Home
Technical essays
Density, capture, and trainable signal.
Notes from World Archive DataOps on what actually moves IL and VLA fine-tuning — written for robotics and frontier lab teams evaluating egocentric corpora. Essays publish on Substack first; this site embeds the live posts.
June 2026
The density advantage: labels per minute that actually train policies
Why annotation density beats raw video hours for IL and VLA fine-tuning.
June 2026
Beyond the Monocular Plateau: How DataOps Wins the Next Year
Scaling laws are real — but monocular ego alone is hitting diminishing returns. The next winners deliver dense, multimodal, rights-clear signal at scale.