Rohan Paul@rohanpaul_ai搜索原本围绕 1 个人输入 1 个查询并检查结果来设计。 AI 智能体并不这样使用它。它们同时通过 50 条路径探索 1 个问题。 如今的搜索 API 仍然通过端点复制相同的人类流程。 Octen 为机器重建了整个栈：索引、排序和服务层。 每次搜索 62 毫秒。每 1000 次调用 1 美元。 界面早已进化。引擎终于也跟上了。
Search was designed around 1 human entering 1 query and checking results.
AI agents do not use it like that. They explore 1 question through 50 paths at once.
Today's search APIs still copy the same human process through an endpoint.
Octen rebuilt the stack for machines： index， ranking， and serving layer.
62ms per search. $1 per 1，000 calls.
The interface moved on long ago. The engine is finally there too.
Rohan Paul@rohanpaul_ai · X搜索原本围绕 1 个人输入 1 个查询并检查结果来设计。 AI 智能体并不这样使用它。它们同时通过 50 条路径探索 1 个问题。 如今的搜索 API 仍然通过端点复制相同的人类流程。 Octen 为机器重建了整个栈：索引、排序和服务层。 每次搜索 62 毫秒。每 1000 次调用 1 美元。 界面早已进化。引擎终于也跟上了。
Search was designed around 1 human entering 1 query and checking results.
AI agents do not use it like that. They explore 1 question through 50 paths at once.
Today's search APIs still copy the same human process through an endpoint.
Octen rebuilt the stack for machines： index， ranking， and serving layer.
62ms per search. $1 per 1，000 calls.
The interface moved on long ago. The engine is finally there too.