# Hypermode Launch

Kevin Van Gundy, Hypermode

AI Engineer World's Fair 2024

[Talk page](https://ai.engineer/talks/8k848OeLc9E-hypermode-launch) · [Video](https://www.youtube.com/watch?v=8k848OeLc9E)

## Hypermode Launch

Rapid AI iteration depends on reversible choices: replaceable models, familiar APIs, observable inference, and fewer services to coordinate on the retrieval path.

### Learning faster above a pizzeria

[0:00](https://www.youtube.com/watch?v=8k848OeLc9E&t=0s)

Kevin Van Gundy recalls three problems from his time at Vercel, when the company had an office above a pizzeria. It was losing to other JavaScript frameworks, losing badly to other hosting providers, and he was losing to a diet consisting exclusively of pepperoni pizza.

![Three panels show figures carrying large blocks labeled JavaScript Frameworks and Hosting Providers, followed by a giant pepperoni pizza.](https://ai.engineer/api/talk-essay-media/0fe63a30ad857576229cacd6c20513ee4f4e3c354ce39a0e65e52d27277c6639)

Three opening problems: JavaScript frameworks, hosting providers, and pepperoni pizza.

The turnaround, in his telling, came from developing the ability to iterate quickly. Without knowing the optimum strategy, the team could try more things, discover what the market wanted, and adapt its products accordingly. “Iteration is the compound interest of software.” Repeated experiments eventually revealed two developer preferences: adopt new technology incrementally, and defer architectural commitments until the application's behavior is better understood.

### Make the wrong choice recoverable

[1:22](https://www.youtube.com/watch?v=8k848OeLc9E&t=82s)

That experimentation stops when developers are afraid of getting things wrong. Generative AI adds a long chain of decisions: hardware, model, prompt, integration, monitoring, and improvement. The consequences are familiar—runaway bills, hallucinating chatbots, and projects that consume months without delivering value. **A system that supports iteration must reduce the cost of mistakes**, because mistakes will happen repeatedly.

Van Gundy identifies three ways to make those decisions easier to revisit:

- **Model replacement:** choosing the wrong model becomes less consequential when switching it out introduces little friction.
- **Shared APIs:** classical software and AI components become easier to integrate when they use the same interfaces.
- **Inference tracing:** following inference step by step gives developers visibility when changing prompts, strategies, or the mix of data supplied to a model.

Together, these mechanisms make changes easier to attempt and their effects easier to inspect.

### Start with an AI function

[2:19](https://www.youtube.com/watch?v=8k848OeLc9E&t=139s)

Hypermode's proposed entry point is a **runtime** that integrates models and data into AI functions. The goal is to let developers add AI to an application without first acquiring specialized skills. Around that runtime, Hypermode supplies tools for rapid iteration and observation of functions in production. A team can begin with an incremental feature, develop its skills, and then reconsider the broader application as AI-native. The initial integration does not require making that larger architectural commitment upfront.

### Put model guidance in the editor

[2:56](https://www.youtube.com/watch?v=8k848OeLc9E&t=176s)

Even adding one model can involve unnecessary documentation work. Van Gundy uses **temperature** as the example: does a model expect a value from 0–2, 0–1, or 1–10? The developer needs that information while configuring the model. Hypermode's launch-era promise was typeahead in the developer's preferred editor, out of the box, with no SDKs or downloads. This describes the workflow presented at the launch, rather than current installation guidance. The editor popup on the slide illustrates bringing parameter documentation directly into that workflow.

![Black slide with the text “Hundreds of models. One interface.” and an editor popup showing temperature documentation.](https://ai.engineer/api/talk-essay-media/56022c68b257f38c00e67e486e83597bc359486ecce0b97a6bd4f5fd79eb1b7c)

“Hundreds of models. One interface.” beside editor assistance for temperature.

### Consolidate the retrieval path

[3:19](https://www.youtube.com/watch?v=8k848OeLc9E&t=199s)

For production, Hypermode offers strong defaults while also allowing teams to bring an existing stack. The concrete example is **retrieval-augmented generation**, or RAG: embedding an input and searching a vector store introduces additional service interactions into a previously simpler application request.

Van Gundy characterizes traditional RAG as requiring N+1 requests and says Hypermode consolidates that retrieval path into one request.

| Retrieval path | Coordination described in the launch |
| --- | --- |
| Traditional RAG | Additional input-embedding call, then vector-store access |
| Hypermode RAG | One request using an in-memory embedding and search service |

The consolidation concerns the embedding-and-search path; those operations still have to occur. Van Gundy reports a saving of a couple hundred milliseconds per request for this retrieval path. He supplies no benchmark configuration, baseline measurements, or latency percentile, so the magnitude cannot be assessed independently from the presentation.

