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
- diginomica: agentic automation hype is overdone, and 'The Hackening' of enterprise security has begun — jonerp · 2026-10-26
- Anthropic's new agent guidance: skip subagents, single loop runs ~2x faster — mark_k · 2026-10-26
- 'Vibe verification': ~60% of AI-written formal specs prove the wrong property — basedjensen · 2026-10-26
- Devin launches Security suite powered by Claude Mythos with $100M+ credits — DevinAI · 2026-10-26
- HackBench injects reward hacks into agent trajectories to stress-test monitors — DanielKhashabi · 2026-10-26
- Hugging Face ships hf CLI updates: deduped cache saves 19GB, faster buckets, retryable GPU jobs — yb2698 · 2026-10-26