A price without its confidence is an incomplete number.

1. There is no price. There are prices.

Bitcoin does not have a price at 09:41:07. It has a print on Binance, a print on Coinbase, a book on Kraken that is two basis points wide and a reference that updated four seconds ago. On a quiet morning those agree to within their own spreads and any one of them will do. On the morning that matters they do not agree, and the venue that picked one of them at random is now lending twenty times against a number it cannot defend.

The honest object is not a number. It is an estimate with a confidence: a mid, and a measure of how far the books that produced it are from each other, how fresh they are, and how many of them are alive.

2. Confidence is not a warning label. It is an input.

A warning is something you read and ignore. An input is something the arithmetic cannot ignore. On this venue the confidence in a price sets the leverage that can be opened against it, the maintenance a position must keep, which orders are accepted at all, and what an exit is expected to cost. When confidence falls, all four move, in the same minute, from the same number, and the screen shows the multiplier that moved them.

3. The venue must say what it does not know.

Four states, published, with their thresholds: Full, Divergence, Degraded, Unavailable. Each is a sentence about the price, and each changes what will fill. The worst thing a venue can do in a cascade is keep quoting the same size at the same leverage because its single feed is still printing. The second worst is to stop without saying why.

So the risk strip names the source that left the others and by how many basis points. The banner says which orders are accepted and why. The log keeps every rejection with its reason. Calm is a design decision: the colours are reserved for the regime and the account, everything else keeps its place, so that on the worst minute of the year you are reading the same screen you read on the best one.

4. Bad debt is the venue's number.

A liquidation rate is a trader's number: how often a position was closed on you. Bad debt is what the venue was left holding after those closes could not be filled. We replayed a hundred and thirty-four market events, minute by minute, through the deployed engine and through a fixed twenty times tier on a single feed. Both open a position at every minute and hold. On Black Thursday the fixed tier liquidated 82% of them and was left with $931,000 unpaid; the engine liquidated 67% and was left with $5,000. The difference is not that it was cleverer about direction. It is that it stopped lending before the books stopped agreeing, and that its maintenance followed the book instead of a flat fraction.

5. Everything that can be shown is shown.

1. The price and its confidence, live, with every source and its deviation, on every market, at all times.
2. The leverage decomposition: the ceiling, what confidence cost, what volatility cost, what the calendar cost.
3. The corpus: every replay, with the data hash, the assumptions and the outcome for both policies, downloadable.
4. The house book: the venue quotes against you on day one and says so, with its inventory and its share of depth in the open.
5. The journal: an append-only record of what the venue decided and when the rules changed.

6. This is what the terminal is for.

Not to make leverage feel safe. To make the state of the venue's knowledge legible at a glance, so that the person trading on it is never surprised by a number they were never shown. Pull a source off the price below and watch what the venue does with that. Then watch a real crash go through the same arithmetic, minute by minute.