A simple SWE style browser agent framework that achieves SOTA results on long horizon web tasks — We

追加: 2026-06-01

[00:00:00]Microsoft just beat the browser agent state of the art by 15 points.

[00:00:05]It's Web Rider, a coding agent that writes Python instead of clicking pixels. 86.7% on online Mintue web. 60.1% on the long Horizon ODC's benchmark. The core agent loop is a single 450 line file. No graph engine, no plugin layer. Works with Claude code, Codex, and Hermes as a skill.

[00:00:27]Description has the repo link.