Jan Leike wrote that Claude was producing almost all of his team's research code, mostly autonomously running standard sampling, evaluation, and fine-tuning workflows, auditing models, and reading transcripts. He called this the start of recursive self-improvement, while arguing alignment looks increasingly solvable and warning that faster automated research may leave little time to align superintelligence.
Sign in to join the discussion →
No comments yet. Start the discussion.
A frontier-lab alignment lead's operational account indicates that AI is already automating important parts of AI research, which can accelerate capability feedback loops; confidence is capped because the internal productivity claims are self-reported and the recursive effect is inferred rather than directly measured.
Reviewed videos from Artificial Intelligence Videos. These links add context and never change a DoomBench score.
Historical backfill of a distinct first-person RSI and automated-research assessment.
17 Aug 2026DoomBench assesses “Jan Leike says recursive AI self-improvement has begun in alignment research” as evidence moving toward doom, with magnitude 46 and confidence 58 out of 100 in the capability gains category.
The DoomBench assessment of “Jan Leike says recursive AI self-improvement has begun in alignment research” is based on reporting from Musings on the Alignment Problem and records the editorial rationale, source quality, attribution, and...
DoomBench summarizes “Jan Leike says recursive AI self-improvement has begun in alignment research” as follows: Jan Leike wrote that Claude was producing almost all of his team's research code, mostly autonomously running standard...
