📄 中文摘要
2026年3月，人工智能代理的技术已取得显著进展。过去，许多所谓的“代理”只是在表面上使用提示的工具，缺乏实际的自主性和可靠性。然而，随着基础设施的成熟，代理现在能够在特定的、经过良好仪器化的工作领域中发挥作用。代理不仅仅是一个有任务的语言模型，而是一个能够决定何时使用工具、在循环中操作并检索上下文信息的系统。这一变化标志着人工智能技术的实用性和可靠性得到了提升。
2026年3月人工智能代理的现状及如何构建特定主题的代理
📄 English Summary
The state of AI agents in March 2026, and how to build a topic-specific one
As of March 2026, significant advancements have been made in AI agents. A year prior, many so-called 'agents' were merely prompt-driven tools lacking true autonomy and reliability. However, with the maturation of the surrounding infrastructure, agents are now capable of functioning effectively in narrow, well-instrumented areas of real work. An agent is not just a language model with a task; it is a system that can decide when to use tools, operate in a loop, and retrieve contextual information. This shift marks an enhancement in the practicality and reliability of AI technologies.