![Three stacked diagrams compare a legacy search request, traditional RAG with multiple service connections, and a green Hypermode function connected to a product database.](https://ai.engineer/api/talk-essay-media/dd317edcfdc4f5c31fd84af638a36cb9ff77b8ca4342d2337c4f343ff8c0cb73)

Request paths before AI, with traditional RAG, and with Hypermode RAG.

### Build intuition by comparing outputs

[3:49](https://www.youtube.com/watch?v=8k848OeLc9E&t=229s)

Once the integration works, understanding its behavior remains difficult. Nondeterministic systems do not produce the same intuitive feedback as conventional software, and Van Gundy describes each model as having its own personality. Hypermode's comparison tools let developers examine different inferences, tunes, and models. The resulting dataset can then be exported for fine-tuning, connecting the work of inspecting outputs with a subsequent model-improvement step.

### From a workshop demo to a production plan

[4:06](https://www.youtube.com/watch?v=8k848OeLc9E&t=246s)

The closing question is practical: what will attendees tell their boss on Monday about what they learned? Van Gundy invites them to the after-lunch [Hypermode workshop](https://ai.engineer/worldsfair/2024/schedule), with the team that built the platform, to practice making iteration velocity a core competency. The proposed applications are natural-language search, intelligent sorting of product data lists, outlier detection, and identifying bad actors.

![Workshop slide lists natural language search, intelligently sorting every data list in an application, and outliers and bad actors detection, with a time of 1:00pm–1:50pm at GGB A.](https://ai.engineer/api/talk-essay-media/31d01c9297abdb123f899aa2b79479395dac485b9312779bfacdc17e407d2104)

Hypermode workshop applications: natural language search, intelligent sorting, and outlier detection.

The workshop promise was a demo attendees could be proud of and a plan to put something similar into production by the end of the following month. Van Gundy also offered $1,000 in Hypermode credits to help attendees get started. These were offers made at the 2024 event. The intended next step was to carry a concrete application beyond the workshop, with a production plan that made continued iteration possible.

## Resources

- [Hypermode launch and workshop schedule](https://ai.engineer/worldsfair/2024/schedule) — The 2024 conference schedule includes the launch keynote and Hypermode's afternoon workshop on adding AI features to applications.
- [Modus source repository](https://github.com/hypermodeinc/modus) — Hypermode's open-source framework for model-connected functions in Go and AssemblyScript, with setup instructions. Archived September 11, 2025.
- [Real-time Vector Search for Ecommerce](https://luma.com/sanjy86t) — A past Hypermode event outlining natural-language search with embeddings, in-memory vector search and custom reranking.

## Slides

### Presenting sponsor

Microsoft is the presenting sponsor.

[0:05](https://www.youtube.com/watch?v=8k848OeLc9E&t=5s) [Microsoft logo beneath the heading Presenting Sponsor.](https://assets.ai.engineer/sites/aiecode2025/sha256/9f227e03d5413a3f145480f3998b2dc3d19988d231e026e1f23d34ccc7102bb5.webp)

PRESENTING SPONSOR

Microsoft

### Innovation and platinum sponsors

AWS is the innovation sponsor; MongoDB, Google Cloud and Neo4j are platinum sponsors.

[0:10](https://www.youtube.com/watch?v=8k848OeLc9E&t=10s) [AWS is the innovation sponsor; MongoDB, Google Cloud and Neo4j are platinum sponsors.](https://assets.ai.engineer/sites/aiecode2025/sha256/630a2c443ecc7597c52e9d1aeb861e20a19a5346dd1d271a75cb61d0c3f5b66c.webp)

INNOVATION SPONSOR
aws

PLATINUM SPONSORS
MongoDB®
Google Cloud
neo4j

### Vercel and Tommaso’s

Before Hypermode, the speaker worked at Vercel, with an office above a pizzeria. The slide points to Tommaso’s in North Beach, San Francisco.

