Jev in Practice: Browser Use and Agent Workflows

A Solana project lets users build their own Jev classifiers and strategies, backtest them against new token launches, run them live in a terminal, and rank them on a leaderboard.

The author shares a link to TypeSafe's official agent-skill documentation for developers who want to integrate Jev into their own agent framework.

The author proposes having Jev compute the probability that each sensitive agent decision is a scam or prompt injection and block it, similar to Sentinel but editable and reusable across contexts.

The author argues that as Computer Use costs fall and Jev's latency drops, agents could cheaply migrate user data between platforms, so context may no longer be a moat.

After testing eight or nine models in a week, the author says Jev, a specialized classifier from TypeSafe AI, beat GPT-5-mini on speed, efficiency and price and now handles option selection for their agent.

The author built a chatbot with no LLM at all, using Jev to pick which tool to call and with what arguments across web search, Wikipedia, weather, Todoist and Home Assistant, with instant cited answers.

A demo shows Jev playing 2048 and making beginner mistakes without an advanced harness, which the author finds fun to watch.

Scarlett.ai offers free API and MCP keys, with a 12B model fine-tuned on trading data for understanding and Jev fine-tuned on paper outcomes for trade decisions, ahead of live trading on Hyperliquid.

A video explains how Jev works as a cheap, fast classifier that replaces long LLM prompts in workflows with more deterministic answers, with several demos.

Browser Use open-sourced Jev Ultrafast, which picks the action and target in one shot instead of calling a large model at every step, cutting browser calls from 1,092 to 101 and searching Google Flights in 7.1 seconds.

effect-agent now has a Decision abstraction layer, and Jev is currently the only provider for it.

The author says they are starting to understand Jev's use cases and believes it will be huge in trading, calling the potential generational wealth.