A German escape artist named Kleppini went around saying he knew how Houdini did it.
His claim: no occult powers, no superhuman body — "He simply used elaborate gadgets."1 He also claimed he had already beaten Houdini in a handcuff challenge in Holland.
The first claim was substantially correct. Houdini could and did conceal small tools; the chapter says so twelve lines later.2 The second was a lie.
Houdini let the speculation run. He minded the lie:
Houdini did not mind all kinds of speculation floating around about his methods, but he would not tolerate an outright lie, and in 1902 he challenged Kleppini to a handcuff duel.1
What follows is not an escape. It is a trap, built over weeks, and it is the only thing in Law 30's chapter where Houdini is not the one being watched.
Kleppini went to work. Through a spy, he obtained the secret word that unlocked a pair of French combination-lock cuffs Houdini liked to use.
The plan was clean: get offered a choice of cuffs onstage, pick the French ones, open them, and demonstrate in front of an audience that the great Houdini's genius was a mechanism anybody could operate with the right word.3
On the night, it went exactly as he had planned. Houdini offered a choice. Kleppini chose the combination cuffs. He was even allowed to slip behind a screen with them for a quick test, and came back out seconds later, confident.4
He had opened them. Behind the screen, with the word C-L-E-F-S — French for keys — the cuffs had opened.
Then Houdini refused to lock him in.
Acting as if he sensed fraud, Houdini refused to lock Kleppini in the cuffs. The two men argued and began to fight, even wrestling with each other onstage. After a few minutes of this, an apparently angry, frustrated Houdini gave up and locked Kleppini in the cuffs.5
Kleppini strained. The word did not work. He had opened these same cuffs minutes earlier and now they would not move.
He sweated, racking his brains. Hours went by, the audience left, and finally an exhausted and humiliated Kleppini gave up and asked to be released.5
The cuffs he had opened with C-L-E-F-S now opened only with F-R-A-U-D.6
He never worked out how it had been done. Greene supplies it: during the tussle, Houdini changed the code — a move he had spent weeks practising.7
Run the whole thing from Houdini's side and notice that Kleppini is never once obstructed.
The spy succeeds. "He let his opponent learn the code."7 The intelligence Kleppini thought he had stolen was placed where he would find it.
The choice is offered. Houdini presents a range of cuffs, and Kleppini selects the ones he has been prepared to select. This is Law 31 in its purest form — a menu built so that the branch the target will take is the branch you want.
The test behind the screen is permitted. This is the masterstroke, and it is the step a cruder operator would have refused. Letting Kleppini verify converts a suspicion into a certainty. He walks back out having personally confirmed the plan works, which is a far stronger position to destroy than mere confidence.
The refusal to lock him in. Houdini plays reluctance, which does two things at once: it signals to the audience that Houdini fears the outcome — raising the stakes and Kleppini's credibility — and it manufactures the pretext for physical contact.
The scuffle. Physical proximity, chaotic, unwatchable, apparently emotional. Cover for a few seconds of hand work.
The capitulation. Houdini "gives up" and locks him in, so that the fatal act is performed by a Houdini who appears to have lost the argument.
Six steps, and in every single one Houdini gives Kleppini what Kleppini wants. The entire operation runs on cooperation. There is no point at which the victim is resisted, which is why there is no point at which he becomes suspicious.
The screen test deserves its own look, because it inverts the ordinary logic of deception.
Most deception works by preventing verification. This one supplied it.
Kleppini's danger, from Houdini's perspective, was that he might hedge — pick a different pair of cuffs, or hold something back, or smell a rat and withdraw. The screen test eliminated all of that. Having opened the cuffs himself, in private, moments before, Kleppini had no residual doubt to act on. He was not merely confident. He had evidence, and it was true evidence, and he had gathered it himself.
That is the sharpest thing in this case and Greene does not name it: a verification that is genuine, self-administered, and correct at the time it is performed can be the most effective possible component of a trap, provided the thing verified is still mutable afterwards. The lie is not in the evidence. It is in the assumption that the world holds still between checking and acting.
Hours later, in front of an emptying theatre, Kleppini was still working with a fact that had been true.
The chapter files this under Make Your Accomplishments Seem Effortless, and it does not belong there.
Nothing in the Kleppini duel is about concealing labour to appear graceful. Houdini's weeks of practice at the code-change are mentioned in one clause and are the least interesting part. What the episode demonstrates is controlled options (Law 31), letting an enemy destroy himself with information you planted (Law 3's territory, and Law 42's), and a staged emotional display used as cover.
The visible-effort question does not arise. Kleppini is not discounting Houdini's achievement because he saw the work; he is disputing that there is an achievement.
So why is it here? Because Houdini is here. The chapter is organised around a figure, and the second case arrived attached to him — which is a version of a defect the build has found repeatedly in this corpus: cases get filed by protagonist rather than by mechanism, and the law they are printed under is a fact about the book's construction rather than about the case.
Kleppini was right about the gadgets. He is publicly humiliated for being right.
Greene's line — "Kleppini was fundamentally wrong: It was not Houdini's tools but his practice, work, and research that made his escapes possible"2 — is a rescue, and it is doing something worth watching. It converts a factual dispute (are there tools?) into an attributional one (what is the tools' relative contribution?), on which Kleppini can be declared wrong without the factual claim ever being addressed.
And the demonstration that settled it in the room settled nothing about the question. The duel proved that Houdini could beat Kleppini at a rigged contest. It did not touch whether he used gadgets.
