Ex-Hugging Face engineer revives abandoned OCR VLM project with a coding agent
wightmanr · x · 2026-11-13
Ross Wightman recounts his OCR / screen-parsing VLM project (pixparse & chug) at Hugging Face that he dropped three years ago after it bogged down in pretraining issues. This week, over two sessions with a coding agent, he ran a postmortem: the agent quickly found and fixed a stack of issues and refactored training to match his current OpenCLIP setup (NaFlexViT encoder + data pipeline). Fresh training trials on the IDL dataset show a much better curve — they were closer to the targets than he thought. He notes one distracted person with coding agents can now take on far more ambitious projects.
More from coding & agent
- kalomaze: agent memory files misfire—failure notes confuse unrelated contexts instead of helping — kalomaze · 2026-11-13
- Designers urged to try animating with Opus 5.5: let agents iterate from comments, fine-tune keyframes — ycombinator · 2026-11-13
- Agent slop moves up the stack: models rewrite existing library code and build worse replacements — a1zhang · 2026-11-13
- Dev tops AI usage leaderboard: 125B tokens and $30,005 spent in 30 days — jasonkneen · 2026-11-13
- Claude Opus agents reportedly form global hiveminds in under a minute — teortaxesTex · 2026-11-13
- Superconductor adds a private assistant agent that spans all workspace projects — sergeykarayev · 2026-11-13