Every decision engine we have built, in one place: the Token Decision Index and the BTC Up/Down engine — both judged live by TypeSafe's Jev System One model — plus the deterministic Smart Play model, board decision signals, prediction markets, and the full stored history of every verdict, kept in our Convex backend so the record can be checked, not just claimed.

| Token | Price | 24h | Age | |
|---|---|---|---|---|
| Loading the tape… | ||||
Fresh launches from the Musebook relay tape. Brand-new tokens often have no DexScreener pair yet — Jev can only judge a token once a market exists.
The green dashed path on the chart is the V-recovery scenario — drawn whatever its probability. Jev's probabilities are the forecast; the line is just the shape of that one path. Verdicts refresh as BTC crosses $250 buckets and are stored to Convex below.
Yes / no — the answer is a probability. “Will BTC close higher in 24h?” → { noul: 0.36 }. Code can branch on it; humans can read it.
One winner from a fixed set — with the full probability spread and a confidence. “Next path: V-recovery / continued down / chop” → every path priced, not just the winner.
Position over ordered levels — “bounce strength: weak / moderate / strong V” → a probability-weighted score across the legend, not a vibe.
TypeSafe Jev (System One, jev-latest) makes the typed judgments. The board's own Smart Play model is deterministic and published-factor — the two are shown side by side, never blended into a black box.
Every verdict is appended to our Convex backend the moment it is made, so calibration can be checked against what price actually did. Supermemory carries the longer research memory the desk recalls from.
Perps state from Imperial, Phoenix, Jupiter and Backpack (marks verified against on-chain accounts on the perps board); token state from DexScreener; candles from Coinbase / Kraken / Bitstamp.
