# Data for models and agents

A pack of 10 talks from the AI Engineer YouTube channel, in the order to watch them. 3h 01m of video.
Page: https://aietalks.com/packs/data-for-models-and-agents

Training data is not a pile of tokens waiting to become a model. It is a set of choices about what the system should learn, which mistakes count, and how closely a task resembles the world outside the lab. Useful examples have to be selected, visual variety preserved, teachers chosen and failures found before they disappear into scale. Agent environments make the same problem easier to see. Broken tasks reward the wrong behavior, synthetic users invent details unless they are grounded, and even a realistic sandbox leaves out parts of production. Opt-in product traces can repair the next training set, but only after filtering turns raw behavior into examples worth learning from.

## This pack is for you if

- You have plenty of raw data but no clear way to decide which examples deserve training compute.
- Your synthetic tasks look plausible, yet you cannot tell whether their failures teach a useful skill.
- You want production behavior to improve the next model without treating every trace as ground truth.

## The talks, in order

### 1. The Hierarchy of Needs for Training Dataset Development

Chang She, LanceDB & Noah Shpak, Character AI | 16:32 | AI Engineer World's Fair 2024
Video: https://www.youtube.com/watch?v=i2vBaFzCEJw
Summary: https://aietalks.com/talks/the-hierarchy-of-needs-for-training-dataset-development.md

Why first: She and Shpak treat dataset development as an engineering system rather than a one-off export. Cleaning, versioning, random access, streaming and analysis all have to work before anyone can make a defensible selection. Morcos follows by asking which examples should make it through that system.

### 2. Data Quality Is the Compute Multiplier

Ari Morcos, DatologyAI | 19:05 | AI Engineer World's Fair 2026
Video: https://www.youtube.com/watch?v=_PdK6x7PQNM
Summary: https://aietalks.com/talks/data-quality-is-the-compute-multiplier.md

Why second: She and Shpak build the data machinery. Morcos gives it a target: more information per token for the task that matters. His examples show curation changing training cost, serving cost and domain performance. Lee then makes the same point in a medium where a bad choice is visible at a glance.

### 3. Training Krea 2: What Matters in Generative Model Training

Sangwu Lee, Krea.ai | 21:46 | AI Engineer World's Fair 2026
Video: https://www.youtube.com/watch?v=-tviRdpmHvs
Summary: https://aietalks.com/talks/training-krea-2-what-matters-in-generative-model-training.md

Why here: Morcos argues that selection determines what a model can learn. Lee shows how selection also determines its taste. A caption that omits the white wall teaches the wrong visual rule, while aggressive quality filters erase useful styles. Marten takes that experimental mindset into reasoning data.

### 4. OpenThoughts: Data Recipes for Reasoning Models

Ryan Marten, Bespoke Labs | 19:59 | AI Engineer World's Fair 2025
Video: https://www.youtube.com/watch?v=liG97YXaTSA
Summary: https://aietalks.com/talks/openthoughts-data-recipes-for-reasoning-models.md

Why here: Lee curates billions of images with staged filters. Marten applies the same discipline to a smaller reasoning pipeline and tests each choice: question sources, repeated traces, teachers and answer filters. The next talk shows what happens when a good-looking synthetic recipe meets a full training run.

### 5. The Messy Reality of Scale: Synthetic Data and Pre-Training

Marah Abdin & Robert McHardy, poolside | 17:31 | AI Engineer World's Fair 2026
Video: https://www.youtube.com/watch?v=KhYifX22yhE
Summary: https://aietalks.com/talks/the-messy-reality-of-scale-synthetic-data-and-pre-training.md

Why here: Marten finds useful synthetic recipes with controlled experiments. This talk follows them into scale, where repetition, replica hashes, precision changes and silent corruption can turn a data result into a training failure. Crawford narrows the problem again by asking whether an agent task deserves to be trained on at all.

### 6. Task Fidelity Scaling Laws

Kobie Crawford, Snorkel | 20:40 | AI Engineer Europe 2026
Video: https://www.youtube.com/watch?v=YYH0DMQr30A
Summary: https://aietalks.com/talks/task-fidelity-scaling-laws.md

Why here: the previous talk shows that a recipe can fail in the machinery around it. Crawford shows that the task itself can be broken. An impossible specification or unreliable environment produces failure without a useful learning signal. Anand then applies the same suspicion to synthetic people.

### 7. Persona Engineering: A Field Guide to AI Synthetic Personas

Ishan Anand, InsightSciences.ai | 21:09 | AI Engineer World's Fair 2026
Video: https://www.youtube.com/watch?v=YnNF55QV0zs
Summary: https://aietalks.com/talks/persona-engineering-a-field-guide-to-ai-synthetic-personas.md

Why here: Crawford separates genuine difficulty from noise in a task definition. Anand does the same for synthetic personas. Models invent missing context, react to answer order and predict statements better than actions. Gupta and Rajpal follow with a production system that grounds those simulated users in tools and account state.

### 8. SimulationMaxxing: How We Ship Agents 20× Faster

Aman Gupta, Nubank & Shreya Rajpal, Snowglobe | 16:29 | AI Engineer World's Fair 2026
Video: https://www.youtube.com/watch?v=KMR_RBoCa4M
Summary: https://aietalks.com/talks/simulationmaxxing-how-we-ship-agents-20-faster.md

Why here: Anand says synthetic people are forecasts that need checking against reality. Gupta and Rajpal show that check in practice. Nubank compares simulated and real conversations before using the synthetic set to screen agent changes. Wang then asks how much of the surrounding workplace a simulation must contain.

### 9. The Data for Fully Autonomous Software Engineers and Companies

Joseph Wang, Emulated | 16:33 | AI Engineer World's Fair 2026
Video: https://www.youtube.com/watch?v=zkX03APVj0M
Summary: https://aietalks.com/talks/the-data-for-fully-autonomous-software-engineers-and-companies.md

Why here: Nubank simulates a customer, account state and mocked tools around an existing agent. Wang widens the environment to projects, incidents, deployments, distributed failures and real cloud resources. Zhao and colleagues close the pack by showing another source of difficult data: the product as people actually use it.

### 10. How We Built Zeta2: Training an Edit Prediction Model in Production

Ben Kunkle, Zed | 10:50 | AI Engineer Europe 2026
Video: https://www.youtube.com/watch?v=phchDt63qAA
Summary: https://aietalks.com/talks/how-we-built-zeta2-training-an-edit-prediction-model-in-production.md

Why last: Wang builds a richer imitation of production. Zhao and colleagues begin with opt-in production snapshots, repair teacher predictions, remove noisy moments and train a small model for the next keystroke. It is a fitting end because the dataset is no longer upstream of the product. Use generates evidence, and that evidence changes the next model.