Ex-Apple engineer open-sources Harness Tuner: Claude Code auto-tunes agents, 2.5x fewer failures
davernow · reddit · 2026-11-14
An ex-Apple engineer released Harness Tuner (MIT, open source), which lets Claude Code automatically tune AI agents—rewriting tools, tightening prompts/skills, and trying other models—while measuring every change against your full eval suite. On a real production agent it achieved 2.5x fewer failures, 6x lower cost, and 2x speedup, validated on held-out conversations enforced in code. A loop of small Claude Code agents plans experiments, builds them, and decides what to keep; Codex and OpenCode also work. Install with uv tool install harness-tuner and try harness-tuner --demo.
More from coding & agent
- Intern shipped a multiplayer app built with Claude Code, got hired, then presented it on stage — Alec_Coughlin · 2026-11-14
- Dev: Working at a High Level with AI Is Great but Exhausting — jonathan_wilke · 2026-11-14
- Open-source bastyn statically scans MCP and coding-agent configs for risks — claudio-rootedlogic · 2026-11-14
- 'We'll forget how to use apps': devs say agents make hand-coding feel unbearable — robleclerc · 2026-11-14
- OpenComputer launches as open-source, self-hostable Devin alternative claiming 3.3x faster coding agents — zeeg · 2026-11-14
- 15 building blocks of AI agent memory, distilled from building 400+ agents — MaryamMiradi · 2026-11-14