Harness Co-Training: Models shaped inside agent loops become new industry norm
DynamicWebPaige · x · 2026-08-18
The article analyzes the trend of "Harness Co-Training," where models are trained within agent scaffolds before release. It contrasts strategies from poolside, Cursor, Z.ai, and Anthropic: poolside trains across multiple harnesses on purpose, Cursor injects corrections at tool failure points, Z.ai reports best-harness scores, while Anthropic argues for leaving the model untouched.
Commercially, model capabilities become tied to specific scaffolds, meaning leaderboard scores reflect a "model-plus-harness" pair. Switching environments may degrade performance.
More from coding & agent
- Bionic adds support for agent skills, reusing Codex and Claude shortcuts — mattturck · 2026-08-18
- YC Startup Codag Launches to Compress Agent Context 3x and Monitor Tool Usage — ycombinator · 2026-08-18
- Master AI Evaluations: Build Workflows from Traces to Automation — realmadhuguru · 2026-08-18
- ComfyUI-MiniMax-H3-LongMedia: Continuity for long-form generation — Independent-Ear-3035 · 2026-08-18
- 28.65M Secrets Leaked on GitHub in 2025; AI Coding Tools Leak More Than Humans — Thionne_WTZ · 2026-08-18
- Mailflare: Self-Hosted AI Email Inbox on Cloudflare with Custom Domains — tom_doerr · 2026-08-18