Field Notes

Engineering the data supply chain for AI.

Working notes on AI training data — engineering structured, rights-cleared training data for the model gaps it's built to close.

Humanoid robot working at an industrial bench with subtle data overlays in a magenta and blue robotics lab.

The Great Robot Data Glut: Millions of Hours of Egocentric Video, and Most of it is Worthless

A bipedal robot stands in a well-lit industrial warehouse with translucent bounding box annotation overlays visible around it, lit in magenta and electric blue.

The Data Bottleneck in Physical AI: What We Learned at CVPR 2026

Cinematic data-center corridor with fragmented, magenta-lit data on the left resolving into clean parallel electric-blue data lanes feeding a GPU rack on the right — raw training data refined into training-ready infrastructure.

The $100B Data Tax: Where AI Training Budgets Actually Go

AI training data for generative video models

DataPacks: Training Data Organized Around the Gap You’re Trying to Close

Humanoid robot standing in a research lab, lit by magenta and bright blue stage lighting, with translucent data streams swirling around its torso.

GPU-Ready: Data Engineered for Spatial Intelligence

Related Tags

For AI Developers

License High-Quality Data for Training your LLM