jev / gallery
← Back to collection

Author test

jev-ultrafast

Let Jev choose browser actions

How this project uses Jev

Input
Task goal and DOM controls
Jev decides
Action type, target element, and completion
Code executes
The browser executes actions; Mercury enters text

EDITOR’S NOTES

Where to start

Who it is for

For browser-agent developers exploring a separation between action selection and text generation.

Setup requirements

Requires a browser execution environment and the model services specified by the project. Jev selects actions while Mercury assists with text entry. Check dependencies and costs in the repository first.

Suggested first experiment

Start on public pages without logins, purchases or messaging. Record the goal, controls, model choice and execution result; compare task completion and full elapsed time.

Before adopting

The reported 7.092 seconds comes from three runs of one flight task, excluding initial navigation. Remeasure on other pages and require confirmation for consequential actions.

Editorial guidance based on linked sources; not an independent runtime test.

Evidence and limitations

The author reports a 7.092-second median after optimization across three paired runs of one flight task. Initial navigation is excluded; this is not a general browser benchmark.

This project has not been run independently here. Author-reported results are not independently verified results.

Original sources

Explore related use cases

Browse this category →

Learn the approach