Blog

Notes from inside the loop

What we're learning about physical-AI data and infrastructure — published here, on LinkedIn, and wherever the conversation lives.

All posts

Latest writing

Slide through the posts — full articles open on their own page.

Why the bottleneck has moved from data to the loop behind it — provenance, quality gates, real-world evals, and failure attribution.

Why failure-and-recovery arcs are first-class training data, not noise to delete.

A show-floor read of the World Robot Conference — what separated the mature exhibits from the demos was rarely the robot.

What a scaling law fit on human video means for where embodied data value concentrates.

The real data bottlenecks crippling embodied AI — and how top teams build pipelines that beat them.

Follow along on LinkedIn — new essays land here and there.