A man is standing in a hotel suite with a certified cheque for 250,000 francs in his hand, and he has just started to think.
Monsieur P. had won the bidding to buy the Eiffel Tower for scrap. He had been told to bring the cheque to the Hotel Crillon within two days. And now, in the room, listening to the deputy director general discuss arrangements, the doubts arrive.
"Why meet in a hotel instead of a government building? Why hadn't he heard from other officials? Was this a hoax, a scam? As he listened to Lustig discuss the arrangements for the scrapping of the tower, he hesitated, and contemplated backing out."1
Three questions. All of them correct. All of them, at that moment, unanswered.
This is the hinge of the most famous con in the book, and what happens next is the single most instructive thing in Law 28.
It is worth pausing on the quality of Monsieur P.'s reasoning, because he is not a fool and the chapter treats him as one.
Why a hotel and not a ministry? A genuine anomaly. Government business happens in government buildings.
Why no other officials? A structural point, and the sharpest of the three. A transaction of this size and irreversibility would involve a department, not a man.
Is this a hoax? He has arrived at the correct hypothesis, unprompted, holding the cheque, before handing it over.
He is doing everything a careful person is supposed to do. He has assembled a threat model — what would a fraud look like? — and every item in it is satisfied.
And it does not save him, because the thing that arrives next is not on the list.
"Suddenly, however, he realized that the director had changed his tone. Instead of talking about the tower, he was complaining about his low salary, about his wife's desire for a fur coat, about how galling it was to work hard and be unappreciated. It dawned on Monsieur P. that this high government official was asking for a bribe."2
Nothing is stated. Lustig makes no request. He complains — about salary, about a wife's wants, about being unappreciated — and lets the mark assemble the conclusion himself.
That is the same technique the vault has logged across this whole build: the operator supplies a shape and the target supplies the content, so that the belief the target ends up holding is one he wrote. Borri hinted at his accomplishments. Greene's step one instructs vagueness so the listener "will make their own connections."
Here it is deployed at the point of maximum suspicion, and it works precisely because the mark has to do the interpretive work. A man who has just deduced that he is being asked for a bribe is not, in that moment, entertaining the possibility that he is being conned. He is congratulating himself on having read the room.
Greene's sentence is exact and worth quoting whole.
"The effect on him, though, was not outrage but relief. Now he was sure that Lustig was for real, since in all of his previous encounters with French bureaucrats, they had inevitably asked for a little greasing of the palm. His confidence restored, Monsieur P. slipped the director several thousand francs in bills, then handed him the certified check."2
Read the reasoning he actually performs. Not this man is corrupt, therefore he is dangerous. Rather: real officials ask for bribes; this man asked for a bribe; therefore this man is a real official.
That is a base-rate inference and it is sound. If the population of French bureaucrats Monsieur P. had dealt with inevitably wanted greasing, then bribe-asking is a genuine feature of the category, and observing it is genuine evidence of membership.
The inference is valid. The premise is true. And the conclusion is false, because the trait he tested for is one an impostor can produce at zero cost.
The tell was chosen precisely because it was cheap to fake and expensive to fake wrongly — a fraud who had never dealt with the French bureaucracy would not know to ask.
Monsieur P.'s threat model has a shape, and it is the ordinary one: a fraud is a person trying to get money out of me.
Every item on his list follows from that. Wrong venue — a fraud avoids official buildings. No other officials — a fraud works alone. Is this a hoax — the general form.
Now put the bribe against that model. A man trying to get money out of me has just asked for more money, at the last moment, creating a new obstacle between himself and the cheque he has spent a fortnight securing.
That is the behaviour of someone with nothing to lose by delay — which is to say, someone whose position is real.
So the bribe does not defeat the threat model. It falls outside it, and arrives as evidence for the very proposition being doubted. Greene's general statement of the principle is the best sentence in the law: "If you sense that the sucker has suspicions, do as the intrepid Lustig did: Instead of backing down, or lowering his price, he simply raised his price higher. Asking for more puts the other person on the defensive, cuts out the nibbling effect of compromise and doubt."3
And note the second effect, which is quieter and does more damage. Having paid the bribe, Monsieur P. is now complicit. He has corrupted a public official, as he understands it. That makes the transaction something he has an interest in not examining — and it is a substantial part of why he never went to the police.
