Measured, not modeled β a real Claude Code agent solving real GitHub issues, run with and without Lettuce.
13 open-source repositories the system had never been tuned against β axios, chalk, click, date-fns, express, fastapi, flask, hono, markupsafe, pydantic, requests, rich, and zod. Three real bug issues each β no feature discussions.
For every issue, the same Claude Code agent located the relevant code twice: once with only grep + file reads, once with Lettuce. Each arm ran 4 times; the median is reported. Tokens come from the agent's own usage meter.
Last run 5/25/2026.