16 October 2026
Anthropic's Jack Clark Forecasts Recursive AI Self-Improvement by 2027
First reported
Deep Learning Weekly and Transformer ran this on , all on the same day.
- Jack Clark, co-founder of Anthropic (the company behind the Claude chatbot), now thinks AI systems may fully improve themselves by 2027.
- Patel urges public release of the evidence behind that prediction, so outsiders can judge it.
Our read
Only 2 of 30 newsletters covered this story, Transformer and Deep Learning Weekly, with Deep Learning Weekly first reporting it on 2026-10-16. The two outlets framed it differently, with Transformer treating it as a forecast to be checked and Deep Learning Weekly treating it as support from a research paper.
Written by Newskeryx from which newsletters covered this story and from our own archive.
Where they differ
Transformertreats the story as a forecast from inside Anthropic that it calls for public evidence.
Deep Learning Weeklytreats it as a research result, the MASS paper, that it presents as supporting evidence for self-improvement.
What each one covered
Predictions on recursive self-improvement timelines and calls for public evidence
The MASS alternating optimization method, Qwen3.6-27B results, and training efficiency of multi-agent traces