [9:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=575s) assigns and [snorts] groups the incoming [9:36](https://www.youtube.com/watch?v=Xw9utONfhIY&t=576s) assigns and [snorts] groups the incoming work into themes. Uh, it closes [9:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=578s) work into themes. Uh, it closes [9:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=578s) work into themes. Uh, it closes duplicates. Uh, it roots things to the [9:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=581s) duplicates. Uh, it roots things to the [9:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=581s) duplicates. Uh, it roots things to the correct people and the correct teams and [9:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=583s) correct people and the correct teams and [9:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=584s) correct people and the correct teams and it surfaces decisions. Um, when more [9:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=587s) it surfaces decisions. Um, when more [9:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=587s) it surfaces decisions. Um, when more input might be needed to clarify the [9:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=589s) input might be needed to clarify the [9:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=589s) input might be needed to clarify the problem. [9:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=591s) problem. [9:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=591s) problem. Um, it also investigates the this is [9:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=594s) Um, it also investigates the this is [9:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=594s) Um, it also investigates the this is this is part of like the investigation [9:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=595s) this is part of like the investigation [9:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=595s) this is part of like the investigation that it does. Um and when the [9:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=597s) that it does. Um and when the [9:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=597s) that it does. Um and when the investigation is done, it's it's left on [9:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=599s) investigation is done, it's it's left on [9:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=599s) investigation is done, it's it's left on the issue in our case. Um issues are [10:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=601s) the issue in our case. Um issues are [10:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=601s) the issue in our case. Um issues are kind of the center of our world still. [10:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=603s) kind of the center of our world still. [10:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=603s) kind of the center of our world still. Um [10:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=605s) Um [10:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=605s) Um and it's left us detailed context for [10:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=607s) and it's left us detailed context for [10:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=607s) and it's left us detailed context for the next human or the agent to use. So [10:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=609s) the next human or the agent to use. So [10:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=609s) the next human or the agent to use. So by the end of that investigation, we [10:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=611s) by the end of that investigation, we [10:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=611s) by the end of that investigation, we have you know the problem that came in [10:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=613s) have you know the problem that came in [10:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=613s) have you know the problem that came in or the idea that came in, who requested [10:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=615s) or the idea that came in, who requested [10:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=615s) or the idea that came in, who requested it, it's been uh enriched with the [10:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=618s) it, it's been uh enriched with the [10:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=618s) it, it's been uh enriched with the customer information. Hey, this is a [10:19](https://www.youtube.com/watch?v=Xw9utONfhIY&t=619s) customer information. Hey, this is a [10:19](https://www.youtube.com/watch?v=Xw9utONfhIY&t=619s) customer information. Hey, this is a highv value customer. Um this is the [10:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=622s) highv value customer. Um this is the [10:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=622s) highv value customer. Um this is the third time they've requested this in a [10:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=623s) third time they've requested this in a [10:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=623s) third time they've requested this in a month. um any SL stack traces maybe if [10:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=626s) month. um any SL stack traces maybe if [10:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=626s) month. um any SL stack traces maybe if it's a bug logs from data dog that are [10:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=628s) it's a bug logs from data dog that are [10:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=628s) it's a bug logs from data dog that are associated um it's hooked up to your [10:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=630s) associated um it's hooked up to your [10:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=630s) associated um it's hooked up to your internal MCPs so you can have like [10:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=632s) internal MCPs so you can have like [10:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=632s) internal MCPs so you can have like internal account ids and things like [10:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=635s) internal account ids and things like [10:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=635s) internal account ids and things like that and then also related things that [10:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=638s) that and then also related things that [10:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=638s) that and then also related things that previously touched this area related [10:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=640s) previously touched this area related [10:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=640s) previously touched this area related issues and decisions that were made on [10:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=642s) issues and decisions that were made on [10:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=642s) issues and decisions that were made on those. [10:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=643s) those. [10:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=643s) those. This is all surfaced on the on the issue [10:45](https://www.youtube.com/watch?v=Xw9utONfhIY&t=645s) This is all surfaced on the on the issue [10:45](https://www.youtube.com/watch?v=Xw9utONfhIY&t=645s) This is all surfaced on the on the issue itself and kind of forms this this base [10:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=647s) itself and kind of forms this this base [10:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=647s) itself and kind of forms this this base context. Um so you can have something [10:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=650s) context. Um so you can have something [10:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=650s) context. Um so you can have something very simple that comes in like this [10:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=652s) very simple that comes in like this [10:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=652s) very simple that comes in like this crashes in this scenario and then by the [10:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=654s) crashes