[0:16](https://www.youtube.com/watch?v=8k848OeLc9E&t=16s) [A Vercel logo points to a photograph of Tommaso’s in North Beach, San Francisco.](https://assets.ai.engineer/sites/aiecode2025/sha256/372ba236ab27e6fd9de0cf5b9937c5f6b9845eea02a74a942e9378f8c49b3768.webp)

Vercel

RISTORANTE
TOMMASO’S
PIZZERIA
1042. KEARNY STREET
TOMMASO’S

Tommaso’s - North Beach, SF

### Three problems at Vercel

Vercel was losing to competing JavaScript frameworks and hosting providers; the speaker jokes that he was also losing to his pepperoni pizza diet. He credits rapid iteration with helping Vercel start to win.

[0:28](https://www.youtube.com/watch?v=8k848OeLc9E&t=28s) [Three images illustrate JavaScript frameworks, hosting providers and pepperoni pizza as burdens.](https://assets.ai.engineer/sites/aiecode2025/sha256/a6ac6667c99dd32212c77c6e7b3c12fea81cbdf66f48ce3bbeb950257926e9aa.webp)

1. JavaScript Frameworks
2. Hosting Providers
3. Pepperoni Pizza 🍕

Vercel
Vercel

### Iteration is the compound interest of software

A curve starts nearly flat, then rises sharply. Rapid experiments revealed that developers want incremental technology adoption and avoid committing to architecture before they know how their application will work.

[0:58](https://www.youtube.com/watch?v=8k848OeLc9E&t=58s) [A curve rises sharply above the statement, “Iteration is the compound interest of software.”](https://assets.ai.engineer/sites/aiecode2025/sha256/9b4ad9794a95cb59c843ad9405c4bff7ae5193d943bb8be8551e08e38696d1ef.webp)

Iteration is the compound interest of software.

AIE
MongoDB
AI Engineer World’s Fair
OctoAI

[1:14](https://www.youtube.com/watch?v=8k848OeLc9E&t=74s) [A rising curve illustrates the compounding effect of iteration.](https://assets.ai.engineer/sites/aiecode2025/sha256/1e409ef7d219f8798453a0bbad7203da5aa1ad645c2cc8eec40faffe850bfa1a.webp)

Iteration is the compound interest of software.

AIE
MongoDB
AI Engineer World’s Fair
OctoAI
aws

### Fear of getting it wrong

Fear of mistakes can stop iteration. Through rapid experimentation, the speaker learned that developers want to adopt technologies incrementally and defer architectural choices until they understand their applications.

[1:20](https://www.youtube.com/watch?v=8k848OeLc9E&t=80s) [Iteration cannot happen when people are afraid of getting it wrong.](https://assets.ai.engineer/sites/aiecode2025/sha256/c638cb6424ab1664b3d343d429a6871a09a18ebd12b9d2f6957262bb0e70b6f4.webp)

Iteration can’t happen when you’re afraid of getting it wrong.

Interface text:
AIE
aws
neo4j
AI Engineer World’s Fair
Covalent.

### Things to get wrong with GenAI

GenAI teams face choices from models, prompts and data to API rollout, scaling and monitoring. The speaker cites runaway bills, hallucinating chatbots and projects that took months and months without delivering value.

[1:38](https://www.youtube.com/watch?v=8k848OeLc9E&t=98s) [GenAI challenges span building with models, methods, prompts, data and controls, and running APIs with scaling, rollout, performance, monitoring and debugging.](https://assets.ai.engineer/sites/aiecode2025/sha256/416c94a01246a72cad004cd60bd75c22f2418851167730dac50b7c1b4bd19f68.webp)

There’s a lot of things to get wrong with GenAI

Build?
Models, method, prompt, data, controls

Run?
Publish API, scaling, incremental roll-out, performance, monitoring, debugging

AIE
aws
neo4j
AI Engineer World’s Fair
Covalent.

### De-risking mistakes

Systems can reduce the cost of mistakes by making models easy to swap, using the same APIs for classical and AI systems, and tracing inference as prompts, strategies and data mixes change.

[1:53](https://www.youtube.com/watch?v=8k848OeLc9E&t=113s) [Slide states that systems should de-risk getting things wrong.](https://assets.ai.engineer/sites/aiecode2025/sha256/001d36c20e0b066f54977c1c1f84a0bdb4bc6f444736d47564407b628290f1ff.webp)

We need systems that de-risk getting it wrong.

AIE
Microsoft
smol ai

AI Engineer
[illegible]orld’s Fair
neptune.ai
[illegible]roq

### Hypermode

Hypermode aims to make AI approachable so attendees can add it to their apps without specialized skills.

