A benchmark of the Jev model for browser automation shows it is 13x faster than LLM-driven approaches because it returns probability distributions over options rather than generating text tokens. The speed advantage comes from reduced output, not raw model speed, though a critical setup mistake—truncating candidate lists—silently destroyed accuracy by excluding correct answers.
Shobr is a job search automation CLI tool that uses browser automation via your authenticated daily-driver browser, event sourcing for data management, and minimal LLM integration to automate job applications while maintaining human control over final submissions. It emphasizes stealth, Unix philosophy, and provider-agnostic LLM support.
Jev-browse is a tool that enables coding agents like Claude Code to perform browser tasks more efficiently by batching multiple steps into single calls handled by a lightweight TypeSafe Jev model, reducing costs by 2.1–5.3× and improving speed 1.6–3.0× compared to traditional agent-driven browsing.
Ego-jev is a browser agent skill that makes typed decisions in ~0.4 seconds per DOM step using TypeSafe's System One API, replacing full LLM calls. It numbers interactive elements, makes one API call to pick an operation and target together, then executes via ego-browser, escalating complex tasks like logins and payments back to the planner.