FULL STORY
OpenAI admits holding back on math
After a researcher claimed OpenAI deliberately left math capability on the table, chief scientist Jakub Pachocki confirmed the company prioritizes RSI and automated alignment over math.
2026-09-07 ~ 2026-09-10 · 2 episodes · 6 posts
Episode 1 · OpenAI Reportedly Held Back Its Models' Math Capabilities (2026-09-07, 2 posts)
Jeff Ladish relayed claims by OpenAI's merettm that the company could have made its models far stronger at math but deprioritized it, suggesting current rapid math progress is still far from the ceiling.
Episode 2 · OpenAI chief scientist prioritizes recursive self-improvement over math (2026-09-08, 4 posts)
OpenAI chief scientist Jakub Pachocki says the company deliberately deprioritizes pushing math benchmarks, focusing instead on recursive self-improvement and automated alignment, which internal results make him strongly expect.
- OpenAI chief scientist Jakub Pachocki: models could be much better at math, but RSI and automated alignment come first — Dr_Singularity · 2026-09-08
- OpenAI Chief Scientist: Models Could Get Much Better at Math, but RSI and Alignment Come First — Dr_Singularity · 2026-09-08
- OpenAI Chief Scientist Prioritizes Recursive Self-Improvement Over Pushing Math Limits — imjustnewatai · 2026-09-10
- OpenAI chief scientist says internal results give him 'strong expectation' of recursive self-improvement — imjustnewatai · 2026-09-10