A public defeat was accepted as a refutation of a claim the contest was not designed to test. That is worth carrying, because it is the general form of what the chapter recommends: the audience walked out convinced, and what convinced them was a change of subject conducted at high volume.
Somebody is saying a true thing about you that you would rather they did not say.
The instinct is to deny it. Don't — the denial is where you get caught, and the thing is true.
Look instead for the part of their position that is not true. Kleppini's gadget claim was accurate; his Holland claim was invented. Houdini went after the invention and let the accurate part stand, unaddressed, forever.
Then, if there is to be a contest, do not defend. Build the terms. The person attacking you wants something specific — to be proved right in public. Give them a route to it. Let them gather their evidence, let them verify it, do not obstruct a single step, and make sure the thing they verified is something you can still change.
And when the moment comes, appear to lose the argument first. The trap should be sprung by a version of you that has just been overruled.
One warning, because this is a technique with a shape: everything here works because Houdini was materially more capable than his opponent. Strip that out and the same six steps produce a man who elaborately arranged his own exposure.
Strongest evidence. The two codes — C-L-E-F-S and F-R-A-U-D — are specific, quoted, and mechanically coherent with a combination lock. The 1902 date and the Holland claim give the account edges.
Tension — Greene denies and confirms the tools within twelve lines. "It was not Houdini's tools"2 against "no one could check the inside of his throat, where he could have concealed small tools."2
Tension — the case is filed under the wrong law. Documented above.
Tension — F-R-A-U-D is a six-letter word chosen for the story. Worth flagging plainly: the detail that makes this anecdote memorable is also the detail most likely to have been improved in retelling. It requires that Houdini, mid-wrestle, set a specific six-character combination — and that the word he happened to have prepared was the perfect one. Possible. Also exactly what an anecdote collection produces.
🚩 [POPULAR SOURCE] · 🚩 SINGLE SOURCE · 🚩 SECONDARY WITHOUT PRIMARY — no source, no venue, no press account. ⚠ Date inconsistency in the source itself: the Birmingham challenge is dated 1904 and the Kleppini duel 1902, but the Kleppini passage introduces the rivalry as a response to Houdini's established fame, and the chapter narrates them in the reverse order.
Open questions. What did Kleppini do afterwards, and is there any account of the episode that is not Houdini's? Every element of this story — the spy, the plant, the code change — is knowable only from the man who ran it.
The two Houdini pages in this chapter want opposite things from him, and setting them side by side is more useful than either alone.
Houdini and the Birmingham Manacles is Houdini as subject — watched, tested, concealing preparation to produce grace. This page is Houdini as operator — unwatched, running a scheme on a person. The first is a performance discipline; the second is a con. Greene admires both in the same register and does not mark the transition.
That matters for reading the rest of the corpus, because the same slide happens elsewhere. Ivan IV appears as a boy managing his own visibility at Law 28 and as a czar engineering a nation's options at Law 31. The book's recurrence structure keeps a figure constant while the kind of thing they are doing changes underneath, and the law number is where it lands, not what it is.
Against Law 3, this belongs there. Kleppini is destroyed by information he believes he stole — the same shape as the Sam Geezil con Greene explicitly cross-references at Law 31. A planted secret is more persuasive than a stated one, because the target authenticates it themselves in the act of acquiring it.
Psychology — the self-gathered evidence that cannot be dislodged. The Question You Actually Answered describes the mind quietly swapping a hard question for an easy one and returning the answer with undiminished confidence, the swap itself never surfacing.
Kleppini's swap: the hard question is will these cuffs open when it counts; the easy one is did these cuffs just open. He answered the easy one, correctly, behind a screen, and carried the answer onstage as though it settled the hard one.
The insight neither frame gives alone: the substitution literature is built on cases where the easy question is a noisy proxy for the hard one. Here it is a temporally valid proxy — the answer was completely correct at the moment he obtained it, and the only defect was that time passed. That is a distinct and nastier failure mode, because none of the usual defences reach it. Scepticism does not help; he was sceptical enough to test. More evidence does not help; he had first-hand evidence. The only defence is asking whether the verified fact is the kind of thing that can change between the check and the act, and almost nobody asks that, because a fact you established yourself does not feel like a claim with a shelf life.
Security practice — the moment of check versus the moment of use. Signal Mimicry: The Arms Race Between Scam and Legitimate Branding tracks how verification signals get counterfeited, and how each defensive signal invites the imitation that degrades it.
The Kleppini trap sits in a different place on that map and shows the map is incomplete. Houdini counterfeited no signal. The cuffs were real, the code was real, the test was real. What he exploited was the interval — the gap between verification and reliance, which no amount of signal integrity closes.
What the pairing produces: an arms race over signal authenticity is fighting the previous war. A defender who achieves perfect, unforgeable verification has still not addressed a state that is genuinely verified and then genuinely altered, and the attack requires no forgery at all — only access between the two moments. It is the reason a screen test is worse than useless when the opponent controls the object in between, and the reason the strongest confidence in the theatre that night belonged to the man with the best evidence.
Sharpest implication. A verification that is real, self-administered, and correct when performed is the most powerful component available to a trap, provided the verified thing is still mutable afterwards. Kleppini did not fail to check. He checked, he was right, and he lost — because the only question he did not ask was whether the world would hold still between checking and acting. Every step of his defeat was something he was allowed to do.
Generative questions.