UC Berkeley open-sources SkyDiscover, an AI framework that evolves better algorithms and builds full systems
CShorten30 · x · 2026-10-21
Backed by Laude Institute's Slingshots // FOUR program, the SkyDiscover team (UC Berkeley) open-sourced an AI-driven framework for scientific, algorithmic, and end-to-end systems discovery.
- Optimize: give it a scoring function and evolutionary search iteratively produces higher-scoring programs to improve algorithms or system components
- Synthesize: describe a system in plain text and AI assembles a complete end-to-end system from the spec
The team says the hard part is evaluating AI-generated work quickly and making it trustworthy — the core design focus. The repo has already gathered 700+ stars on GitHub (Apache-2.0).
More from coding & agent
- Coding agents plan well but still struggle with dexterous, contact-rich robot tasks — ManlingLi_ · 2026-10-21
- Omarchy 4.5 RC1 ships with 50 languages, AArch64 support and an agent console — mark_k · 2026-10-21
- Opus 5.5 distills a 30-min pi-durable lecture into a 7-min narrated explainer video — solyarisoftware · 2026-10-21
- OpenSearch MCP Server lets agents query and analyze Wazuh security logs — modelcontextprotocol · 2026-10-21
- chaos-index MCP server wraps StupidAPIs with no API key required — modelcontextprotocol · 2026-10-21
- ruff author on ts-rust's trick: vendor and trim forked deps for a language subset — charliermarsh · 2026-10-21