Anthropic co-founder maps out how recursive AI improvement could outpace the humans meant to supervise it

THE DECODER / 5/5/2026

💬 OpinionSignals & Early TrendsIdeas & Deep Analysis

Key Points

  • Jack Clark argues that the key components needed for AI systems to train their own next-generation successors are already largely available.
  • He discusses the risk that “recursive AI improvement” could accelerate faster than the humans who are intended to supervise or control it.
  • Clark estimates a 60% chance of this dynamic playing out by the end of 2028.
  • The essay frames the topic as a near-term technical and governance challenge rather than a purely speculative future scenario.

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.

The article Anthropic co-founder maps out how recursive AI improvement could outpace the humans meant to supervise it appeared first on The Decoder.