Browser agents need better eyes
A stub for the argument behind my AI Engineer talk: many browser-agent failures are observation failures before they are reasoning failures.
Writing
Writing, notes, and arguments in progress.
This should eventually be the most durable part of the site. For now it is a small archive of notes from work that keeps tugging at me.
Correct the data before blaming the model
A short note from the WAT 2025 caption-correction work: sometimes the useful move is not a bigger model, but a cleaner signal.
Future prediction is not planning
A working note around robotics, world models, and the difference between predicting frames and choosing actions.