OpenEvolve
Run 'AI discovers algorithms' yourself
What it is
An open evolutionary coding agent in the lineage of DeepMind's AlphaEvolve: it uses LLMs as mutation operators inside a quality-diversity loop with island-based parallel populations, evolving programs against automated evaluators. It has reproduced state-of-the-art results and discovered real GPU-kernel speedups.
Why it's interesting
The most accessible way to run algorithm-discovery experiments yourself — a genuine research frontier packaged so that someone with an API key or a local model can hunt for novel optimizations over a weekend.
Use cases
- Automated optimization of hot code paths and GPU kernels
- Heuristic and algorithm discovery for well-specified problems
- Research into LLM-guided evolutionary search
Who it's for
Researchers, performance engineers, adventurous tinkerers
Setup
Moderate. Python 3.10+, an OpenAI-compatible LLM endpoint, and a scoring function you write for your problem
Limitations & cautions
Only works on problems with an automatable fitness metric, can burn large token budgets, and evolved code needs human review before production.
Editorial takeaway
AlphaEvolve made headlines and stayed closed. This is the version you can point at your own bottleneck tonight.