Before you press play
What you’ll find in the video
- Give Jev a description of the situation and a bounded question: choose an option, score against a scale or estimate a yes answer. Other software gathers that description, executes the action and reports what changed.
- Check what a demonstration actually measures. The flight example searches rather than buys, and uses another model to write text; the StarCraft experiment pauses the game while Jev decides.
- Test whether a compacted history preserves facts needed later. The viral compaction animation is scripted and never calls the API, so its speed is not a model benchmark. A structured answer can still be wrong.
Worth knowing
This is a secondary survey, not an independent hands-on evaluation. Original creators' costs, timings and success rates remain attributed reports. Pinned public source confirms the flight search and separate typing model, the StarCraft decision pauses and the scripted compaction animation. No projects or workflows were run, and the linked X posts were not independently inspected. Public application code does not mean Jev's model weights are open. These notes use retrieved untimed caption text and selected video evidence, not a full audiovisual watch.