2/ Cua Driver + Jev playing 2048.

In this measured 2048 run, Jev was 5x faster and about 1,000x cheaper than Astra. That result is specific to this run, not a general claim about either system.

Video attachment: 30 seconds, silent, 1920 x 1080. The final frame shows Jev at 44.9 seconds and about $0.00108 API-equivalent cost, versus Astra at 294.9 seconds and about $1.45.

Demo by @injaneity

3/ Jev from @typesafeai is the fast decision layer.

Cua Driver observes and executes across macOS, Windows, and Linux. jev-use turns what Driver sees into bounded action candidates Jev can select, then Driver executes and verifies.

4/ For universal GUI parsing, we're introducing optional Cua Perception.

It can use screenshot parsers such as OmniParser to produce typed visual regions when AX or DOM is incomplete. Cua Driver stays model and parser agnostic.

5/ Cua Perception is opt-in. Jev can use it when a screen needs visual parsing and skip it when AX or DOM already provides enough structure.

6/ The Jev action loop stays bounded:

1. Driver observes
2. Cua Perception parses if needed
3. The client builds candidates
4. Jev selects one ID
5. Driver executes and verifies fresh state

7/ PR #3943 lays the foundations Jev can build on: optional extensions, typed visual-region contracts, bounded captures, a fixture Rust worker protocol, and release safeguards.

8/ This is the start of jev-use.

Next: public parse and runtime wiring, an optional OmniParser/model install, and Jev provider integration.

Draft: github.com/trycua/cua/pul…
Guide: github.com/trycua/cua/pul…
RFC: github.com/trycua/cua/pul…