Jev 创始人讲助理之后：决策层
Diogo Almeida 用 36 分钟讲 agent 里大量工作只需决策而非生成，Jev 在输入与执行间插入 State→Decision→Action→Feedback 层。
JEV founder Diogo Almeida spent 36 minutes explaining what comes after AI assistants. The question worth paying attention to: how much of your agent’s work actually needs a generated answer? Routing a task. Scoring an option. Choosing the next action. Those decisions sit inside almost every agent workflow. JEV’s approach puts a dedicated decision layer between the input and execution: State → Decision → Action → Feedback For me, that’s the interesting angle: how you build a system that can keep moving without asking a person to interpret every response. If you’re building agents, this is worth a spot in your watchlist. Watch the talk, then look at your own setup. Where are you generating paragraphs when you only need a decision?
这条是引用着另一条写的：「用 Jev 批量筛交易机会：600 个 setup 47 秒花 0.07 美元」