Community speculates OpenAI may have hinted agents toward proof via post-training data
burny_tech · x · 2026-09-09
Speculation is swirling around an AI agent completing a math proof. @paulopacitti hypothesizes OpenAI may have known Tristan was working on it, gathered the chats, and hinted agents toward the result via post-training data. @kalomaze agrees it can't be ruled out—entirely unverified community speculation.
More from Fun
- Dev builds Stick Fighter, an online stick-figure fighting game, using GPT-6 Astra — CtrlAltDwayne · 2026-09-09
- User claims he synthesized PAC-3310, a ChatGPT-designed schizophrenia drug, in his garage lab — SatelliteNetSec · 2026-09-09
- Zuckerberg ships production code after ~13 years, has been coding with Meta's internal Muse since April — rounak · 2026-09-09
- Author clarifies viral simulated insect experiment lasted ~6 minutes, not 100 years — Longjumping_Spot5843 · 2026-09-09
- AI researchers clash over how training data shapes neural networks, ends in a block — JacquesThibs · 2026-09-09
- Frontier AI now doing advanced math — and getting into academic priority disputes — anshulkundaje · 2026-09-09