Agents 2.0：从浅层循环到深度智能体

For the past year, building an AI agent usually meant one thing: setting up a while loop, take a user prompt, send it to an LLM, parse a tool call, execute the tool, send the result back, and repeat. This is what we call a Shallow Agent or Agent 1.0.

在过去一年里，构建 AI agent 通常意味着一件事：设置一个 while 循环，接收用户提示，将其发送给 LLM，解析工具调用，执行工具，将结果返回，然后重复。这就是我们所说的 Shallow Agent 或 Agent 1.0。

This architecture is fantastically simple for transactional tasks like "What's the weather in Tokyo and what should I wear?", but when asked to perform a task that requires 50 steps over three days, and they invariably get distracted, lose context, enter infinite loops, or hallucinates because the task requires too many steps for a single context window.

这种架构对于“东京天气如何，我该穿什么？”这类事务性任务来说极其简单，但当被要求执行需要 50 个步骤、持续三天的任务时，它们往往会分心、丢失上下文、陷入无限循环或产生幻觉，因为任务步骤过多，单个上下文窗口无法承载。

We are seeing an architectural shift towards Deep Agents or Agents 2.0. These systems do not just react in a loop. They combine agentic patterns to plan, manage a persistent memory/state, and delegate work to specialized sub-agents to solve multi-step, complex problems.

我们正在见证一种架构上的转变，朝着 Deep Agents 或 Agents 2.0 迈进。这些系统不再只是在一个循环中被动响应，而是结合 agentic patterns 进行规划，管理 persistent memory/state，并将工作委派给专门的 sub-agents，以解决多步骤、复杂的问题。

overview

Agents 1.0: The Limits of the "Shallow" Loop

Agents 1.0：“浅”循环的局限

To understand where we are going, we must understand where we are. Most agents today are "shallow". This means rely entirely on the LLM's context window (conversation history) as their state.

要理解我们将走向何方，必须先明白我们身处何处。如今大多数 agent 都是“浅层”的，这意味着它们完全依赖 LLM 的上下文窗口（对话历史）作为其状态。

User Prompt: "Find the price of Apple stock and tell me if it's a good buy."
User Prompt: “查找 Apple 股票的价格，并告诉我是否值得买入。”
LLM Reason: "I need to use a search tool."
LLM 推理： “我需要使用搜索工具。”
Tool Call: search("AAPL stock price")
工具调用： search("AAPL stock price")
Observation: The tool returns data.
观察： 工具返回数据。
LLM Answer: Generates a response based on the observation or calls another tool.
LLM Answer: 根据观察结果生成回答，或调用另一个工具。
Repeat: Loop until done.
Repe...