tszzl: models aren't autonomously producing research ideas — RSI is not here yet

i_dg23 · x · 2026-09-13

Responding to discussion about AI-driven research progress, tszzl argues that recursive self-improvement (RSI) is simply not here: models are straightforwardly not autonomously producing research ideas, and can only marginally help run experiments — which is not the same thing. He acknowledges private frontier models may be far better but doubts it.

Related event: Devs Debate Whether Recursive Self-Improvement Is Real or Hype(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →