“I now think it's more likely we'll see full RSI [recursive self-improvement] become possible in 2027, whereas before I thought it was more likely in 2028.”
Said Oct 12, 2026Deadline Dec 31, 2027Import AI 476: Creativity and RSI; worse-than-China open weight models; AI goes to space, October 12, 2026
Clark pulled his recursive-self-improvement timeline forward by a year after the TasteVal benchmark showed frontier models' experimental research taste doubling roughly every three months since December 2025, with the top model exceeding the human expert baseline. Deadline is end of 2027.