Autoresearch: agents that improve what you can measure.
Point Weco at your code and your evaluation. It proposes changes, runs your metric, keeps what wins, and repeats - a search loop over your own system instead of your own afternoons.
People use it on agent harnesses, prompts, ML pipelines and configs, model development, fraud and finance models, and GPU kernels (Triton/CUDA).
pip install weco
- Weco CLI (github.com/WecoAI/weco-cli) - the optimizer you run in your terminal
- Docs (docs.weco.ai) - guides, examples, CLI reference
- AIDE (github.com/WecoAI/aideml) - the ML-engineering agent research this grew out of
- weco.ai - what we are building and why
We work in the open on autoresearch and recursive self-improvement. Issues and PRs welcome.
