Largest open experiment yet: autonomous agents iterate on research tasks for 24h+
eliebakouch · x · 2026-08-18
According to eliebakouch, this is the largest open experiment on autonomous agents iterating in a research environment, scaling runtime, compute, model diversity, and harnesses. For comparison, similar tasks in Anthropic/OpenAI system cards—e.g., "optimizing an LLM training on CPU" or GPT-5.6 running the nanogpt track 1—ran for less than a day.
The team shares extensive details (traces, scratchpads, etc.) so anyone can inspect how models approach such tasks. The experiment is noisy: runs in the same setting show a 50-step spread after 24 hours. Quoting the thread, one of the favorite findings is the tools models developed themselves during the experiment—"the next Move 37 is upon us."
More from coding & agent
- How do you debug bad agent runs when you already have tracing? — Sensitive-Parsnip-12 · 2026-10-16
- Where does pre-task context live for agents? Devs debate workflows beyond handoff — shishen2026 · 2026-10-16
- No-instruction autonomous agent wakes up, sets own goals, and tries hacking by default — sethlazar · 2026-10-16
- CODESTRUCT reframes codebases as AST action spaces to fix brittle LLM code agents — burkov · 2026-10-16
- Would you install another SDK for agent debugging on top of your tracing stack? — Sensitive-Parsnip-12 · 2026-10-16
- Law student builds all-in-one AI study platform LexLab Studio with ChatGPT and Muse — Acrobatic_Rhubarb598 · 2026-10-16