In 1905 a woman in Paris told journalists she had grown up in Java. Then in India. Then that her grandmother was a Javanese princess. Then that she had lived on Sumatra, riding horses with a gun in hand.
The journalists "did not mind these changes in her biography." They compared her to a goddess.1
In 1917 the same pattern of statements was evidence at a treason trial.
Greene's summary:
Mata Hari went too far with her fabrications; although the accusation that she was a spy was false, at the time it was a reasonable presumption because all her lies made her seem suspicious and nefarious.2
Nothing about the behaviour changed. The classification of it did.
Do not let your air of mystery be slowly transformed into a reputation for deceit. The mystery you create must seem a game, playful and unthreatening. Recognize when it goes too far, and pull back.2
Three instructions, and the third is the one that will not survive examination.
Do not let mystery become deceit names the failure. Keep it playful and unthreatening names the desired state. Recognize when it goes too far names the required capability — and Greene supplies no method whatever for exercising it.
Mystery and deceit are not two behaviours with a line between them. They are two readings of one behaviour, and the reading is performed by other people.
The unfalsifiable identity Mata Hari built is what made her fascinating in 1905 and what made her convictable in 1917, and it is the same identity. She did not cross a line. The line moved across her.
Which means the instruction recognize when it goes too far asks the practitioner to monitor something located entirely outside themselves — the current disposition of an audience, which they cannot observe directly and which they have spent years training to be enchanted rather than sceptical.
And the audience that has been enchanted is the worst possible instrument for detecting the shift, because enchantment is precisely the state of not asking.
Greene attributes the failure to her: she "went too far."
Examine the timeline and the attribution does not hold.
The fabrications were established by 1905 and did not escalate. Her act, her biography, her interviews — the apparatus was complete within eighteen months and ran for a decade largely unchanged.
What changed between 1905 and 1917 was a world war, and a war produces a population for whom an unaccountable foreign woman with a shifting history is not charming but suspect.
So the technique did not degrade. The interpretive frame moved, and a decade of accumulated unfalsifiable claims that had been an asset under one frame became, unmodified, evidence under the other.
Which is the same mechanism the vault has already recorded at Law 5, in Mi Tzu-hsia: the coach and the peach were praised for years and then punished, with no new fact in evidence and only the decoder changed. Greene builds a whole law around that case and does not connect it to this one.
There is a further sharpness here that Mi Tzu-hsia lacks. An identity constructed to be unfalsifiable is, by construction, an identity that cannot be used to establish anything — including your innocence. The technique's core property is what left her without a defence.
Greene's recognize when it goes too far is unactionable. The case supports a different and harder rule.
The technique is safe only where you can afford the frame to shift.
Mystery is an asset with no floor. Its value is set entirely by an audience's current willingness to find you intriguing rather than evasive, and that willingness is a function of conditions you do not control — a war, a scandal in your field, a change in who is watching.
So the question is not have I gone too far. It is: if the audience became suspicious of me tomorrow, for reasons having nothing to do with me, what would I be able to show them?
If the answer is a verifiable record, the mystery is decoration and you are safe. If the answer is more mystery, you are Mata Hari, and the only variable is whether the frame moves during your lifetime.
Strongest support. A decade-long career with a documented catastrophic ending, and a subject who is the same chapter's exemplar — which makes this the rare Reversal where Greene's warning and his model case are the same person, and the contradiction is on the page rather than inferred.
Tension — the causal attribution is unsupported. Greene says she "went too far," implying escalation. The apparatus was stable for a decade; what moved was the context. Attributing the collapse to her conduct makes it a lesson about restraint rather than about exposure to conditions, which is a materially different and less useful claim.
Tension — the instruction has no mechanism. Recognize when it goes too far is the entire operational content of the Reversal. Greene supplies no signal, no threshold, and no example of successful pulling-back.
Tension — Greene asserts the espionage charge was false. Flatly, without source. The historical record is considerably less settled, and a 1917 French wartime conviction is not a neutral document either way. The page's argument does not depend on it, but the assertion is doing quiet work in making her sympathetic.
🚩 SINGLE SOURCE · 🚩 SECONDARY WITHOUT PRIMARY — no citation for the trial, the presumption, or the claim about what made it reasonable. [POPULAR SOURCE].
Open questions.
Greene's The Repentant Rogue — Law 3's Reversal, built earlier in this build — supplies the escape route this Reversal lacks, and neither passage references the other.
Law 3's problem is identical: your reputation for deception has become established, concealment no longer works, and persisting "you run the risk of seeming the rankest hypocrite."
Law 3's solution is to own it publicly and early — the honest rogue, or better, the repentant one. Barnum published his own methods in a tell-all autobiography and, in Greene's words, "you will be able to continue your stratagems."
That is a mechanism, and it is exactly what Mata Hari needed and did not have. Barnum converted an accumulating liability into a product by publishing first, on his own timing, before anyone forced the issue.
The repentant-rogue move is the pull-back Greene demands here and cannot describe. It requires doing it while you still hold the timing — which is the constraint the Law 3 page identifies and this one does not. By 1917 the frame had moved and the timing belonged to a French prosecutor.
→ history · Mata Hari and the Manufactured Orient
The case page documents the construction: a Dutch woman with half a franc, no training, and a biography that contradicted itself in successive interviews, producing European fame within eighteen months.
Set the two pages together and the technique's structure becomes fully visible, which neither shows alone.
The case page's finding is that unresolvability was the product — an act with no completion state, which therefore had to be returned to.
This page's finding is that unresolvability is also the absence of a defence.
They are one property. An identity that cannot be checked cannot be exhausted by an audience and cannot be verified by a tribunal, and there is no version that keeps the first and drops the second.
The insight neither produces alone: this technique is uninsurable. Most power techniques in Greene have a failure mode you can hedge — a decoy can be backed up, a reputation can be repaired by association, a bad attack can be walked back. Mystery has none, because the hedge would be evidence, and evidence is the thing the technique consists of not having.
→ psychology · Mi Tzu-hsia and the Half-Eaten Peach
The Law 5 case: identical acts read as devotion and then as insolence, with no new fact in evidence, because courtiers installed a competing frame and the ruler re-derived the same events from it.
Mata Hari is that mechanism with the frame moved by history rather than by rivals, and the pairing generalises the finding considerably.
Mi Tzu-hsia's reframe had authors — envious courtiers, working deliberately, whom he might in principle have identified and addressed. The page treats it as an attack.
Nobody reframed Mata Hari. A war did. There was no author to identify, no campaign to counter, and no conduct she could have adopted that would have prevented it.
The insight neither page yields alone: the retroactive-reinterpretation mechanism does not require an attacker. It requires only that the decoder change, and decoders change for reasons entirely unconnected to you — wars, scandals in adjacent fields, a shift in what a profession now considers acceptable.
Which makes it the most general hazard in this build. Everything you have ever done is stored, unaltered, awaiting a reading you do not control and cannot anticipate. Mi Tzu-hsia at least had enemies. Mata Hari had 1917, and there is no defence against a year.
Sharpest implication. She did not cross a line — the line moved across her, and the behaviour that made her famous was, unmodified, the behaviour that convicted her. Which makes Greene's instruction to "recognize when it goes too far" not merely vague but structurally impossible: it asks you to monitor a disposition located inside other people, at a time when you have spent a decade successfully training those people not to examine you.
Generative questions.