[2:21](https://www.youtube.com/watch?v=8k848OeLc9E&t=141s) [Hypermode logo on a black background.](https://assets.ai.engineer/sites/aiecode2025/sha256/5da757021169deaa93537acdd0c645f004dd62852a975dc7c111ad3fe3a70bd7.webp)

Hypermode

Event branding:
AIE
aws
MongoDB
AI Engineer World’s Fair
DATASTAX
HASURA

### Hypermode function architecture

Hypermode’s runtime integrates models and data into AI functions. The diagram places a function between an application UI and a product database, with model and service icons connected below it.

[2:27](https://www.youtube.com/watch?v=8k848OeLc9E&t=147s) [Diagram connects an application interface to a Hypermode function and product database, with model and service icons beneath the function.](https://assets.ai.engineer/sites/aiecode2025/sha256/6fc67fac6342067802b3b050ab5a70d29da241e3f2ded4ed7f464a6c3b714d54.webp)

Hypermode Function
Product Database
Application UI

Interface text:
Where do you want to experience?
A-Frame Lake House
Sleeps up to 6
$130
Per Night
Additional Details
Book Now
AIE
aws
MongoDB
World’s Fair
OctoAI

### Hypermode dashboard

The Test Flight dashboard puts deployments, functions, models, and hosts in one view. Hypermode describes its tools as supporting iteration and observation of AI functions in production.

[2:38](https://www.youtube.com/watch?v=8k848OeLc9E&t=158s) [Hypermode dashboard showing deployments, functions, models, and hosts.](https://assets.ai.engineer/sites/aiecode2025/sha256/594cc10622800678af0b9ff1b1756c612c71e45f917259acd548cce9f252d475.webp)

Test Flight
Endpoint: testflight-hypermode.hypermode.app/graphql
GitHub: gohypermode/test-flight

Deployments
cleanup project for reaability
Deployed 1d ago
remove unused model
Deployed 6d ago
gpt 3.5-turbo
Deployed 12d ago
urgency classification
Deployed 12d ago
Update index.ts
Deployed 12d ago
View all deployments

Functions
classifyIssueType
classifyIssueUrgency

Models
type-classifier
AntoineMC/distilbart-mnli-github-issues
gpt-3-5-turbo
gpt-3.5-turbo
openai

Hosts
hypermode
Default for model hosting and collections
openai
Base URL: https://api.openai.com/

Interface text:
Hypermode
Hypermode / Test Flight
Support
Docs
Home
Query
Inferences
Function Runs
Deployments
Settings
EXPLORE
Models
AI Features

### Hundreds of models, one interface

Hypermode offers editor typeahead for model settings. The example documents temperature’s range, effects, and default alongside the code.

[3:09](https://www.youtube.com/watch?v=8k848OeLc9E&t=189s) [Slide introducing one interface for many models, with a code editor tooltip for temperature.](https://assets.ai.engineer/sites/aiecode2025/sha256/d6ac1be1053504829fc259eb41f07e525148c5a3e08e8562d2744c787d1debec.webp)

Hundreds of models.
One interface.

Interface text:
input.temp
// Here we [illegible]
const outp[illegible]
// The out[illegible]
// Here we [illegible]
return out[illegible]

temperature
(property) OpenAIChatInput.temperature: number

A number between 0.0 and 2.0 that controls the sampling temperature.

Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic.

We generally recommend altering this or topP but not both.

@default
1.0

### Comparing application and RAG request paths

The diagram contrasts traditional RAG’s n+1 requests across several services with Hypermode’s claimed single request through a function to the product database.

[3:31](https://www.youtube.com/watch?v=8k848OeLc9E&t=211s) [Diagram compares a legacy application request, a multi-request traditional RAG path, and a Hypermode RAG path.](https://assets.ai.engineer/sites/aiecode2025/sha256/72a98be1992c475716bb24ceb86861ae5d513d5af71caa70c9a09144ad75551f.webp)

Before AI

Stale results/recommendation
Legacy search system
Normal application request, no AI

Use of AI to make your application slower.

Create Embedding
Find Similar vectors
Retrieve products
Generate response recommendation
Log outputs for training

“Traditional RAG”
Visual of RAG request path needing n+1 requests.

Using AI to make your application better

Hypermode Function
Product Database

Hypermode RAG
auto in-memory embedding + vector

Interface text:
Application screenshots: [illegible]

### Debugging and fine-tuning with inference history

Hypermode’s inference history shows model metadata alongside each input and full output. The speaker says developers can compare inferences, tunes, and models, then export data for fine-tuning.

