Report: OpenAI has automated much of experimental model training—AI writes GPU kernels, agents collaborate

Dr_Singularity · x · 2026-09-22

OpenAI has reportedly automated much of its experimental model training pipeline. Internal models can write and optimize GPU kernels, run optimization work for weeks from a single example, and collaborate with other AI agents without human involvement. Employees say this level of automation only became possible in the last few months—the training loop is starting to close.

Related event: Report: OpenAI Has Automated Most Experimental Model Training(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →