# // From Language Models to World-Acting Systems // A critical review of agentic AI， and a framework…

- 来源：DAIR.AI (@dair_ai)
- 发布时间：2026-09-08 02:00
- AIHOT 分数：45
- AIHOT 链接：https://aihot.virxact.com/items/cmtrkrflj07marotnz7k9w0gy
- 原文链接：https://x.com/dair_ai/status/2097022152088445034

## 正文

// From Language Models to World-Acting Systems //

A critical review of agentic AI, and a framework that is genuinely useful for deciding how much authority to hand an agent.

Here is how it works.

The review separates three things the field routinely treats as one. Model competence, harness integration, and the authority a deployment actually grants are pulled apart and assessed separately.

Evidence gets organized along delegated authority, temporal persistence and environmental coupling, and the model, the harness and the environment stay distinct when a result is attributed.

The finding across the papers examined is that expansion of action interfaces is documented far more convincingly than robust completion, recovery, authorization or independent verification. MCP and Agent2Agent improve interoperability without establishing that delegation is trustworthy. Multi-agent organization buys specialization along with cost and correlated failure.

Paper: https://academy.dair.ai/papers/from-language-models-to-world-acting-systems-progress-and-limits-of-agentic-ai-a-2609.04894