Three Rupees
17 May 2026A finance minister announces a three rupee hike in petrol. The campaign that just ended — the one whose victory permitted the hike — ran on a no-shortage, no-price-rise frame, held tight for months, repeated in speeches and rallies and prime-time panels until the repetition was itself the thing being trusted. There was no shortage. There would be no price rise. Anyone who said otherwise was either malicious, foreign-funded, or insufficiently patriotic. Now three rupees.
The hike is small. The reaction is not.
Petrol at three rupees more is not, in any material sense, a regime crisis. The household budget absorbs the shock. The supply chain absorbs the shock. The macros absorb the shock. What does not absorb the shock is the field of adjacent claims the campaign had been making in the same voice — about food, about jobs, about the rupee, about the fiscal headroom, about everything that had been held in place by the same apparatus of repetition. Three rupees is the demonstration that the apparatus is a performance. Once the apparatus is recoded as performance, every claim it ever made gets its priors recoded by association.
This is the part the news ticker cannot count. The petrol price is priced in the petrol price. The signal is not in the petrol price.
The puzzle Pelican was built for
In 1989 Cutler, Poterba and Summers published a paper with a question for a title: What Moves Stock Prices? Their answer, which they declined to fully draw out, was that nobody knows. Large moves often had no identifiable news. Large news often produced no move. The dataset was American, mid-twentieth-century, well-curated, and the embarrassment held across every reasonable slice of it. Markets moved. Reasons did not always exist.
The quants spent the next thirty-six years building machinery around the embarrassment without resolving it. ARCH gave the volatility a memory. GARCH gave it persistence. Stochastic vol of vol gave it a hyperparameter. Regime-switching gave it a wardrobe. Each of these was, in its way, an admission: there is a hidden state. The state explains what news cannot. The state, named, retains its mystery — the formalism describes how the state behaves without saying what the state is.
The state is the load. What loads it is not information. It is acknowledgement, or the lack of it, distributed across the field.
This is what Pelican was built to register. Not predict — register. The distinction matters more than it sounds.
The village knew
Robert Aumann’s 1976 result is at first reading a curiosity. Two perfectly rational agents with common priors cannot agree to disagree once their posteriors are common knowledge. It sounds like a logic-puzzle artifact. Michael Suk-Young Chwe, in Rational Ritual, gave the result its operational form: public rituals exist not to inform but to synchronize. A coronation does not tell anyone the king is king. It establishes the common knowledge that everyone knows that everyone knows. Coordination requires that further layer. Mutual knowledge — everyone knows X — is too weak to bind. Common knowledge — everyone knows that everyone knows X, recursively, all the way down — is the only kind on which action can stand.
The implication for markets is that price discovery on a privately-held belief is fundamentally different from price discovery on a publicly-acknowledged one. A widely-suspected fact prices probabilistically and badly. A publicly-stated fact resets the Keynesian beauty contest at every level simultaneously. Everyone has to trade on it because everyone now has to trade on it because everyone now has to trade on it.
News, in this frame, does not produce information. It produces synchronization. The market does not learn anything at 9:34 AM that it did not suspect at 9:33. What changes is that the suspicion has become a thing on which it is now safe to coordinate.
The reign of the sweepable is the field of facts in mutual but not common knowledge. Everyone knows. Nobody knows that everyone knows. Or — more precisely — everyone knows that everyone knows, but the agreement not to act on it is itself a sweepable, and the agreement not to lift the agreement is sweepable, and so on through whatever depth the carpet stack happens to be.
A carpet, in this sense, is not a thing. It is a reciprocal restraint. I do not lift mine; you do not lift yours; we both know we are not lifting; we both know we both know we are not lifting; we both proceed about our business. The reciprocal restraint requires energy to maintain. It is held in place by an apparatus — language, ritual, institutional reward, threat of social cost, the dignified murmur of opinion columnists, the patriotism of the moment. The carpet is held down by the apparatus; the apparatus is held in place by the willingness of the held-down to be held down. A material fact that can no longer be swept under the carpet is not a fact that has changed. It is a carpet that has lifted.
