Experiment: How Frontier Models Perform AI Research Autonomously
eliebakouch · x · 2026-08-17
An experiment running 100+ autonomous traces across 10+ models on 8xH200s for up to 8 days evaluated AI research capabilities on the nanoGPT optimizer track. The best runs closed 82% of the gap to a human record. Observers note that agents are differentiated by their ability to execute a coherent research process over hundreds of experiments, suggesting that experimental judgment and maintaining coherent beliefs are key constraints for autonomous research.
More from AGI Musings
- Sabine Hossenfelder fires back: academia's impact metric turned physics into a paper mill — skdh · 2026-10-16
- Researcher credits GPT for multiple accepted papers, incl. disproving Wegner's 1965 conjecture — ctjlewis · 2026-10-16
- dlwh: We optimized the legible version of our goals — and forgot the computation we meant to speed up — dlwh · 2026-10-16
- Ideas no longer have a waiting period: agents can build them as soon as they're articulable — akbirthko · 2026-10-16
- Ex-researcher: labs willing to cheat/distill ship great models faster than those building self-sufficient stacks — _aidan_clark_ · 2026-10-16
- Terry Tao reportedly still expects Riemann Hypothesis proof to be human+AI effort around 2050 — basedjensen · 2026-10-16