Jev 开放注册：用类型化决策替代文本解析
Jev 不写段落，只返回 Choice/Score/Noul 等带置信度的类型化决策，作者用邮件分流 agent 举例说明调用方式与价格。
Jev is now open to everyone. Here's what it actually is. AI models read your text and write text back. Then your code has to parse that text and hope it's in the right shape. Jev skips that. It reads plain text or JSON, but it never writes a paragraph. It answers with typed decisions your code can use directly. Say you are building an email agent and this lands in the inbox: "Loved the demo! Can you send pricing? We want to start next week." You give it to Jev and ask: – Which team should get this? (a choice) – How excited is this person, 0 to 10? (a score) – Is this a hot lead? (yes/no with a probability) It answers all of it in one go, in about 70 to 500 milliseconds, with a confidence level on every answer. So your code can act when it's sure and send it to a human when it's not. Think of it as an AI you call like a function, not one you chat with. Super cheap: input costs $0.042 per million tokens. Output is free.
这条是引用着另一条写的：「Jev 正式开放，取消等待名单」