algorithmicsuperintelligence/openevolve is a Python tool that implements DeepMind's AlphaEvolve as open source. The latest recorded release is v0.4.0. OpenEvolve takes an initial program and an evaluator as input and iteratively evolves the code. The evaluator is the criterion used to judge the program's results. However, the provided information does not state what changed in v0.4.0, so the changes in this version cannot be pinned down.
The purpose of this project is to solve optimization problems through code evolution. The README explains how to call it as a library and shows examples of running it from the command line. It also introduces extension points for splitting the search into islands and assigning work, and for choosing the policy that manages the population. A population is the set of program candidates managed during the search. The examples include function optimization, circle packing, sorting, signal processing, and metal kernel optimization. The language is Python and the license is Apache-2.0. The recorded star count is 7652. This setup offers a concrete entry point to consult when implementing a code-evolution search.
The research record's momentum score is 4.8 out of 10, and no novelty, technology, or quality scores were given. In particular, the change log for the latest release was not provided either, so there is no basis for describing the meaning of v0.4.0 as a feature change. This repository can therefore be examined within the scope of the AlphaEvolve-inspired implementation and its examples, but the improvements in this release are difficult to judge from the current information alone.