Andrew Ho leaves OpenAI to build reinforcement-learning datasets for scientific AI
The GeneBench-Pro co-author is betting frontier labs will pay for verifiable biology and statistics tasks that web-scale training does not supply.
By Ryan Merket · Published
Primary source: Andrew Ho on X
Why it matters
Ho is turning the benchmark methodology he built inside OpenAI into a data-supply company. The bet places scarce, verifiable expert tasks at the center of the next phase of model training.
Andrew Ho (@andrewho03) left OpenAI on July 29th after eight months and said in an announcement on X that he is starting a company to produce reinforcement-learning datasets for AI labs, beginning with biology and statistical reasoning.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.