Agent harnesses should be retested every time the model changes
hugobowne · x · 2026-07-23
Hugo Bowne-Anderson argues that much of an agent harness is really a guess about what the model still cannot do.
His advice is to re-test compaction, retries, routing, and planning scaffolds every time the model changes, and remove anything that no longer improves the system.
The linked slide makes the point bluntly: yesterday’s necessary workaround can become tomorrow’s dead weight.
More from coding & agent
- HyperFrame shows how it rebuilt the demo and says Claude Code plus Fable can train HQ agents — op7418 · 2026-07-23
- Training HQ Agents Using Claude Code and Fable via CLI — jacob_posel · 2026-07-23
- AI Engineer talk launches ActiveGraph, a durable runtime for long-running agents — altryne · 2026-07-23
- Claude Code repo packages an Indian equities trading bot with Zerodha integration — tom_doerr · 2026-07-23
- OpenAI Codex community hackathon draws a full room of builders and judges — paw_lean · 2026-07-23
- Cohere Labs posts a hiring message built around collaborative science culture — Cohere_Labs · 2026-07-23