Someone you are dealing with is about to pull out, and you can feel it.
The emails have got shorter. There is a new person copied in. The question that arrives is not about the work but about the arrangement — remind me how the invoicing works, who else has done this with you.
Your instinct is to reassure, and reassurance takes a specific and recognisable form: you make yourself easier. A discount, a flexible term, a faster start, an offer to reduce the commitment. Every one of those feels generous and every one of them confirms the hypothesis they are running, which is this person needs this deal more than I do.
The counter-move here is genuinely useful and it does not require anything dishonest. Add a condition instead of removing one.
Ask for something a person in a weak position could not ask for. A longer notice period. A named contact on their side. Payment terms that are worse for them. A pause while you check something.
Watch what that does to the doubt. The hypothesis this is precarious does not predict a counterparty who makes the deal harder, and the prediction failing is worth more than any amount of reassurance.
And here is the line, because Lustig is on the other side of it. The condition has to be one you would actually enforce. If you add a term purely for the signal and would drop it the moment they pushed, you have manufactured a tell rather than revealed one — which is exactly what the fur coat was. The honest version of this move costs you something real; the dishonest version costs nothing, which is why it is available to a man with no ministry behind him.
Greene tells this episode in two places: here at Law 28, and at Law 12's Reversal, Use Selective Honesty and Generosity to Disarm Your Victim.
The build-once rule assigns it here, and the two placements are worth comparing because they read the same event differently.
At Law 12 the bribe belongs to a family of moves in which a small piece of genuine-seeming candour buys a large amount of trust — the operator reveals a flaw in order to be believed about everything else.
At Law 28 it is filed as boldness: raise your price when doubted.
Both readings are available in the event, and the vault holds it here because Law 28 is where Greene states the general principle. But the Law 12 framing catches something the boldness framing misses. What Lustig discloses is a vice, and disclosing a vice is a specific kind of move: it makes the discloser legible, which is the property the mark had found missing. Three doubts, all of the form this man does not fit the category, answered by a single unattractive detail that puts him inside it.
The vault's Law 12 treatment sits at Selective Honesty as a Control Mechanism — itself one of the superseded pages rebuilt in this ingest, and the page where the corpus's first fabricated law number was found.
Greene reports the sum as "several thousand francs in bills."2
Set against a certified cheque for more than 250,000, that is roughly a rounding error — and it is the entire consideration Lustig paid for the mark's confidence.
Which is the economic shape of the move and the reason it generalises. The bribe is not revenue; it is a purchase. Lustig buys the collapse of three correct doubts for something under two per cent of the take, at the exact moment when the alternative was losing all of it.
And the price is set by the category, not by the deal. A believable bribe from a middling official has to be small — a large one would itself be suspicious. So the cheaper the tell, the better it works, which inverts the usual relationship between the cost of a signal and its credibility.
That is worth marking against the vault's signalling material. Costly signals are credible because a defector could not afford them. This is a cheap signal that is credible because it is embarrassing — and embarrassment, unlike expense, is not something a fraud is normally willing to manufacture.
🚩 SINGLE SOURCE · 🚩 SECONDARY WITHOUT PRIMARY. Uncited throughout. Monsieur P.'s interior sequence — the three doubts, the dawning realisation, the relief — is reported as known mental content of an unnamed man in 1925.
🚩 The claim about French bureaucrats is the load-bearing premise and is presented as ordinary knowledge: "in all of his previous encounters with French bureaucrats, they had inevitably asked for a little greasing of the palm."2 The mechanism requires it to be true, and nothing supports it beyond the assertion.
Tension with the chapter's frame. Law 28 files this under boldness. The move is not bold in any sense the six mechanisms describe — nobody is impressed, no fear is struck, nothing is made to seem larger. It is a precisely calibrated small gesture, and calling it audacity obscures what it does.