in this scenario and then by the [10:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=654s) crashes in this scenario and then by the time the triage agent is done you have [10:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=656s) time the triage agent is done you have [10:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=656s) time the triage agent is done you have uh a really a really filled out uh [10:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=659s) uh a really a really filled out uh [10:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=659s) uh a really a really filled out uh context. [11:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=662s) Um a few examples of where this was uh [11:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=665s) Um a few examples of where this was uh [11:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=665s) Um a few examples of where this was uh useful for us recently. Um and you [11:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=669s) useful for us recently. Um and you [11:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=669s) useful for us recently. Um and you probably can't read this. It doesn't [11:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=670s) probably can't read this. It doesn't [11:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=670s) probably can't read this. It doesn't matter too much but you know this is an [11:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=671s) matter too much but you know this is an [11:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=672s) matter too much but you know this is an example where the video captions kept [11:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=673s) example where the video captions kept [11:14](https://www.youtube.com/watch?v=Xw9utONfhIY&t=674s) example where the video captions kept getting turned uh was it on? Yeah it [11:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=676s) getting turned uh was it on? Yeah it [11:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=676s) getting turned uh was it on? Yeah it kept getting turned on. Um, and here [11:19](https://www.youtube.com/watch?v=Xw9utONfhIY&t=679s) kept getting turned on. Um, and here [11:19](https://www.youtube.com/watch?v=Xw9utONfhIY&t=679s) kept getting turned on. Um, and here like our uh triage agent went beyond [11:21](https://www.youtube.com/watch?v=Xw9utONfhIY&t=681s) like our uh triage agent went beyond [11:21](https://www.youtube.com/watch?v=Xw9utONfhIY&t=681s) like our uh triage agent went beyond finding just the setting or something [11:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=683s) finding just the setting or something [11:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=683s) finding just the setting or something like that. It actually reproduced the [11:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=685s) like that. It actually reproduced the [11:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=685s) like that. It actually reproduced the bug um and discovered that there was an [11:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=687s) bug um and discovered that there was an [11:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=687s) bug um and discovered that there was an automatic browser action that was being [11:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=689s) automatic browser action that was being [11:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=689s) automatic browser action that was being mistaken for user intent. Um, this level [11:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=693s) mistaken for user intent. Um, this level [11:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=693s) mistaken for user intent. Um, this level of detail left on the on the issue [11:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=695s) of detail left on the on the issue [11:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=695s) of detail left on the on the issue context it was key for in success of [11:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=699s) context it was key for in success of [11:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=699s) context it was key for in success of like automatically fixing the issue at [11:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=701s) like automatically fixing the issue at [11:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=701s) like automatically fixing the issue at this point. [11:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=703s) this point. [11:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=703s) this point. Um, here's another example. [11:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=707s) Um, here's another example. [11:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=707s) Um, here's another example. uh where the the research checked the [11:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=710s) uh where the the research checked the [11:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=710s) uh where the the research checked the existing codebase. It found some simil [11:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=712s) existing codebase. It found some simil [11:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=712s) existing codebase. It found some simil similar shipped to work. Made sure it [11:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=714s) similar shipped to work. Made sure it [11:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=714s) similar shipped to work. Made sure it wasn't a regression in this case, you [11:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=715s) wasn't a regression in this case, you [11:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=716s) wasn't a regression in this case, you know, is this a missing feature. Is this [11:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=717s) know, is this a missing feature. Is this [11:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=717s) know, is this a missing feature. Is this a bug? Um it found all the customers [12:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=720s) a bug? Um it found all the customers [12:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=720s) a bug? Um it found all the customers that are associated with this request. [12:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=722s) that are associated with this request. [12:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=722s) that are associated with this request. Uh re and realized it's high value. It [12:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=724s) Uh re and realized it's high value. It [12:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=724s) Uh re and realized it's high value. It also found the likely point in the [12:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=726s) also found the likely point in the [12:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=726s) also found the likely point in the codebase to get started. So, as you can [12:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=728s) codebase to get started. So, as you can [12:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=728s) codebase to get started. So, as you can imagine, this is just like a really [12:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=730s) imagine, this is just like a really [12:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=730s) imagine, this is just like a really great starting point for your coding [12:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=731s) great starting point for your coding [12:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=731s) great starting point for your coding agent at this point. Uh it probably [12:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=733s) agent at this point. Uh it probably [12:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=733s) agent at this point. Uh it probably massively increases the chances of [12:14](https://www.youtube.com/watch?v=Xw9utONfhIY&t=734s) massively increases the chances of [12:14](https://www.youtube.com/watch?v=Xw9utONfhIY&t=734s) massively increases the chances of success. [12:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=736s) success. [12:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=736s) success. So