CASE FILE / Model & agent routing

Okay so Jev can actually do computer use really well Without any screens…

Okay so Jev can actually do computer use really well Without any screenshots, or LLMs and no Pixels leave my mac I dont even read the Dom elements A local CoreML model segments every button and UI element on screen. On-device OCR reads the labels. That text is all Jev gets. It returns a probability across those elements and tells me the best one to click. Then it clicks, re-runs detection, and decides again. In a loop until the goal is done. ~90ms per decision. Faster than any LLM computer use I've tried. Blazing fast computer use, without any latency @typesafeai is building something really interesting

By Milind S @milindlabs

● ● ●CASE FILEGUIDE

Milind S · @milindlabs

Original post

Okay so Jev can actually do computer use really well Without any screenshots, or LLMs and no Pixels leave my mac I dont even read the Dom elements A local CoreML model segments every button and UI element on screen. On-device OCR reads the labels. That text is all Jev gets. It returns a probability across those elements and tells me the best one to click. Then it clicks, re-runs detection, and decides again. In a loop until the goal is done. ~90ms per decision. Faster than any LLM computer use I've tried. Blazing fast computer use, without any latency @typesafeai is building something really interesting

Views 111,583Likes 1,619Reposts 85

Chinese translation

好吧，Jev实际上可以很好地使用电脑没有任何屏幕截图，或者LLM和没有像素离开我的Mac 我甚至没有读过Dom元素本地CoreML模型分割了屏幕上的每个按钮和UI元素。设备上OCR读取标签。这条短信就是杰夫得到的全部内容。它返回这些元素的概率，并告诉我最适合单击的元素。然后它单击，重新运行检测，并再次做出决定。循环，直到目标完成。每次决定~ 90 ms。比我尝试过的任何LLM计算机都快。快速使用计算机，没有任何延迟 @typesafeai正在构建一些非常有趣的东西

What this case shows

Let JEV choose the model or tool first
Auto-route on high confidence, escalate otherwise
This is the most common production pattern