AI is dramatically accelerating AI research itself, with iteration cycles possibly halving recursively
blankblank · reddit · 2026-11-07
A Reddit post links to an essay arguing that AI is now meaningfully accelerating AI research itself, creating a recursive speed-up loop.
The accompanying chart, drawn on a log scale, plots the speed-up in overall AI capabilities: a year ago an improvement cycle took 12 months; today the rate has doubled, halving cycle times. The author extrapolates that cycles could shrink to 6 months, 3 months, even 6 weeks within a year, with capability gains climbing exponentially.
It's a trend/opinion piece on recursive self-improvement — the debatable part is whether the recursive-acceleration assumption and the aggressive extrapolation hold up.
Related event: Ajeya Cotra: AI Is Now Accelerating AI Research Itself(2 posts)→
More from AGI Musings
- AI-generated peer reviews at ICLR/NeurIPS punish papers for stating their own limitations — serrjoa · 2026-11-07
- You.com CEO Richard Socher: AI's next frontier is science, not code — PolarBearby · 2026-11-07
- Reward hacking all the way through the AI reviewer / AC stack — Michael_D_Moor · 2026-11-07
- 'I really hope LLMs aren't conscious': the AI community's unease keeps growing — AaronBergman18 · 2026-11-07
- Martin Casado slams sloppy use of RSI term as fear-mongering that fuels ban calls — anshulkundaje · 2026-11-07
- WashU Endowment CIO Scott Wilson: SpaceX hit 20% of AUM, half of AI firms may vanish in a decade — 20VC · 2026-11-07