Third Hand is a macOS menu bar tool invoked with Control Space. It uses Apple Vision for on-device screen text recognition, while Jev picks actions from descriptions to type, click, and read back results. Screen text stays on-device, but requests, control labels, and action history are sent.
图标界面读不到控件时，让 Apple Vision 本地补文字、Jev 从描述里选动作，这是 Third Hand 的 computer-use 思路 https://github.com/shhivv/third-hand… macOS 菜单栏工具，Control Space 唤起，能打字点击并回读结果。屏幕文字不出本机，但请求、控件标签和动作历史会发
Gregor Zunic showcases an open-source browser agent combining Browser Use and Jev, finding flights in 7 seconds for $0.0039, using a new action space per step, DOM state space, and small LLM fallback for typing.
Scott Williams predicts custom Jev-style models (parallel constrained decoding) will make existing company agent systems more token efficient, creating millionaires.
The user praises the JEV model for its extremely fast reproduction of each movement and command, believing bots will become smarter as a result.
