Stack
A coin that is valuable because it makes pocket structural inefficiencies cheaper to find and cheaper to diligence — and for no other reason. Not because a thesis pays off. Not because anyone is right. $STACK meters access to a shared detection-and-evidence corpus and pays the people who build it. Every trade run off this machine could lose money and the token would still be worth exactly what it was, because what you are paying for is the cost of finding out.
The correction, stated plainly
The previous design tied the token to whether claims came true. That was wrong, and wrong in an expensive way: it needed an oracle, it locked capital for years waiting on outcomes, it selected for exciting claims over checkable ones, and it walked straight into securities regulation by making the token a bet on securities.
Sell the pickaxe. The token meters a machine that makes the work cheaper, and the machine earns whether or not anybody using it is any good. Bloomberg is not worth less when a subscriber has a bad year. A drilling-data consortium is not worth less when a member drills a dry hole.
So the only question that decides whether $STACK has value is: is finding and diligencing a POSI materially cheaper with this than without it? Everything below is an answer to that and nothing else.
The machine, in three layers
The reason nobody has built this is that it is unglamorous, enormous and long-tailed — every fund running POSI strategies rebuilds a private, worse version of the same three layers, and none of them shares. That is the definition of a thing that should be a commons.
Signals
The twenty-two standing alerts, running permanently. A Form 15. A dividend suspension. A terminated merger. An analyst count going to zero. A 13D amendment. Raw, dated, unclassified events, thousands a month.
Free forever, because the value was never in the event — it is public. The value is in what happens to it next.
Costs to read: nothing. This layer exists to make the book impossible to be empty and to give contributors a queue to work.Dossiers
A Signal becomes a Dossier when somebody does the work: classify it into the archetypes, name the forced actor with evidence, pull the 13F set, read the local-language filing, find the prospectus mandate language, and write the traps.
Versioned, attributed, and assembled by many hands. One person adds the register analysis, another the index rulebook citation, another the translation.
Costs to read: $STACK, per Dossier. Split to every contributor whose attestations are in the version you read.The Book
Every Dossier, ranked by Stack Score — how many independent forcings sit on one name at once. This is the output nobody can compute alone, because it requires the whole classified corpus to exist first.
It is also the thing an allocator actually wants: not a thesis, but a live map of where in the market forced sellers are currently concentrated, and where several are stacked on one ticker.
Costs to read: a $STACK subscription or stake. This is the largest revenue line and the easiest first sale.A Dossier
The atomic unit of paid work. Its sections are the four fields off every POSI card — forced actor, the tell, the clock, the trap — because the corpus should have the shape of the taxonomy. Every factual line is an attestation: contributed by someone, checkable against a document, and challengeable by anyone.
The refutation bounty
The single design opinion this whole thing rests on, now that outcomes are gone.
Attacking a Dossier pays three times what adding to one pays.
Every research venue ever built pays for advocacy. You get paid to publish the case, and nobody is ever paid to kill it — so nobody does, and a corpus of un-attacked theses is worth close to nothing, because the reader has to do the entire diligence job themselves anyway.
Here the highest-earning activity on the platform is finding the thing that makes a Dossier wrong: the covenant in note 14, the fourth holder who is not actually mandated, the amortisation that does not run off when the author said. A Dossier that has survived twenty paid attacks is worth paying to read. One that has survived none is a first draft, and the Book marks it as such.
Stack Score
The flagship output, and the one number the corpus exists to produce. The POSI page’s counting discipline says stacked archetypes must be independent forcings — if your three are three descriptions of one selling event, you have a stack of one. So the score discounts archetypes drawn from the same family. Try it.
Why the token has value when nobody is right
The question the whole design has to answer. There are four sinks, and not one of them references an outcome, a price target, or whether a POSI paid off.
| Demand sink | Who spends | What they get | Independent of outcomes because… |
|---|---|---|---|
| Metered queries | Anyone screening | per query | You pay to ask “every name with 2+ independent forcings and a verified forced actor.” The answer has value whether or not you then make money on it — same as any screen you rent today. |
| Dossier access | Readers | per Dossier | You are buying assembled evidence and a list of attacks it survived. That saves you a week of work regardless of what you conclude from it. |
| Book subscription | Funds, allocators | stake or subscribe | A live map of where forced sellers are concentrated. It is a dataset, priced like a dataset. Nobody asks whether a dataset was right. |
| Bounty posting | Anyone with a question | directed spend | “Pay 40k $STACK to whoever proves or disproves that this register is mandated.” You are buying labour on a checkable fact, with a deadline. |
| Everything collected | → contributors | Fees route to the people whose attestations were actually read and actually survived challenge. The treasury takes a cut for gap bounties and pipeline upkeep, nothing else. | |
The loop
Four nodes, and the only thing that has to be true for it to turn is that the corpus saves its readers more than it costs them.
Usage
Someone queries the Book, buys a Dossier, or posts a bounty. $STACK is spent.
