Run an auto-research loop with Codex to speed up a GPU kernel
A developer ran Codex in an auto-research loop for a GPU Mode contest and reached a 232x speedup over baseline on QR decomposition.
Real things people did with AI work, sourced and with recipes.
A developer ran Codex in an auto-research loop for a GPU Mode contest and reached a 232x speedup over baseline on QR decomposition.
A developer trained an agent whose job is to write reinforcement-learning jobs for small models, for about $1.3k, and open-sourced it.
Five real AI agent use cases each morning, each with a recipe and a source. Plus a Sunday roundup. Free.