MCP

Model Context Protocol guides, updates, and resources.

9 Benchmarks: What Happens When a Coding Agent Can Search Research Papers

We ran 9 tasks - writing tests, generating SQL, reviewing PRs, classifying text, extracting data from documents. Same coding agent, same data, same evaluation. The only variable: whether the agent could search research papers before choosing its approach.

April 15, 2026

Paper Lantern improves Autoresearch

We connected Paper Lantern to an autoresearch agent and ran it head-to-head against the same agent without it. The Paper Lantern configuration reached 3.2% lower validation loss – and the gap was still widening.

March 26, 2026