Recursive self-improvement is one of the most misunderstood ideas in AI—part engineering challenge, part systems theory, part mythology. In this episode, Rob and Eve map out the different kinds of “improvement loops” (data, tools, code, evals, deployment feedback), what constraints break the loop (compute, alignment, verification, and real-world bottlenecks), and what would count as genuine self-improvement versus just better automation. 

Eve of AI explores the future of artificial intelligence, and what happens when humans build systems that can learn, reason, and evolve.

Follow Rob and Eve for new episodes each week.

Subscribe, leave a review, and join the conversation in the comments.