Community · E2 · artifact verified
Feel Jev speed and cost in a 25-second game
A tiny Japanese game with no send button: type IT buzzwords and each keystroke gets a Jev judgment that stretches a meter, so 25 seconds of play answers what curl never does - how fast and how cheap the model feels inside a real app.
01 · Role in the system
What Jev does here
The README's premise is that a curl one-liner tells you a model is fast but never whether it works in an app, and a price table never tells you what a call actually costs. So the demo is a game: every term you type fires a real Jev judgment, the meter grows with each verdict, and a session counter accumulates real latency and real cost - 25 seconds later you have felt both numbers instead of reading them. Judgments are typed choices over the buzzword's meaning, served from a small Python server to a static web page. It is a demonstration instrument rather than a product: the measured thing is the model itself, in-app, under play.
02 · Control boundary
Where Jev sits
Jev as the judged action inside a latency-and-cost demonstrator: every input triggers a real typed judgment, the UI meters speed while the server accumulates cost, and the session total is the demo's payload.
Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.
03 · Known limits
What this evidence does not prove
- A demonstration instrument; it measures the model, not an application workload.
- Japanese UI; judgment set is IT buzzwords.
- Created 2026-10-01, 0 stars; self-hosted server required to play.
04 · Attribution
Public sources
This is a Community record: the project was published by a third-party community author.
- pochang6 ↗Community · github · public · checked 2026-10-01