Full recursive self-improvement (RSI) is now more likely to become possible in 2027 than in 2028
“I now think it’s more likely we’ll see full RSI become possible in 2027, whereas before I thought it was more likely in 2028.”
Context: Jack Clark, Anthropic co-founder, in "Import AI 476: Creativity and RSI; worse-than-China open weight models; AI goes to space" (personal newsletter, 12 Oct 2026), after discussing P-Zero Research's TasteVal benchmark ("experimental research taste of frontier models has doubled every 3.0 months since December 2025") and recent creative-seeming AI math results. RSI = recursive self-improvement; in Import AI 455 (4 May 2026) he described it as "no-human-involved AI R&D – an AI system powerful enough that it could plausibly autonomously build its own successor" and gave about 60% by end-2028 and 30% for 2027 (clark-automated-ai-rnd-60pct-2028). He gave no new probability. Sources: (1) Import AI 476 on jack-clark.net (primary); (2) NDTV, "AI Could Start Improving Itself By 2027: Anthropic Co-Founder" (same words in quotes); (3) Techmeme headline, 13 Oct 2026 (same words); (4) The Decoder, "Anthropic co-founder Jack Clark moves his prediction for self-improving AI forward to 2027" (25 Oct 2026).
How we judge it
Our interpretation: correct if by 31 Dec 2027 a frontier AI system is broadly recognized by independent experts as capable of full recursive self-improvement, meaning it can autonomously do the AI research and training that produce a more capable successor without humans doing that work (his own Import AI 455 definition of no-human-involved AI R&D); a developer's own claim does not count on its own. Wrong if this has not happened by 31 Dec 2027. He named 2027 as the more likely year without a probability, so it is graded as a dated prediction, not on calibration.
Resolution criteria are our interpretation of the claim, written before grading. Methodology
- Said
- Resolve by
- Due in 420 days
- Verdict
- Open
- Stated probability
- None stated
- Specificity
- 2 of 3 (dated, some interpretation)
Share this scorecard
Post on X LinkedIn Share image (PNG)
Embed / Cite
The embedded card is live: it shows the current verdict, with the date the data was read. If a verdict is corrected later, the card changes too, so cite the page and its date for a fixed record.