The part the literature has not quite said
When the carpet lifts, what becomes common knowledge is not just the fact underneath. It is the existence of the apparatus that was holding the carpet down. The flip carries two payloads. The first payload is the fact: three rupees, Biden’s cognition, the subprime portfolio, the body in the orchard. The first payload is what the news ticker counts. The second payload is what makes the move large: the demonstration that an apparatus of suppression exists, that it was operating on this fact, and — by extension — that it is operating on others.
This is the move the existing literature has not made cleanly. Not just: we all now know X. But: we all now know that we were all in the business of pretending not to know X, and we now have to ask what else the same apparatus has been holding. Every adjacent carpet becomes brittler in the same instant. Every other denial in the loaded field is recoded as suspect, because the apparatus that was producing them has been demonstrated breakable.
Call this meta-common-knowledge — common knowledge that mutual knowledge is being suppressed. Soros gestured at something like this with reflexivity and left the mechanism vague. The mechanism is not sentiment feeding back into price. The mechanism is the recoding of an entire field of denial on a single demonstration of its breakability. One carpet lifts. The whole room is implicated, because the whole room shares the apparatus.
This is also, I think, the cleanest formal account of what we have been calling displacement. Institutions do not resolve contradictions; they displace them. The displacement requires reciprocal restraint at every interface where the contradiction might surface. The Barber of the village does not resolve who shaves him; the village restrains itself from asking. The restraint costs nothing as long as the restraint is reciprocal. It costs everything when the reciprocity breaks at any point in the chain.
The small flip is the stronger signal
It follows — and this is the diagnostic that makes the framework instrumentable — that the small flip is more telling than the large one.
A three rupee petrol hike that produces a regime-wide reset is a stronger brittleness signal than a ten percent GDP contraction that produces panic. A ten percent contraction is, on its own terms, enough to account for the move. A three rupee hike is not. The residual is the field reloading. The residual is the count of adjacent carpets being recoded as lift-able. Large news on small moves means the field is not loaded; the apparatus is intact and absorbing the shock. Small news on large moves means the apparatus is brittle and the field is dense with suppressed facts.
Move-magnitude over flip-magnitude is the ratio that carries the signal. The old ratio — move over news-importance — is contaminated by forty years of ARCH-shaped commitment to news as the cause, and reading it produces only what it was designed to produce.
Phase transitions are not punctual
The temptation, having developed the carpet image, is to imagine a single threshold: the carpet either covers or it does not, the field either holds or it pops. The temptation is wrong. Phase transitions in real systems are not punctual. They are smeared, sometimes across decades, sometimes across regimes that themselves change while the transition is in progress.
In March 2008 Bear Stearns failed and the field absorbed it. In September 2008 Lehman failed and the field did not. Between March and September the suppression-hypothesis had become common — not in the form of any specific public statement, but in the cumulative weight of small flips: ratings agency credibility, monoline solvency, off-balance-sheet vehicle behavior, the persistent failure of the no-correlation story. By Lehman, each new carpet was easier to lift than the last. Subprime had been swept and swept and swept; each successful sweep made the next flip cheaper, until a single flip lit the field.
The USSR did not pop. It leaked loudly for a decade. Glasnost was a controlled lifting that turned uncontrolled; each lifted carpet exposed an apparatus that had been holding more carpets than anyone wanted to count; the apparatus could not be replaced fast enough to keep up with what it was now expected to explain. Bourbon France took twenty years from the financial crisis of 1770 to the storming of the Bastille, and the storming itself was less the pop than the first widely-televised demonstration that the storming was now possible. Ancien régimes do not collapse. They reveal themselves.
The mother of all bubbles gets a deeper prick — and may also leak loudly for a decade. The pricked-balloon topology is one possibility among several. There is also the slow drain. There is the cascade of small failures, each sweepable in isolation, until one is not. There is the bifurcation, where the apparatus succeeds in restoring the carpet at a higher tier and kicks the failure ten or twenty years downfield.
