The post describes @awlevin's trick: OCR the screen, then let TypeSafe's Jev pick the next click. It compares costs with Opus 5 and suggests writing a handcrafted date parser for better value.
Computer use for $0.0002 a step. Opus 5 on the same screenshot: $0.032. @awlevin's trick: OCR the screen, then TypeSafe's Jev picks the next click. Catch: Opus reads dates off pixels. Here you code it by hand. For 155x, I'd write the date parser. https://github.com/awlevin/typesafe-computer-use…
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.