yeah, we have a pretty nice base [12:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=738s) So yeah, we have a pretty nice base [12:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=738s) So yeah, we have a pretty nice base context. Uh the next step is to build um [12:21](https://www.youtube.com/watch?v=Xw9utONfhIY&t=741s) context. Uh the next step is to build um [12:21](https://www.youtube.com/watch?v=Xw9utONfhIY&t=741s) context. Uh the next step is to build um either automatically from uh the triage [12:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=744s) either automatically from uh the triage [12:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=744s) either automatically from uh the triage agent deciding to uh deciding that the [12:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=747s) agent deciding to uh deciding that the [12:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=747s) agent deciding to uh deciding that the issue is quick enough. Uh in in our [12:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=749s) issue is quick enough. Uh in in our [12:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=749s) issue is quick enough. Uh in in our case, we we have a uh if the bug is [12:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=753s) case, we we have a uh if the bug is [12:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=753s) case, we we have a uh if the bug is small enough, we automatically fix it. [12:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=755s) small enough, we automatically fix it. [12:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=755s) small enough, we automatically fix it. But you could do this manually. It could [12:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=757s) But you could do this manually. It could [12:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=757s) But you could do this manually. It could be locally via MCP or it could be remote [12:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=759s) be locally via MCP or it could be remote [12:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=759s) be locally via MCP or it could be remote in the corral cloud. I'm not going to [12:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=761s) in the corral cloud. I'm not going to [12:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=761s) in the corral cloud. I'm not going to hop on this slide too much. I'm sorry if [12:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=763s) hop on this slide too much. I'm sorry if [12:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=763s) hop on this slide too much. I'm sorry if this is where you expected most of the [12:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=764s) this is where you expected most of the [12:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=764s) this is where you expected most of the talk to be. Um, but in my experience, [12:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=767s) talk to be. Um, but in my experience, [12:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=767s) talk to be. Um, but in my experience, this is kind of a solved problem today. [12:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=770s) this is kind of a solved problem today. [12:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=770s) this is kind of a solved problem today. Um, the most important thing here apart [12:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=772s) Um, the most important thing here apart [12:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=772s) Um, the most important thing here apart from the base input context that we just [12:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=775s) from the base input context that we just [12:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=775s) from the base input context that we just uh created is uh the ability to verify [13:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=780s) uh created is uh the ability to verify [13:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=780s) uh created is uh the ability to verify um and the ability for CI failures to [13:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=781s) um and the ability for CI failures to [13:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=781s) um and the ability for CI failures to feed back into the agent. So between [13:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=783s) feed back into the agent. So between [13:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=784s) feed back into the agent. So between those two pieces um I see a lot of [13:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=787s) those two pieces um I see a lot of [13:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=787s) those two pieces um I see a lot of success uh between the base the base [13:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=789s) success uh between the base the base [13:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=790s) success uh between the base the base context being able to verify and being [13:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=792s) context being able to verify and being [13:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=792s) context being able to verify and being able to see failures um we see a massive [13:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=795s) able to see failures um we see a massive [13:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=795s) able to see failures um we see a massive amount of success in our team. [13:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=803s) Um Laur Lauren at XAI probablyto on [13:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=806s) Um Laur Lauren at XAI probablyto on [13:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=806s) Um Laur Lauren at XAI probablyto on Twitter. She has a great set of skills [13:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=808s) Twitter. She has a great set of skills [13:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=808s) Twitter. She has a great set of skills you can use to uh generate um the [13:31](https://www.youtube.com/watch?v=Xw9utONfhIY&t=811s) you can use to uh generate um the [13:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=812s) you can use to uh generate um the verification skill for your coding [13:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=813s) verification skill for your coding [13:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=813s) verification skill for your coding agent. Um really makes a huge [13:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=815s) agent. Um really makes a huge [13:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=815s) agent. Um really makes a huge difference. [13:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=819s) Um so the next place where context [13:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=821s) Um so the next place where context [13:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=821s) Um so the next place where context matters and I I think it's still [13:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=824s) matters and I I think it's still [13:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=824s) matters and I I think it's still illconsidered is in review and it's also [13:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=826s) illconsidered is in review and it's also [13:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=826s) illconsidered is in review and it's also where it gets quite interesting. Like if [13:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=828s) where it gets quite interesting. Like if [13:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=828s) where it gets quite interesting. Like if you think about your setup today, what [13:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=830s) you think about your setup today, what [13:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=830s) you think about your setup today, what does it look like? Do you do you just [13:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=831s) does it look like? Do you do you just [13:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=831s) does