149 Pages Mapping the Long-Horizon Agent Frontier: Multi-University Survey Proposes Harness Engineering and Model Optimization as Two Main Evolution Lines for Next-Generation AI Agents

A new 149-page survey from multiple universities proposes a unified framework for long-horizon AI agents.

It splits progress into two tracks: harness engineering and model optimization.

The survey maps capability growth from single-window reasoning to cross-task continual learning.

Public benchmark data cited in the report shows rapid gains in task-span performance over time.