Changelog

The desk was called Basis. The name collided with basis points — the unit every term of the schedule is quoted in — so a reader met the same word meaning two different things on the same table, and the one that mattered was the unit. It is now Nobid, which is what the headline on the front page has said all along.

Nothing measured changed and no term of the schedule changed. What did change is worth an entry: the consent record moved from basis.consent.v1 to nobid.consent.v1, and the cookie from basis_consent to nobid_consent. Old keys are not read, so an answer given before today is not carried across and the question is asked once more. The alternative — reading the old key silently — would have meant inheriting a permission granted to a different name.

/api/book returns the survey as JSON, /api/status returns the desk’s own state, and /api/quote runs the published schedule over a position described in the request. No key, no account. They are documented at API.

All three read the same files the pages read, so an answer from the endpoint and a figure on a screen cannot disagree. If they ever do, the endpoint is the thing that is wrong, and this line is where to start looking.

The book, the desk, this documentation, and the session clock — which works the US regular session out from the exchange calendar and stops at 2026-12-31 rather than asserting a market it has no table for.

The standing disclosure sits at the head of every documentation page and in the footer of every screen: the book is published, the desk is not open, and nothing here is an offer. It is written once and repeated verbatim rather than paraphrased into something softer further down the page.

Every term as a number with a name: venue at 145 bp against an equity token the desk can offset during the session and 420 bp against one it cannot; the out-of-hours term at 202 bp, which is z = 0.84 against an assumed overnight σ of 2.4% and nothing else; a size ladder in bands, taking nothing below 2,500 USD and stopping at a conversation above $250,000; a curve term that rises as the depth beneath the position thins; and the desk’s own fee at 25 bp, the only line that is revenue rather than risk.

All-in for the case this desk exists for — a position with no venue, quoted out of hours, in the standard size band — is 682 bp. The lines add rather than compound, so the arithmetic can be done on paper by anyone who wants to argue with it.

Being the first version, nobody has been quoted under an earlier one. That will not be true of the second entry of this kind, which is why this page was written before it was needed.

The aggregate split is measured. The per-asset breakdown is not: establishing which equity token each of the 115,094 stranded curves is denominated in needs a quoteToken() read per curve against an RPC that limits by IP, and that run is scheduled for 2026-09-14. One quote asset is confirmed by direct call — AMC — and it is the only one this site names.

census.complete is false in the book and in the API. A plausible breakdown would have been quicker to publish and is the single thing that would have made everything else here worthless.

A full enumeration of the Pons factory. 244,487 launches recorded, 240,350 still holding a reserve, 4,137 retired. Of the live population 125,256 are quoted in ether, selected across the population by the constant product k = V × S = 1.68e9 — identical for every curve denominated in the same asset, and therefore a fingerprint — and cross-checked by direct isNativeQuote() calls on the shortlist.

That leaves 115,094 curves, 47.89% of the live chain, priced in tokenised equity with no pair against WETH or USDG on the canonical factory. The reference for the day was 2459.19 USD per ether, stated so that every dollar figure on this site can be re-derived instead of trusted.

Recorded in the same pass: the deepest curve on the chain prices in AMC and reports a reserve of 1,625, which a screener assuming ether prints as $3,996,184.