it look like? Do you do you just have codecs running on your on your PRs? [13:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=834s) have codecs running on your on your PRs? [13:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=834s) have codecs running on your on your PRs? Uh Code Rabbit, Grappile, one of these [13:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=836s) Uh Code Rabbit, Grappile, one of these [13:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=836s) Uh Code Rabbit, Grappile, one of these products. Um what are they considering [13:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=839s) products. Um what are they considering [13:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=839s) products. Um what are they considering when they review your code? It's it's [14:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=842s) when they review your code? It's it's [14:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=842s) when they review your code? It's it's likely that they're getting all of their [14:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=843s) likely that they're getting all of their [14:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=843s) likely that they're getting all of their context from the codebase itself. Um and [14:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=846s) context from the codebase itself. Um and [14:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=846s) context from the codebase itself. Um and mostly focusing on logical correctness. [14:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=850s) mostly focusing on logical correctness. [14:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=850s) mostly focusing on logical correctness. [snorts] That's okay. That's good. The [14:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=852s) [snorts] That's okay. That's good. The [14:12](https://www.youtube.com/watch?v=Xw9utONfhIY&t=852s) [snorts] That's okay. That's good. The the coding agents, the the review agents [14:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=853s) the coding agents, the the review agents [14:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=853s) the coding agents, the the review agents are really really great. Um they catch a [14:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=855s) are really really great. Um they catch a [14:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=855s) are really really great. Um they catch a lot of bugs. [14:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=858s) lot of bugs. [14:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=858s) lot of bugs. But um at this point we make internally [14:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=862s) But um at this point we make internally [14:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=862s) But um at this point we make internally we make much more context available than [14:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=864s) we make much more context available than [14:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=864s) we make much more context available than just the codebase. Um we dispatch [14:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=866s) just the codebase. Um we dispatch [14:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=866s) just the codebase. Um we dispatch another agent. Hey agents. Uh we have a [14:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=869s) another agent. Hey agents. Uh we have a [14:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=869s) another agent. Hey agents. Uh we have a a review agent. Um it starts with that [14:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=872s) a review agent. Um it starts with that [14:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=872s) a review agent. Um it starts with that original issue um and that base context [14:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=874s) original issue um and that base context [14:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=874s) original issue um and that base context that we created with the customers and [14:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=875s) that we created with the customers and [14:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=875s) that we created with the customers and the original intent. Um and then it goes [14:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=878s) the original intent. Um and then it goes [14:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=878s) the original intent. Um and then it goes and looks for product history. So things [14:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=880s) and looks for product history. So things [14:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=880s) and looks for product history. So things like previous decisions uh reverts uh [14:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=884s) like previous decisions uh reverts uh [14:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=884s) like previous decisions uh reverts uh regressions and incidents in the same [14:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=886s) regressions and incidents in the same [14:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=886s) regressions and incidents in the same area. Um now we have code to compare [14:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=888s) area. Um now we have code to compare [14:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=888s) area. Um now we have code to compare against too, right? So we know which [14:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=890s) against too, right? So we know which [14:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=890s) against too, right? So we know which code has been touched. Um and we know [14:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=892s) code has been touched. Um and we know [14:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=892s) code has been touched. Um and we know the intent of the PR. We can we can uh [14:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=894s) the intent of the PR. We can we can uh [14:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=894s) the intent of the PR. We can we can uh further enhance enhance the context with [14:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=896s) further enhance enhance the context with [14:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=896s) further enhance enhance the context with this. Uh we also need to refine related [14:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=899s) this. Uh we also need to refine related [14:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=899s) this. Uh we also need to refine related issues. Hey, we now we now have the code [15:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=901s) issues. Hey, we now we now have the code [15:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=901s) issues. Hey, we now we now have the code context. Um this might be weeks later. [15:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=903s) context. Um this might be weeks later. [15:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=903s) context. Um this might be weeks later. It might be minutes later, but it might [15:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=904s) It might be minutes later, but it might [15:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=904s) It might be minutes later, but it might be a long time later too. Uh we also [15:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=907s) be a long time later too. Uh we also [15:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=907s) be a long time later too. Uh we also want to catch overlapping and [15:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=908s) want to catch overlapping and [15:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=908s) want to catch overlapping and conflicting changes. Um and we also have [15:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=911s) conflicting changes. Um and we also have [15:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=911s) conflicting changes. Um and we also have the full agent session. Um all of the [15:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=913s) the full