Open question. The bribe works because it is a trait the mark had verified across a whole category. Which raises the general form: every category has such a trait, and every one of them is a fake-able credential. Is there any category marker that is expensive enough to be reliable and still commonly checked?
Greene's Law 3 — Conceal Your Intentions — and this case pull in opposite directions and the tension is instructive.
Law 3's operator hides what he wants. Lustig, at the decisive moment, reveals a want — a petty, discreditable, entirely believable one — and the revelation is what closes the deal.
What the pair produces is a distinction the book does not draw. Concealing your intention makes you unreadable; revealing a small dishonourable one makes you readable in the wrong direction. The second is stronger against a suspicious counterparty, because suspicion is generated by illegibility, and an unattractive motive is the cheapest legibility available.
There is also a convergence with the vault's Law 27 material. Greene's step one instructs vagueness so that followers "make their own connections and see what they want to see." Lustig's complaint about a fur coat is that instruction executed in a single conversation: he never asks for anything, and the mark supplies the request, the interpretation and the conclusion.
Psychology — The Question You Actually Answered
Standing in front of a used car with no knowledge of engines, you cannot answer is this a good car. So the mind quietly answers a different question — do I trust the person selling it — and hands you the substitute with the confidence the original would have earned, without your ever noticing the swap.
Monsieur P. is running that substitution and it is worth being precise about which questions are in play.
The question he cannot answer is does this man hold the office he claims. Verifying that requires a call to a ministry, a directory, a colleague — an hour of work he does not do, standing in a hotel suite with a cheque in his hand.
The question he can answer is does this man behave like the officials I have known. That one is cheap, immediate, and resolvable from the evidence in the room.
And the bribe is aimed exactly at the second. It contains no information about the first — an impostor can complain about a fur coat as easily as a director general can — but it settles the substitute completely.
What the pair adds to the substitution account is the operator's side. The usual framing is that substitution is a failure the subject falls into under difficulty. Lustig's case shows it being aimed at: he does not merely benefit from the swap, he identifies which question the mark is able to answer and supplies decisive evidence for that one. The manipulation is in the selection of the answerable question, not in any false statement.
Which names the defence, and it is not scepticism. It is noticing that you have answered something easy. I have concluded he is genuine — what exactly did I observe, and does it bear on the thing I was worried about? Monsieur P.'s three doubts were about office; the evidence he accepted was about manner.
Eastern spirituality — The Asura Boon-Loophole
The demon has earned a wish and is no fool. He writes himself a contract with no gaps: not by man, not by beast, not by day, not by night, not indoors, not outdoors, not with a weapon, not with bare hands. He closes every door he can think of and walks away feeling immortal. Then something that is neither man nor beast kills him at dusk on a threshold with its claws — and every clause remains true.
Monsieur P. built the same object in a hotel suite.
His three doubts are an enumeration: wrong venue, no other officials, is this a hoax. It is a genuine threat model, honestly constructed, and it covers the ways a fraud presents itself to a careful man.
The bribe is not an attack on any clause. It arrives from the region the enumeration does not cover — and, like the demon's killer, it satisfies every stated condition while producing the opposite of the intended result. Nothing in a fraud avoids official buildings or a fraud works alone has anything to say about a fraud who asks for more money.
The shared principle, which neither text states: a defence is written against an enumerated threat, and the enumeration is always shorter than the world. Neither the demon nor the mark is out-argued. Both are approached from a direction their list had no term for, and in both cases the list was good — that is what makes the pairing worth having, because the failure is not carelessness.
And it names the operator's actual skill. Not deception, and not audacity. Reading somebody's threat model well enough to find its complement — which is a much larger space than the space of anticipated attacks, and which grows every time a clause is added.
Sharpest implication. The bribe is a falsifiable-looking test that is cheap to pass. Monsieur P. verified a real property of a real category, using sound reasoning from a true premise, and arrived at a false conclusion — because the trait he chose as his credential was one an impostor could produce for a few thousand francs. Which means the danger is not credulity or laziness. It is that every category has a familiar marker, every familiar marker is a cheap signal, and checking one feels exactly like due diligence.
Generative questions