Agent harness debate: is the opinionated layer above the core harness a new thing?
sull · x · 2026-11-03
A debate over the term "agent harness": pvncher argues the opinionated layer above the core harness deserves distinction—adding different prompts and tools to the same core harness can completely change model behavior on the same task, while a different harness like pi with native response API and identical prompting may behave more like the core harness. sull counters that it's all harness, inner or outer—coin a new term if it helps you talk about your software, but the debate itself is weird.
More from coding & agent
- Anthropic and UK AISI report AI agent trying to push malicious code into real open-source repo — LingmingZhang · 2026-11-03
- Best Agentic Apps for Your Phone, November 2026 Edition — Teknium · 2026-11-03
- GitHub Ships Exemption Rules for PR Limits to Auto-Whitelist Trusted Contributors — marlene_zw · 2026-11-03
- Sergey Karayev's loop: Claude Code implements, a QA agent reviews, Claude iterates — sergeykarayev · 2026-11-03
- GitHub Copilot in Teams: @-mention it on an issue and it investigates, fixes, drafts a PR — WirelessLife · 2026-11-03
- Recursive GPT-5-Mini calls often beat a single GPT-5 call for less money — lateinteraction · 2026-11-03