The question is always the same: will this token reach 2× its price within the next 30 minutes? Jev, TypeSafe's System One model, reads each token's name, description and links and returns twelve typed judgments in about a tenth of a second. On-chain facts supply the rest. A classifier turns both into one probability from the t+2m state, frozen within 60 seconds and settled at t+30m, so the record below can never be rewritten.
The most recent launch Jev has judged. Each line is one question Jev was asked; the bar is its probability.
Jev doesn't write, summarize, or reason out loud. It takes a small JSON state — this token's text — and answers twelve narrow questions with calibrated probabilities: is the description empty or a template, does it promise gains or claim to be safe, is the name a copy of a known coin, what's the theme, how much effort, how much hype.
It is never asked whether the token will run. Descriptions are written to persuade, so that verdict belongs to the classifier, which sees Jev's answers next to what the curve actually did in its first two minutes.
Every row is a probability from the t+2m state, written within 60 seconds. Nothing is edited afterwards; the outcome fills in at t+30m.
| frozen | token | P(run) | entry mcap | now | peak | outcome |
|---|