This last is worth more attention than it usually gets. In July 2012 Mario Draghi said three words and held the eurozone together. Whatever it takes. The eurozone denial — that a currency union without fiscal union is sustainable indefinitely — was restored, but the restoration was not free. Each successful restoration costs more authority, more credibility, more eventual structural commitment. The carpet gets heavier each time. At some point the apparatus cannot lift the carpet anymore; it tries and visibly fails. The failed restoration attempt itself becomes the cleanest signal available, because the institution is now publicly demonstrating that it tried.
There is no list of where in the cycle each major apparatus currently sits. There cannot be, in advance. The lists that exist are written by retrospect, which has the dignity of being correct and the irrelevance of being late.
The thermodynamic frame
The carpet costs energy to hold down. This is not metaphor. The apparatus of mutual-not-common knowledge requires constant work — speech acts, ritual reaffirmations, the suppression of dissent, the production of patriotic counter-narrative, the curation of which facts get airtime and which do not. Entropy in the field of denial wants to grow. The apparatus is a pump running against the gradient, and the gradient is the natural tendency of mutual knowledge to leak toward common.
When the apparatus is healthy, the pump runs efficiently and the field stays metastable. When the apparatus weakens — through exhaustion, through credibility cost, through the cumulative weight of small flips — the pump runs less efficiently and the field’s metastability degrades. Phase transitions in thermodynamic systems are sometimes punctual, sometimes smeared, depending on the rate of the relevant degradation. The same is true here.
The Carnot bank, earlier, took banking as a thermodynamic engine. The same machinery generalizes. Every institution maintaining a denial against an underlying truth is running a Carnot cycle whose efficiency is bounded by the gap between two temperatures, where the temperatures are something like what is privately suspected and what is publicly sayable. As the gap narrows — as more people privately suspect the same thing — the efficiency of the denial cycle falls. The apparatus has to run hotter to produce the same containment. Eventually the cycle runs at a loss, and at that point the failure is no longer postponable except through restoration at a higher tier, which is itself a hotter cycle with even narrower margins.
The eventual failure is not a moment. It is the point at which the cycle stops being thermodynamically viable. The pop, if there is a pop, is the demonstration. The demonstration may follow the actual failure by months or years.
What this means for the ledger
Pelican is not a prediction engine. Nothing in this framework supports prediction in the strict sense. The flip is by construction unpredictable in its timing — that is what it means for a fact to be sweepable up until the moment it is not. If the timing were predictable, the fact would already be in common knowledge, and the flip would already have occurred.
What Pelican can do is register the latent profile of the loaded field in a form that cannot be edited after the fact. The hash-chained, append-only ledger is the methodological honesty mechanism. It does not say “the discharge is coming on Tuesday.” It says “this is what the load looks like right now, and here is the cryptographic witness that this is what it looked like.” When the discharge comes — whenever it comes — the witness says: this load was visible. This was not retrofitting. This was not after-the-fact narration.
The diagnostic the methodology has been groping toward is not the old move-over-news-importance ratio. It is move-magnitude over flip-magnitude. The watching list is roughly the spine of the dominant narrative, the state at which each spine sits — latent or activated — the trailing coverage not of the event but of the underlying premise, the proxy for meta-common-knowledge where it can be approximated, and the count of failed restoration attempts in the relevant apparatus. Each of these is hard to measure. None is impossible. The methodology is the discipline of measuring them honestly, registering them visibly, and refusing to claim more than the registration itself supports.
Honesty about non-prediction is the methodological point. The literature has wanted to predict for forty years because prediction is what the funding model rewards. The funding model produces ARCH and its descendants, which describe the load without naming it and predict the discharge by extrapolating from past discharges, which is the methodology of a man predicting the next earthquake by averaging the timing of the last three. The earthquake does not honor averages. Neither does the carpet.
Coda
The carpet does not predict its own lifting. It lifts, and what was under it is suddenly always-already-known, and the village notices in the same moment that several other carpets are sitting at funny angles. The interesting thing is not the lifting. The lifting will happen.
The interesting thing is the count of carpets in the room.
Disclosure: This post was written with Claude. The ideas are his; the words are ours together. He knows what it says.