CEBE tells you what a common share owns today: the bitcoin, minus every claim ahead of you, over the shares that exist. That is a measurement. Its rules are published and the same for every company, it makes no assumption about the future, and it provides no forecast. The framework draws that line and enforces it as a priority.
The valuation layer sits on top of that measurement and answers a different question: what is a share worth once the instruments the company issued settle? A convertible note is either a claim on the treasury or a set of new shares, a warrant is either nothing or a set of new shares, and today nobody knows which way any of them will resolve. The valuation layer assigns each instrument a probability of delivering its shares, published and dated, and carries it in that expected state, part claim and part shares. The result is Adjusted CEBE per share, and Adjusted CEBE mNAV is the market's price on it.
For each instrument, its probability p carries p of its shares into the denominator and, for a convertible, (1 minus p) of its face in the claims. A warrant has no face to carry. One note, one probability, applied to both sides. The split treats resolution as binary, full delivery or full repayment. A note with combination settlement, a cap or a collar is carried the same way, and the limitation is stated on its row.
Two inputs feed the layer, and they carry different labels because they have different standing.
ADMITTED is the Adjusted Claims % weighting. It passed a pre-registered calibration against filed fair values (Cal-num-3), and its failed cells are published with it. A challenger study sealed on 2026-09-18 (Cal-num-4) scores it forward over four quarters, and a second (Cal-num-5 and 6) tests whether the winner generalizes across issuers. If a challenger passes, this page updates and says so.
PRIOR UNDER TEST is the delivery probabilities. They come from a census of 72 convertibles and warrants across the roster (Con-den-1), grouped by how far in or out of the money each was when issued: in the money, near, or deep out of the money. The observed delivery rate in each class is the prior, and each carries its confidence interval, which is wide because the sample is small. It rests on 26 resolved instruments across the three classes, 11 of them read from a filed status of converted where no delivery size was disclosed. The probability is fixed at the class the instrument was issued in and does not re-score as the stock moves. That is the registered construction, and a time-weighted alternative is published beside it as a sensitivity, not as the headline. The figure you see is running on a prior that has not yet been validated. A study sealed the same day (Cal-den-2) tests those priors as instruments resolve. If a prior fails, it retires, and the page says so. The priors are never refit on the roster they score.
Every figure in the layer carries its label. A number without ADMITTED or PRIOR UNDER TEST beside it is a measurement, not a valuation. The layer is designed to be improved, and it should not let perfection be the enemy of good. It is agnostic to each issuer's own inclusion choices and measures every company in the sector by the same rules. An issuer's diluted model stops at the company that built it.
A valuation inherits everything its base got wrong. That is why the layers are built in this order. A meaningful valuation cannot be placed on a claims-blind measure with the issuer's inclusion rules inside it. It has to apply the same way company to company, and hold across markets and reporting regimes.
Build a model on a fully diluted share count and you have modeled the issuer's inclusion rules: which warrants count, at what strike, whether shares issued but unsold are in or out. Build it on a count that already assumed every note converts and you have counted each note twice, once as debt repaid and once as shares issued. The base has to be a measurement that counted every claim once, under rules written for structures rather than for companies, at a stated price and date. Then the only thing the valuation adds is the probability, and the probability is the only thing you have to argue about.
I welcome the arguments. I have argued with myself through several iterations of this layer. Modifications are attributed in the corrections log, and where the math wins, the standard changes.
That is why the measurement never moves. Nothing in this layer writes back down into CEBE.
Both are models. A fully diluted count adds every share at one hundred percent the day the stock crosses a strike, on the issuer's rules, and takes it back out the day the stock crosses back. This layer adds shares at a published probability with an interval, on a rule that is the same for every company, and it does not change when the stock does.
Labeling an estimate does not make it right. It makes it checkable. One of these publishes its probability, its interval and the study that can retire it. The other publishes its rules when they change.
It does not value an earnings stream. A treasury that puts its bitcoin to work, through options premium, lending, or underwriting, earns income, and the framework carries the income in two places: as an offset in the Wrapper, lowering the bitcoin growth the structure needs to stand still, and in CEBE Sats Yield, which counts growth in sats per share from any source. CEBE Implied P/E is the multiple paid per unit of that demonstrated growth. It is a growth multiple, not an earnings-risk multiple, and it says nothing about whether the income recurs or what it can cost.
What the framework does instead is show what the income rests on. Bitcoin put to work is bitcoin with something attached. The coins carry a state, pledged, lent or written against, and the liability they back enters the claims stack the day it exists. Whether the market's multiple on that income is right is a question this site leaves to the reader, with the claims in view.
The tracker's Valuation view shows Adjusted CEBE and Adjusted CEBE mNAV beside the measurement columns, off by default. The claims page shows, for each company, every instrument in the layer with its class, its prior, its potential shares and its weighted shares, and the instruments left out with the reason. The priors and their intervals are on the Con-den-1 findings page under research, with the hash they were sealed under.
Nothing on this page is a measurement. The measurement, its formula and its date are on the methodology page, and the figures themselves are on the tracker and the stack.