Rewards
Fees route to contributors whose attestations were read and survived attack — plus the attackers. Doing the boring work pays.
Corpus
More verified attestations, more classified names, more surviving traps. Coverage widens and depth compounds.
Cheapness
Finding a stacked, evidenced POSI gets measurably cheaper than doing it alone. Which pulls more usage. Back to 01.
What got cut, and why
Held against the rule at the top — if it only pays off when somebody’s thesis works, it does not belong. Three things failed it.
Betting on outcomes
Bonds slashed when a thesis turned out wrong, payouts from a resolution pool.
Why: it needed an oracle, which was the single most likely thing to kill the protocol; it locked capital for years; and it made the token’s value a function of other people’s luck.Price targets and theses
Dossiers carry no view, no target, no rating. They carry forcings and evidence.
Why: the moment the corpus contains recommendations, the incentive to inflate them appears, and the whole thing drifts back toward advocacy. The Book maps; it does not advise.Per-thesis tokens
No coin per name, no curve per idea, nothing tradable that tracks a listed security.
Why: that is the manipulation design. It is also unnecessary now — the thing being sold is the corpus, and the corpus does not need a ticker attached to have a price.Does it need a token at all
Asked honestly, because the answer is narrower than it was for the previous design and there is a case where the answer is no.
Contributed-commons economics
The real argument. The corpus is too broad and too long-tailed for one team: it needs the person who reads Korean filings, the one who tracks aviation-lease dockets, the one who knows the index rulebooks cold.
Those people will not spend years building a moat for a company that can change the terms later and keep it. Ownership is what makes contributing rational — it is the Reddit-API problem, solved in advance.
Global micro-payments
Thousands of small payments to pseudonymous contributors across dozens of jurisdictions, settling automatically against usage. Fiat rails do this badly, slowly, or not at all.
Challenges need slashable stake
A disputed attestation is only meaningful if being wrong costs the challenger something. Note the difference from before: the stake is on a checkable fact, not a forecast — so it settles in hours, against a document.
Where the answer is no
If a well-capitalised incumbent simply builds this centrally and pays a hundred analysts, they win. FactSet could. The token version is not better technology — it is a different labour-supply model, and it only wins if the long tail genuinely cannot be hired.
That is a real bet and it should be stated as one rather than assumed away.
What kills it
A different list from the previous design, because removing outcomes removed the two worst risks and promoted a new one to the top.
Paid contributions become slop
The moment you pay per attestation you get volume: thin, technically-true, useless facts filed by the thousand. This is now the most likely cause of death.
Mitigation: pay on read-weighted attestations, not submitted ones — if nobody queried it, it earned nothing. Plus the 3× refutation bounty, which makes hunting bad attestations more profitable than filing mediocre ones.Leakage kills the paywall
A Dossier is text. One subscriber can paste it into a group chat and the marginal buyer disappears — the failure mode of every research subscription ever sold.
Mitigation: price the Book and the queries, not the prose. A live, continuously-updated ranking is much harder to leak usefully than a PDF, and it decays the moment it leaves the machine.Cold start
An empty corpus saves nobody anything, and the first hundred contributors work for a token with no usage behind it yet.
Mitigation: Signals run from day one so the queue is never empty, and the treasury pays gap bounties on the boring archetypes directly — buy the corpus before you need the demand.It drifts back to advocacy
Contributors will want to write theses, because theses are more fun than attestations, and readers will ask for them.
Mitigation: the schema simply has no field for a view. You cannot file an opinion because there is nowhere to put one — the same trick that makes the date field unskippable.One name, end to end
Note that at no point does anybody have to be right about anything.
- Signal. The dividend-suspension alert fires on a $340m industrial. Auto-posted, free, unclassified, sitting in the queue with four hundred others that week.
- Classification. A contributor tags it [01] dividend-cut exile and files the first attestation: three income-mandated holders, 22.4% of the register, prospectus language quoted and linked.
- Second forcing. Someone else notices coverage went 4 → 0 in May and checks why — the broker exited the sector. Files [09] coverage cliff. The Dossier now carries two archetypes from two families. Stack score 2.4.
- Attack. A third person reads note 14 and files a refutation: covenant headroom is thinner than the deck implies, so the cut may be defensive rather than offensive. Paid at 3×. The Dossier is now more valuable, not less.
- Dispute. A challenger bonds against the claim that a fourth holder is mandated. It resolves that afternoon against the fund’s own prospectus. Somebody is wrong about a document, which takes hours, not quarters.
- Demand. A fund screening the Book for two-plus independent forcings with a verified forced actor pays $STACK for the query, then for the Dossier. Fees split across every contributor in the version they read, plus the attacker.
- And then? The fund buys the stock and loses forty per cent. Nothing about the token changes. The machine did its job — it made a hard question cheap to answer, and the answer to the different question of what to do about it was never for sale.