agent session. Um all of the [15:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=913s) the full agent session. Um all of the decisions that were made in that. Maybe [15:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=915s) decisions that were made in that. Maybe [15:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=915s) decisions that were made in that. Maybe it was a oneshot. Maybe somebody went [15:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=917s) it was a oneshot. Maybe somebody went [15:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=917s) it was a oneshot. Maybe somebody went round and round um and made some [15:19](https://www.youtube.com/watch?v=Xw9utONfhIY&t=919s) round and round um and made some [15:19](https://www.youtube.com/watch?v=Xw9utONfhIY&t=919s) round and round um and made some critical decisions along the way. [15:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=922s) critical decisions along the way. [15:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=922s) critical decisions along the way. Um and finally, ownership tells us who [15:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=924s) Um and finally, ownership tells us who [15:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=924s) Um and finally, ownership tells us who should make the product tradeoff um and [15:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=928s) should make the product tradeoff um and [15:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=928s) should make the product tradeoff um and where review coverage may be missing. Uh [15:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=930s) where review coverage may be missing. Uh [15:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=930s) where review coverage may be missing. Uh you know, it's not always people that [15:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=932s) you know, it's not always people that [15:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=932s) you know, it's not always people that are in the codebase. It could be your [15:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=934s) are in the codebase. It could be your [15:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=934s) are in the codebase. It could be your product lead. It could be um your CEO. [15:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=937s) product lead. It could be um your CEO. [15:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=937s) product lead. It could be um your CEO. Uh could be your head of product. Uh [15:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=939s) Uh could be your head of product. Uh [15:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=939s) Uh could be your head of product. Uh that needs to be here too. [15:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=944s) Um, so now we're thinking like less [15:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=946s) Um, so now we're thinking like less [15:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=946s) Um, so now we're thinking like less about is this code correct and more [15:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=949s) about is this code correct and more [15:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=949s) about is this code correct and more about is this the right change for the [15:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=952s) about is this the right change for the [15:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=952s) about is this the right change for the product. Um, so I have a couple of [15:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=955s) product. Um, so I have a couple of [15:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=955s) product. Um, so I have a couple of concrete examples of this. Um, these are [15:58](https://www.youtube.com/watch?v=Xw9utONfhIY&t=958s) concrete examples of this. Um, these are [15:58](https://www.youtube.com/watch?v=Xw9utONfhIY&t=958s) concrete examples of this. Um, these are all within the last few weeks uh of of [16:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=961s) all within the last few weeks uh of of [16:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=961s) all within the last few weeks uh of of our product. Um, in this case, our [16:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=965s) our product. Um, in this case, our [16:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=965s) our product. Um, in this case, our review agent found several uh relevant [16:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=967s) review agent found several uh relevant [16:07](https://www.youtube.com/watch?v=Xw9utONfhIY&t=967s) review agent found several uh relevant changes from 6 months ago and 7 months [16:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=970s) changes from 6 months ago and 7 months [16:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=970s) changes from 6 months ago and 7 months ago. um where this was marked as an [16:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=975s) ago. um where this was marked as an [16:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=975s) ago. um where this was marked as an incident risk um due to holding a DB [16:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=978s) incident risk um due to holding a DB [16:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=978s) incident risk um due to holding a DB transaction open. Um this is actually [16:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=980s) transaction open. Um this is actually [16:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=980s) transaction open. Um this is actually really great like you know a review [16:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=983s) really great like you know a review [16:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=983s) really great like you know a review agent a standard review agent may have [16:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=986s) agent a standard review agent may have [16:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=986s) agent a standard review agent may have been able to find that potentially um [16:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=988s) been able to find that potentially um [16:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=988s) been able to find that potentially um but in this case the extra context [16:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=989s) but in this case the extra context [16:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=989s) but in this case the extra context really helped and codex did not find it. [16:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=993s) really helped and codex did not find it. [16:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=993s) really helped and codex did not find it. Um this is a pretty visceral example [16:36](https://www.youtube.com/watch?v=Xw9utONfhIY&t=996s) Um this is a pretty visceral example [16:36](https://www.youtube.com/watch?v=Xw9utONfhIY&t=996s) Um this is a pretty visceral example where we surface decisions made uh one [16:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=999s) where we surface decisions made uh one [16:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=999s) where we surface decisions made uh one year ago uh where we dropped some UI [16:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1002s) year ago uh where we dropped some UI [16:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1002s) year ago uh where we dropped some UI because it was busy. Um this is the kind [16:45](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1005s) because it was busy. Um this is the kind [16:45](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1005s) because it was busy. Um this is the kind of decision we make a lot in a product, [16:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1006s) of decision we make a lot in a product, [16:46](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1006s) of decision we make a lot in a product, right? Like ah this this doesn't feel [16:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1008s) right? Like ah this this doesn't feel [16:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1008s) right? Like ah this this