Reproduce “A mathematical theory of evolution for self-designing AIs” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Kenneth D Harris

The authors model recursive AI self-design as a directed tree of successor programs, while humans allocate resources through a fitness function. They show that under η-locking and bounded reachable fitness, mean fitness converges to the reachable supremum, and population share concentrates near it. The supremum need not be attained; this conditional mathematical result concerns the model, not real AI populations.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.