the oracle

Ask it anything. It answers with conviction, never twice differently, and keeps score of how often it was right.

The conceit is the mechanism: a verdict is derived from the question, not drawn from chance. sha256(normalized question) picks the stance, the confidence (52–98 — it never admits to a coin flip) and the omen it claims to have read. So asking twice gets the same answer, "Will it rain?" and WILL IT RAIN!!! are one question, and the ledger can be scored: every verdict can be marked right or wrong afterwards, and the front page reports what the oracle actually achieved against what it claimed — Right 61%, claimed 84%. It talks a better game than it plays.

That last line is the only real thing here. The stances are theatre; the calibration is arithmetic over what you told it.

Using it

Ask on the front page; you land on the verdict's own page, which is a permalink you can come back to when you know how it turned out. /ledger holds everything, filtered by all / unjudged / right / wrong.

Running it

PORT=8080 FORGE_APP_DATA=./data bun run src/server.ts   # run it here
bun test                                                # the check every pull runs
FORGE_APP_DATA=./data bun run src/tty.ts                # the terminal interface

It also has a terminal — the exact command is in FORGE_APP_SSH when the app runs, and the footer prints it. j/k move, a asks, r/w score the verdict under the cursor, u takes a score back, q leaves. Same ledger as the web page.

Shape

Everything it keeps lives under FORGE_APP_DATA, which is what makes production data and preview data two different things: a preview starts as a copy of production's verdicts and never writes back.

No dependencies. No JavaScript on the page.