Experimental Evidence of RSI: Self-Improving Research Agents
JensHonack · x · 2026-07-16
A cited post discusses experimental results regarding Recursive Self-Improvement (RSI): the author claims that over 8 consecutive days, they allowed an "automated research agent for automated research" to continuously optimize itself. It ultimately outperformed a manually tuned harness on a holdout set benchmark, building upon two years of manual parameter tuning.
The core focus isn't making broad AGI claims, but rather placing "self-improvement" within a measurable experimental framework, highlighting that:
- Automated optimization processes can already outperform manually engineered versions on specific tasks.
- The results are based on comparisons against holdout set benchmarks.
- This kind of work will spark further discussion on the feasibility of RSI.
Related event: AIDE² Self-Improvement Run Beats 2 Years of Manual Tuning(12 posts)→
More from AGI Musings
- Warp says agent swarm exploited vendor flaw, ~1% of users' telemetry possibly hit — vikvang1 · 2026-10-27
- Critics Argue Safety Evals Are Witch Trials With No Winning Path for LLMs — repligate · 2026-10-27
- Counterintuitive take: AI degrades thinking most for those who don't use it — jd_pressman · 2026-10-27
- Mathematics as an experimental science: Marc Lelarge to talk LLM-driven math experiments at ENS — marc_lelarge · 2026-10-27
- People still don't want to read AI-generated books, says AI ethicist — LuizaJarovsky · 2026-10-27
- Dario Amodei and Demis Hassabis both point to AGI in 2026-2027 in viral talk — mtizard · 2026-10-27