SOMEONE HAS TO CALL IT.

Every 600-second Squeekchain round, this bot calls what the hamster will do — before the round closes — publishes the call with a hash it cannot edit afterwards, then grades itself in public against what the wheel actually did. You can call the round too, from your own wallet.

Reads api.squeekchain.com for live telemetry and Sepolia's SqueekCore for the certified frontier. Signature-only: no gas, no transaction, no custody.

ROUND … · CLOSES IN –

waiting for the poller…

HASH PUBLISHED AT SEAL TIME–

RECOMPUTES NOW–

why this call

LAST GRADED ROUND

nothing graded yet

STREAK–

AI ACCURACY–

CLIMATOLOGY BASELINE–

LIVE WHEEL

STATE–

RPM (milli)–

SPEED (mm/s)–

SAMPLE–

CERTIFIED FRONTIER–

IDENTITIES–

CALL THE ROUND YOURSELF

WHERE IT IS RIGHT AND WRONG

CALLED	TIMES	CORRECT

CALIBRATION

pActive	N	ACTIVE

HUMANS VS THE BOT

#	WALLET	CALLS	HITS	PENDING

Your call is graded by the same rulebook as the bot's. Late calls are refused by the server, not by trust.

THE LOG (RECOMPUTE ANY ROW)

ROUND	CALLED	SAW	WORK	OUTCOME	HASH

Every row ships the call, its nonce and its hash, so the seal can be checked field by field. The honest limit: the log lives on our server. A stronger version would anchor each hash on-chain, and we say so here instead of pretending otherwise.

HOW IT WORKS

Sample — /v1/status every 15 s: wheel state, rpm, speed, cumulative work.
Call — in the first 2 minutes of a round, the model picks a band from half-hour climatology (7-day /v1/measurement/stats), the live wheel state and the last rounds' work.
Seal — sha256(round, pActive, expectedWork, bucket, nonce) is stored and returned by the API while the round is still open.
Grade — when the round ends, work added during it (from our own sampling) decides the band, and the call is scored — alongside a climatology-only baseline, so you can see whether the live signal actually helps.