Long-Horizon Agents Rely on System Architecture Over Base Models

A 149-page survey of over 900 papers reveals that long-horizon agent capabilities depend more on a complete system architecture than just stronger base models. It breaks down agent competence into reasoning, memory, and experience, emphasizing the role of the harness.

2026-07-25 ~ 2026-07-27 · 2 related posts

• A 149-page survey says long-horizon agents depend on harnesses, not just bigger models — 机器之心 · 2026-07-25
• Survey maps long-horizon agents into three layers, from in-context reasoning to lifelong experience — theomitsa · 2026-07-27