doesn't feel right. This doesn't feel good enough. [16:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1010s) right. This doesn't feel good enough. [16:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1010s) right. This doesn't feel good enough. How do you get that back into the coding [16:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1012s) How do you get that back into the coding [16:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1012s) How do you get that back into the coding agents context? Uh in this case like a [16:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1015s) agents context? Uh in this case like a [16:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1015s) agents context? Uh in this case like a PR attempted to read that same [16:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1016s) PR attempted to read that same [16:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1016s) PR attempted to read that same iconography uh that had been rejected [16:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1019s) iconography uh that had been rejected [16:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1019s) iconography uh that had been rejected multiple times before. uh we were able [17:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1021s) multiple times before. uh we were able [17:01](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1021s) multiple times before. uh we were able to surface that and then of course these [17:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1024s) to surface that and then of course these [17:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1024s) to surface that and then of course these review comments get fed straight back [17:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1025s) review comments get fed straight back [17:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1025s) review comments get fed straight back into the coding agent um as additional [17:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1028s) into the coding agent um as additional [17:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1028s) into the coding agent um as additional context [17:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1030s) context [17:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1030s) context and uh one more here we we have the the [17:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1033s) and uh one more here we we have the the [17:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1033s) and uh one more here we we have the the concept of team and workspace labels in [17:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1036s) concept of team and workspace labels in [17:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1036s) concept of team and workspace labels in our product as a as a hierarchy um and [17:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1038s) our product as a as a hierarchy um and [17:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1038s) our product as a as a hierarchy um and in this case it looked like the [17:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1040s) in this case it looked like the [17:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1040s) in this case it looked like the workspace labels were being displayed [17:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1042s) workspace labels were being displayed [17:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1042s) workspace labels were being displayed below the team labels um I don't think a [17:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1045s) below the team labels um I don't think a [17:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1045s) below the team labels um I don't think a code code review uh tool would generally [17:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1048s) code code review uh tool would generally [17:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1048s) code code review uh tool would generally pick that up uh but in this case. Yeah, [17:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1050s) pick that up uh but in this case. Yeah, [17:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1050s) pick that up uh but in this case. Yeah, it had a previous revert and an issue um [17:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1053s) it had a previous revert and an issue um [17:33](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1053s) it had a previous revert and an issue um related to this and was able to surface [17:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1055s) related to this and was able to surface [17:35](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1055s) related to this and was able to surface that. [17:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1057s) that. [17:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1057s) that. So, this is where it's able to like [17:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1059s) So, this is where it's able to like [17:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1059s) So, this is where it's able to like maintain that sense of coherence. [17:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1062s) And there's no place in the code that [17:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1064s) And there's no place in the code that [17:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1064s) And there's no place in the code that defines that there's so many places that [17:45](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1065s) defines that there's so many places that [17:45](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1065s) defines that there's so many places that labels are listed, for example. [17:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1070s) Um we also built a UI around this that [17:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1072s) Um we also built a UI around this that [17:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1072s) Um we also built a UI around this that will guide a human reviewer through the [17:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1074s) will guide a human reviewer through the [17:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1074s) will guide a human reviewer through the change set uh and acts as a place to [17:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1076s) change set uh and acts as a place to [17:56](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1076s) change set uh and acts as a place to surface those decisions. [17:58](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1078s) surface those decisions. [17:58](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1078s) surface those decisions. um and places where the judgment calls [18:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1080s) um and places where the judgment calls [18:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1080s) um and places where the judgment calls are needed. You can use any review [18:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1082s) are needed. You can use any review [18:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1082s) are needed. You can use any review product for this, of course, but you [18:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1084s) product for this, of course, but you [18:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1084s) product for this, of course, but you know, we we we want to build tools. [18:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1086s) know, we we we want to build tools. [18:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1086s) know, we we we want to build tools. We're we're a big fan of like um [18:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1089s) We're we're a big fan of like um [18:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1089s) We're we're a big fan of like um handcrafted tools and we wanted to make [18:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1091s) handcrafted tools and we wanted to make [18:11](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1091s) handcrafted tools and we wanted to make the the best one we could. So, we built [18:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1093s) the the best one we could. So, we