AI 导读
长时程智能体不只是多执行几步。 智能体必须评估进展、从陌生情境中学习、记住关键信息、使用工具，并在世界变化时修正自己的计划。 dots3-note preview 是迈向这一未来的一个开放权重步骤。 阅读技术文章： https://t.co/Yi5Nh23pHK 在 OpenRouter 上免费试用： https://t.co/mLEoWCNDVl
正文
Long-horizon agency is more than acting for additional steps.
An agent must evaluate progress, learn from unfamiliar situations, remember what matters, use tools, and revise its plans as the world changes.
dots3-note preview is an open-weight step toward that future.
Read the technical write-up:
https://t.co/Yi5Nh23pHK
Try it free on OpenRouter:
https://t.co/mLEoWCNDVl
来源：@rohanpaul_ai · x.com