Browser Agent Uses Jev for Click Selection

CyrilXBT describes a tiny open-source browser agent that uses Jev instead of an LLM for every click. The agent found a flight search in 7 seconds at a reported cost of $0.0039.

View on X cost$0.0039time7 SECONDS

CyrilXBT@cyrilXBT𝕏

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 fie

Sep 21, 2026X postsView on X

At each step, the agent uses the DOM state as input and gives Jev available actions as a typed choice question. Jev classifies which action to take rather than generating text.

Also filed under Agents & browsers

Building a Harness with Jev
Building a Harness with Jev Learn all about Jev, a new System One model from TypeSafe AI, and how you can use it in your agent harness. Jev is up to 200x ...
How To Build A Harness With Jev | A LangChain x TypeSafe Conversation
How To Build A Harness With Jev | A LangChain x TypeSafe Conversation Hear from Sydney Runkle, Product Manager, OSS at LangChain, Allie Laabs, Founding Developer Advocate at TypeSafe AI, and ...
Open-Weight Jev-Inspired Decision Model
tev1 is an open-weight Jev-inspired decision model fine-tuned from Qwen3.5 4B.
Jev-kit for Claude Code
Cadaverr’s Jev-kit integrates TypeSafe’s Jev model into Claude Code for tasks including tool-call guarding, sub-agent sizing, file search, and browser automation. Jev answers pick-one and yes/no questions in about 0.3s, according to the author.