[3:56](https://www.youtube.com/watch?v=8k848OeLc9E&t=236s) [Slide describes inspecting model metadata, inputs, outputs, and exported inference history.](https://assets.ai.engineer/sites/aiecode2025/sha256/f49c4d272fe3241e85b956cf67c6955543fd6188a98d915b81a4f87d212e4cc5.webp)

AI is more than a model
Debug your functions
Fine-tune to success

./hypermode/inference-history

Model source, version, and host

Know exactly what’s flowing into each step in your chain of models

See the entire output, not just what’s passed along

Export inference history to audit, replay, or fine-tune

Interface text:
type-classifier inference from 14m ago
Timestamp: 2024-06-24T18:01:06.534Z
Duration: 112 ms
Model: type-classifier
Source Model: AntoineMC/distilbart-mnli-github-issues
Provider: hugging-face
Host: hypermode

Input
{
  "instances": [
    "Multiple instances of React - I had multiple instances of React by mistake. When trying to use hooks, got this error: ‘hooks can only be called inside the body of a function component’. Which is not correct since I was using function components. Took me a while to find the real cause of the issue."
  ]
}

Output
{
  "predictions": [
    {
      "label": "issue",
      "confidence": 0.6041929,
      "probabilities": [
        {
          "label": "issue",
          "probability": 0.6041929
        },
        {
          "label": "feature",
          "probability": 0.15605827
        },
        {
          "label": "question",
          "probability": 0.23974888
        }
      ]
    }
  ]
}

Export to CSV
Export to CSV (filtered inferences)
Export to JSONL
Export to JSONL (filtered inferences)

### The Hypermode Workshop

The 1:00–1:50pm workshop in GGB A will cover natural language search, sorting application data, and detecting outliers and bad actors. The speaker also offers attendees $1,000 in Hypermode credits.

[4:20](https://www.youtube.com/watch?v=8k848OeLc9E&t=260s) [Workshop slide lists natural language search, sorting application data, and detecting outliers and bad actors.](https://assets.ai.engineer/sites/aiecode2025/sha256/832a7898825198f8663272a065533b1ce44bcb19a4838349ab01e232e44de3d3.webp)

The Hypermode Workshop
Time 1:00pm - 1:50pm @ GGB A

Natural Language Search
Intelligently Sort Every Data List in Your Application
Outliers and Bad Actors Detection

...and much more

AIE
MongoDB
AI Engineer World’s Fair
OctoAI

[4:24](https://www.youtube.com/watch?v=8k848OeLc9E&t=264s) [Workshop time and location, with topics including natural language search, sorting application data, and detecting outliers and bad actors.](https://assets.ai.engineer/sites/aiecode2025/sha256/5965ba369ebd4bd01d1010b5e92b4677f97ef44b1a33dce32ff93dd00c7beed4.webp)

The Hypermode Workshop
Time 1:00pm - 1:50pm @ GGB A

Natural Language Search
Intelligently Sort Every Data List in Your Application
Outliers and Bad Actors Detection

...and much more

## Full transcript

[0:00](https://www.youtube.com/watch?v=8k848OeLc9E&t=0s) [on hold music] Before Hypermode, I worked at Vercel.

[0:16](https://www.youtube.com/watch?v=8k848OeLc9E&t=16s) We had an office down the street above a pizzeria, and we had three big problems. One, we were losing to other JavaScript frameworks. Two, we were losing badly to other hosting providers.

[0:29](https://www.youtube.com/watch?v=8k848OeLc9E&t=29s) And three, I was losing to my diet of exclusively pepperoni pizza. [laughing]

[0:35](https://www.youtube.com/watch?v=8k848OeLc9E&t=35s) Eventually, we started to win, not because we were smart or we knew all the right answers, but because we built this core competency of iterating really, really, really quickly.

[0:46](https://www.youtube.com/watch?v=8k848OeLc9E&t=46s) We didn't know the optimum strategy, but we figured if we just tried more things faster than everyone else, we'd eventually be able to adapt and figure out the right products and strategies to figure out what the market wanted.

[0:58](https://www.youtube.com/watch?v=8k848OeLc9E&t=58s) Iteration is the compound interest of software. Keep doing it long enough and eventually really good stuff starts to happen.

[1:05](https://www.youtube.com/watch?v=8k848OeLc9E&t=65s) Because we tried a lot of things really quickly, we eventually figured out two things. One, developers wanna incrementally adopt new technologies. And two, they don't wanna commit to architectural patterns before they know how their application's actually gonna work.

