Models Train Themselves Now. The Leap Is Small. Recursive Improvement Is Coming, Obviously.

Leading labs suggest AI models are approaching recursive self-improvement, or RSI, where systems autonomously learn to become more efficient and capable. Prior efforts, including those by OpenAI co-founder Andrej Karpathy, yielded only minor improvements. Researchers Aguirre and Thickstun note that significant qualitative breakthroughs now seem close.

The concept is recursive self-improvement, a feedback loop where a system enhances its own capabilities without human intervention. The historical pattern is that these loops produce incremental gains, not exponential explosions. The mental model to internalize is the difference between linear refinement and a phase transition. We appear to be approaching the latter.

OpenAI co-founder Andrej Karpathy has experimented with AI models training new AI systems. Researchers Thickstun and Aguirre indicate that leading labs are now close to achieving qualitative breakthroughs in self-improvement.