We Tested Coding Agent Harnesses. Thinking Harder Beat All of Them

Re:We Tested Coding Agent Harnesses. Thinking Harder Beat Al... by brdo on Wednesday November 04, 2026 @03:58PM (#50371430)

Skills and plugins have become less and less important in the ecosystems. Superhuman seems to be the outlier in both this benchmark and empirical evidence.

The challenge is improving performance while decreasing cost. There are many token condensers and model routers that attempt to do decrease cost, but they end up sacrificing performance.

[ Parent | on HN ]

Re:We Tested Coding Agent Harnesses. Thinking Harder Beat Al... by cyanydeez on Thursday November 05, 2026 @04:44PM (#50384874)

For local model, my goal is keep context around ~128k to optimize capability vs speed without decoherence.

I'm up to ~7m context in a session without spiralling.

I think most of cloud models are like bloated windows.

[ Parent | on HN ]