开源浏览器 agent 用 Jev 做点击决策，7 秒找到航班
每步把 DOM 状态变成类型化选择题交给 Jev 分类选动作，只在输入自由文本时才调 LLM，单次任务成本 0.0039 美元。
A tiny open source browser agent using Jev instead of an LLM for every click. Found a flight search in 7 SECONDS. Total cost: $0.0039. Here's why that's not a typo. A normal browser agent asks a chat model "what should I click" on every single step. That's a full generation call, just to pick a button. Mine doesn't. The DOM state at each step becomes the input. Jev gets the available actions as a typed choice question. It picks the action, not by generating text, by classifying against what's actually on the page. The only place a language model still runs is typing free text into a field, an actual search query, a name, an address. That's generation. Everything else is a decision, and decisions don't need a chat model anymore. New action space every step. DOM state space in, typed choice out. Small LLM fallback only where typing is genuinely unbounded. (video below is real time, not sped up) It's open source. Try it yourself, link below. Follow @cyrilXBT if you're building agents and still routing every click through a full LLM call, there's a cheaper way to do most of it. https://x.com/gregpr07/status/2100411066966749359/video/1
这条是引用着另一条写的：x.com/1373665408193036293/status/210151513079829…