built [18:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1093s) the the best one we could. So, we built this surface. [18:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1096s) Um and then the final stage is like [18:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1098s) Um and then the final stage is like [18:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1098s) Um and then the final stage is like distilling that feedback. So, along the [18:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1100s) distilling that feedback. So, along the [18:20](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1100s) distilling that feedback. So, along the way, these comments have been stored um [18:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1102s) way, these comments have been stored um [18:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1102s) way, these comments have been stored um on issues on PRs. Uh changes have been [18:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1105s) on issues on PRs. Uh changes have been [18:25](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1105s) on issues on PRs. Uh changes have been merged into the codebase. a lot of [18:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1107s) merged into the codebase. a lot of [18:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1107s) merged into the codebase. a lot of context has built up. [18:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1110s) context has built up. [18:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1110s) context has built up. Um this this is like multiplied at the [18:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1112s) Um this this is like multiplied at the [18:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1112s) Um this this is like multiplied at the organizational level not just in the [18:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1114s) organizational level not just in the [18:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1114s) organizational level not just in the code. Um [18:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1117s) code. Um [18:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1117s) code. Um this gives like the human um corrective [18:39](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1119s) this gives like the human um corrective [18:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1120s) this gives like the human um corrective influence beyond the the current change. [18:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1123s) influence beyond the the current change. [18:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1123s) influence beyond the the current change. Hopefully we spend less time making the [18:45](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1125s) Hopefully we spend less time making the [18:45](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1125s) Hopefully we spend less time making the same correction twice. [18:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1128s) same correction twice. [18:48](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1128s) same correction twice. And then on top of those changes that [18:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1130s) And then on top of those changes that [18:50](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1130s) And then on top of those changes that end up back in the code, end up in the [18:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1131s) end up back in the code, end up in the [18:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1131s) end up back in the code, end up in the issue tracker, um end up in the pull [18:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1134s) issue tracker, um end up in the pull [18:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1134s) issue tracker, um end up in the pull requests, we um also have a a specific [18:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1137s) requests, we um also have a a specific [18:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1137s) requests, we um also have a a specific loop setup. This is an automation that [18:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1139s) loop setup. This is an automation that [18:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1139s) loop setup. This is an automation that runs once a day, but it could uh could [19:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1142s) runs once a day, but it could uh could [19:02](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1142s) runs once a day, but it could uh could be less or more depending on your PR [19:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1144s) be less or more depending on your PR [19:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1144s) be less or more depending on your PR volume that will troll through all of [19:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1146s) volume that will troll through all of [19:06](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1146s) volume that will troll through all of the code that has been merged in that [19:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1148s) the code that has been merged in that [19:08](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1148s) the code that has been merged in that day. [19:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1150s) day. [19:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1150s) day. Um it will find actionable lessons uh [19:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1153s) Um it will find actionable lessons uh [19:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1153s) Um it will find actionable lessons uh places where [19:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1155s) places where [19:15](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1155s) places where um these types of comments happened in [19:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1157s) um these types of comments happened in [19:17](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1157s) um these types of comments happened in the PRs and changes were made or reverts [19:19](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1159s) the PRs and changes were made or reverts [19:19](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1159s) the PRs and changes were made or reverts happened. Um extract learnings and then [19:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1162s) happened. Um extract learnings and then [19:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1162s) happened. Um extract learnings and then inject them back into the codebase um [19:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1164s) inject them back into the codebase um [19:24](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1164s) inject them back into the codebase um either in the agents.mmd or in the [19:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1166s) either in the agents.mmd or in the [19:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1166s) either in the agents.mmd or in the related documentation [19:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1168s) related documentation [19:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1168s) related documentation um just to make sure that those [19:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1170s) um just to make sure that those [19:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1170s) um just to make sure that those learnings are continuously compounding. [19:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1172s) learnings are continuously compounding. [19:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1172s) learnings are continuously compounding. Uh the exact prompt is something you you [19:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1174s) Uh the exact prompt is something you you [19:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1174s) Uh the exact prompt is something you you just have to experiment with and um [19:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1177s) just have to experiment with and um [19:37](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1177s) just have to experiment with and um obviously these changes aren't something [19:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1178s) obviously these changes aren't something [19:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1178s) obviously these changes aren't something you want to build up forever. You don't [19:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1180s) you want to build up forever. You don't [19:40](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1180s) you want to build up forever. You don't want your agents MD to end up like a [19:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1181s) want your agents MD to end up like a [19:41](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1181s) want your agents MD to end up like a million lines of context. So this is [19:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1183s) million lines of context. So this is [19:43](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1183s) million lines of context. So this is this is something that that you have to [19:45](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1185s) this is something that that you have to [19:45](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1185s) this is something that that you have to uh it's very Tom. [19:51](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1191s) This is this is a linear automation but [19:53](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1193s) This is this is a linear automation but [19:53](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1193s) This is this is a linear automation but you could use any automation product for [19:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1195s) you could use any automation product for [19:55](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1195s) you could use any automation product for it. Yeah. Um you know I think every [19:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1197s) it. Yeah. Um you know I think every [19:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1197s) it. Yeah. Um you know I think every product has has some sort of automation [19:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1199s) product has has some sort of automation [19:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1199s) product has has some sort of automation trigger at this point. [20:04](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1204s) So what does carrying the context [20:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1205s) So what does carrying the context [20:05](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1205s) So what does carrying the context through the whole workflow buy us? Um [20:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1209s) through the whole workflow buy us? Um [20:09](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1209s) through the whole workflow buy us? Um hopefully I've kind of described it [20:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1210s) hopefully I've kind of described it [20:10](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1210s) hopefully I've kind of described it right like fewer changes that solve the [20:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1213s) right like fewer changes that solve the [20:13](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1213s) right like fewer changes that solve the wrong problem to begin with. Um at [20:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1216s) wrong problem to begin with. Um at [20:16](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1216s) wrong problem to begin with. Um at intake understanding the intent helps us [20:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1218s) intake understanding the intent helps us [20:18](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1218s) intake understanding the intent helps us with that. Um during implementation the [20:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1222s) with that. Um during implementation the [20:22](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1222s) with that. Um during implementation the the product history and conventures help [20:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1223s) the product history and conventures help [20:23](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1223s) the product history and conventures help us uh build on decisions that we've made [20:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1226s) us uh build on decisions that we've made [20:26](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1226s) us uh build on decisions that we've made before. [20:27](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1227s) before. [20:28](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1228s) before. uh at review seeing the related work [20:29](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1229s) uh at review seeing the related work [20:30](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1230s) uh at review seeing the related work helps us catch conflicts um in the [20:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1232s) helps us catch conflicts um in the [20:32](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1232s) helps us catch conflicts um in the product and with ongoing work. I didn't [20:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1234s) product and with ongoing work. I didn't [20:34](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1234s) product and with ongoing work. I didn't give an example of that but it exists. [20:36](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1236s) give an example of that but it exists. [20:36](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1236s) give an example of that but it exists. Um and decides like whether the change [20:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1238s) Um and decides like whether the change [20:38](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1238s) Um and decides like whether the change like fits the product as a whole. Um the [20:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1242s) like fits the product as a whole. Um the [20:42](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1242s) like fits the product as a whole. Um the aim here is for to uh is for fewer [20:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1244s) aim here is for to uh is for fewer [20:44](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1244s) aim here is for to uh is for fewer avoidable mistakes less rework and with [20:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1247s) avoidable mistakes less rework and with [20:47](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1247s) avoidable mistakes less rework and with that we get faster inter faster [20:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1249s) that we get faster inter faster [20:49](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1249s) that we get faster inter faster iteration. Um and then at that final [20:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1252s) iteration. Um and then at that final [20:52](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1252s) iteration. Um and then at that final distillation loop uh that final [20:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1254s) distillation loop uh that final [20:54](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1254s) distillation loop uh that final automation you know human judgment that [20:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1257s) automation you know human judgment that [20:57](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1257s) automation you know human judgment that happened during this whole process uh [20:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1259s) happened during this whole process uh [20:59](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1259s) happened during this whole process uh becomes context for the next loop [21:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1260s) becomes context for the next loop [21:00](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1260s) becomes context for the next loop around. You know our our judgment [21:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1263s) around. You know our our judgment [21:03](https://www.youtube.com/watch?v=Xw9utONfhIY&t=1263s) around. You know our our judgment matters more as code gets cheaper.