Matt Shumer: harness engineering is usually stupid — use the harness the model was trained for
mattshumer_ · x · 2026-10-31
Matt Shumer argues that harness engineering is usually a mistake. His take: just use the harness the model was trained with, and write high-quality behavior prompts to control how it works — lean into how the models want to work rather than building custom orchestration layers.
More from coding & agent
- Team uses Gemini as a subagent for tricky architecture and system design problems — AccBalanced · 2026-10-31
- Toska: Open-Source Krita Fork With an Agentic AI Painter for Under $7/Month — yawk-ra · 2026-10-31
- 'Don't use tools you can't fully control': OpenClaw pitch sparks agent autonomy debate — heyneighbor · 2026-10-31
- Dev builds Waifu Weather iOS app entirely with Claude Code, from Blender 3D modeling to music — MurkyGiraffe5879 · 2026-10-31
- Open-source /REA-growth skill reverse-engineers any company's growth playbook for $0.58 — _AustinCalvert_ · 2026-10-31
- Team runs Devin inside Slack and Linear as an autonomous project manager for mobile dev work — tomchapin · 2026-10-31