WorkThe story, in brief
Anthropic co-founder maps out how recursive AI improvement could outpace the humans meant to supervise it
60% odds by end of 2028. Anthropic co-founder just mapped out how AI systems could outpace human supervision.
The KeyNews take
Why it matters
A senior AI lab leader is publicly staking a high-confidence forecast on recursive self-improvement timelines—signaling internal thinking at a top lab about AI capability acceleration and the governance challenges it creates.
The key facts
5 to knowJack Clark (Anthropic co-founder) published essay on recursive AI improvement
60% probability recursive AI training systems in place by end of 2028
Core thesis: building blocks for AI systems training successors largely complete
Governance risk: recursive improvement could outpace human supervision capacity
Published May 2026
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: Jack Clark argues in a long essay that the building blocks for AI systems training their own successors are largely in place. He puts the odds at 60 percent by the end of 2028.