Research Uses LLM Code Gen to Drive Evolutionary Algorithms

kenneth0stanley · x · 2026-08-22

New research demonstrates using LLMs as mutation and crossover operators to drive evolutionary algorithms that generate entire programs. Instead of just evolving individual game levels, the system dynamically creates a pool of reusable expert API primitives. This approach combines LLM code generation with evolutionary search to build more complex and modular generators.

Related event: LLM-Powered Evolutionary Algorithm Generates Programs and Expert APIs(2 posts)→

Original post →

More from coding & agent

coding & agent channel →