devwithjevJev Routes and Scores Agent Work
Jev picks a model, runs a worker, scores the result, and gates tool calls. Each decision ships a small state object rather than the transcript.
View on X cost$22.18 on a reasoning model before, $0.05 after
Carnage@0xCarnagee๐
Jev made 973 decisions in one agent run and generated 0 output tokens doing it same run, same output: $22.18 on a reasoning model before, $0.05 after, and the frontier model was never asked once the loop: prompt โ structured state โ Jev picks the model โ worker runs โ Jev scores the result Jev gates the tool call deliverable each decision ships a small state object, not the transcript. that is why
In one agent run, Jev made 973 decisions and generated 0 output tokens. The author reports a cost of $22.18 on a reasoning model before and $0.05 after; the frontier model was never asked.
Also filed under Agents & browsers
- Watchdog for AI Coding Agents
Jev Watch is a watchdog for AI coding agents that stops Codex or OpenCode when they loop, stall, or drift off task, then resumes the same session with a correction.
- Coding-Agent Runtime Using Jev Decisions
Distill is a coding-agent runtime that uses Jev decisions for model routing, context retention, and tool selection.
- Jevry Desktop Browser
Jevry is a desktop browser where Jev selects executable browser actions while a separate text model handles planning.
- CDP Browser Automation with Jev
jev-browser-bridge is a CDP browser library that uses batched Jev Choice questions to select operations and DOM-derived targets.