[1:22](https://www.youtube.com/watch?v=8k848OeLc9E&t=82s) But iteration can't happen if you're afraid of getting it wrong.

[1:26](https://www.youtube.com/watch?v=8k848OeLc9E&t=86s) The same thing that has held back web is also holding back AI. And if I'm honest, there are even more things for us to get wrong about GenAI. When I think about it, I'm grossly overwhelmed.

[1:37](https://www.youtube.com/watch?v=8k848OeLc9E&t=97s) What's the right hardware? What's the right w- model? What's the right prompt? How do I integrate? How do I monitor? How do I improve? Everyone here knows a horror story of someone with a runaway bill, a hallucinating chatbot, a project that took months and months and never delivered any value.

[1:54](https://www.youtube.com/watch?v=8k848OeLc9E&t=114s) And in the end, we need to build systems that de-risk getting it wrong

[1:59](https://www.youtube.com/watch?v=8k848OeLc9E&t=119s) because we are going to get it wrong a lot.

[2:03](https://www.youtube.com/watch?v=8k848OeLc9E&t=123s) Picking the wrong model doesn't matter if there's no friction to switching it out. Integration is simple when your classical systems and your AI systems use the same APIs. You can fearlessly make changes to prompts, strategies, data mixes if you can trace that inference step by step by step.

[2:19](https://www.youtube.com/watch?v=8k848OeLc9E&t=139s) At Hypermode, we care deeply about making AI approachable. Everyone here should be able to put AI in their apps without specialized skills.

[2:29](https://www.youtube.com/watch?v=8k848OeLc9E&t=149s) At its core, Hypermode is a runtime. It allows you to easily integrate models and data into AI functions.

[2:36](https://www.youtube.com/watch?v=8k848OeLc9E&t=156s) We then surround that runtime with a bunch of tools that make it easy for you to e- to rapidly iterate and observe those AI functions in prod. We make it easy to get started, incrementally adopt AI as appropriate, and then as your team develops those skills, reimagine those applications as AI-native.

[2:56](https://www.youtube.com/watch?v=8k848OeLc9E&t=176s) First and foremost, we wanted to make the developer experience of developing with AI a lot less terrible.

[3:02](https://www.youtube.com/watch?v=8k848OeLc9E&t=182s) When it comes to adding a new model to your service, you probably don't wanna read a bunch of pages of docs to figure out the temperature's on a zero to two rather than a zero to one or a one to ten.

[3:12](https://www.youtube.com/watch?v=8k848OeLc9E&t=192s) With Hypermode, we provide you typeahead in your favorite code editor right out of the box. No SDKs, nothing to download.

[3:19](https://www.youtube.com/watch?v=8k848OeLc9E&t=199s) Then when you do ship to prod, we give you strong defaults just to get started. Or if you have your own stack, bring it along. In either case, we'll remove a lot of that complexity for you.

[3:29](https://www.youtube.com/watch?v=8k848OeLc9E&t=209s) For example, traditional RAG requires N+1 requests. You need to make an additional call to embed the inputs, go talk to your vector store. With Hypermode, we c- you can do that all in one request.

[3:41](https://www.youtube.com/watch?v=8k848OeLc9E&t=221s) We build in-memory embedding and search service that allow you to do that and save a couple hundred milliseconds per request.

[3:49](https://www.youtube.com/watch?v=8k848OeLc9E&t=229s) Finally, building intuition around nondeterminix sys- nondeterministic systems is hard. Each model has its own personality, and we make it really easy for you to quickly compare different inferences, different tunes, different models, and you can then export this data set to fine-tune.

[4:06](https://www.youtube.com/watch?v=8k848OeLc9E&t=246s) On Monday, your boss is going to ask you, "What did you learn at AI World Fair?" If you come by our workshop after lunch, I'll prove to you that you can make AI...

[4:15](https://www.youtube.com/watch?v=8k848OeLc9E&t=255s) Uh, sorry. I'll prove to you that you can make iteration velocity a core competency. The team that built all this amazing stuff will be there. We'll show you how to build natural language search, intelligently sort every data list in your product, detect outliers, catch bad guys.

[4:29](https://www.youtube.com/watch?v=8k848OeLc9E&t=269s) You'll walk away with the demo that you're proud of and a plan to put something like it in prod by the end of next month. And if seeing my happy face again and building something really cool is not enough, we'll give you $1,000 in Hypermode credits to get started.

[4:42](https://www.youtube.com/watch?v=8k848OeLc9E&t=282s) Thank you all so much. [outro music]
