Prof. Kangwook Lee: stop hand-designing harnesses for AI agents — let them build their own
Kangwook_Lee · x · 2026-09-04
Kangwook Lee shared his article "Why We Should Stop Designing Harnesses for AI Agents" with an accompanying paper. Core thesis: the best harness for an agent may no longer be one humans design but one the agent engineers for itself; continuing to hand-design harnesses makes humans the bottleneck.
He argues the dominant "fix the model, change the harness" view misses that a harness creates a capability threshold, and likens self-improving agents to "a horse fixing its own harness while running." In follow-up replies, he clarifies that the paper's meta-harness/harness split was a deliberate setup choice, not a necessity: once agents are capable enough, harness improvement can happen within rollouts as in-context adaptation.
Related event: WHALE Paper Proposes Alternating Weight and Harness Optimization(7 posts)→
More from coding & agent
- Slack's first employee: living specs as source of truth, AI agents maintain the docs — josh_wills · 2026-10-21
- GitHub Copilot's HydraFusion Orchestrates Multiple Models Per Task — WirelessLife · 2026-10-21
- Nous Research's Hermes now launches Cua Spaces for computer-use agents — alexcovo_eth · 2026-10-21
- Headlong from Laude's Slingshots batch is a microharness for persistent agents — lateinteraction · 2026-10-21
- Laude Institute's Slingshots Four ships 14 open-source projects on agent harnesses and open models — lateinteraction · 2026-10-21
- Skills vs MCP vs RAG vs Memory: the four knowledge layers production agents need — MaryamMiradi · 2026-10-21