Picture a furnace built to run on a specific fuel, and nothing else. Feed it that fuel and it roars — it heats a room, then a house, then a city block. Cut the fuel and it doesn't just cool down the way an ordinary fire would. It runs its own draft backward, sucks in whatever's nearest, and starts burning the furnace itself. That's the shape of what happened to Rasputin between November 1915 and December 1916, and the vault already has a page that names the mechanism in exactly those terms.1 The question this report asks is whether that's a description of one unlucky Siberian peasant, or a law.
A companion report already sits in this vault on a related question — whether Rasputin was actively managing his own myth in real time, the way Shivaji's chancery was managing a coinage-and-coronation legend at the same historical moment.2 That report found Rasputin genuinely was performing, live, for an audience — the 1912 newspaper interview and the Djanumova visits show a man running a repeatable kit of moves, deflating and inflating in the same breath, aiming every line at the one listener who mattered.3 That's the craft. This report is about what happens once the craft stops working — not how Rasputin built the myth, but what specifically breaks it, and whether the break is a slow leak or a single failed valve.
Strip the vault's model down to the sentence it's actually making. Charisma, on this account, isn't a personal quality a figure carries around like height or a good voice. It's a transaction that has to clear, over and over, in public. A follower projects competence onto the figure — he stopped the bleeding, he moved the minister — and the figure absorbs that projection and acts with the borrowed confidence of everyone who currently believes in him. As long as the world keeps handing back even ambiguous wins the figure can plausibly claim, the loop reinvests its own output and grows. The whole system has exactly one point of failure: the moment reality stops producing anything the figure can credibly claim credit for.1
That's a narrower and more falsifiable claim than "people stopped believing in him." It says something specific about why they stopped: not moral disgust, not a competing story winning out, but a raw supply problem. Rasputin's fuel was legible efficacy — not actual cause and effect, but results an audience would agree to credit to him. For a decade the tsarevich's hemophilia supplied that fuel almost perfectly: a recurring, terrifying, and — crucially — self-limiting crisis, so that when the bleeding stopped on its own schedule, Rasputin could stand at the exit and collect the credit.1 The petitioner economy supplied a second, steadier stream: favors granted, lawsuits dropped, a telephone call answered. Small, constant, replenishable proofs that he bent the world.
Run the tape forward and the break isn't diffuse. It has a date range you can point to, and a shape that repeats across every witness who was in the room. In November 1915 the loop is still humming at full power: the tsarevich is "almost moribund," Rasputin arrives, says "Don't be alarmed. Nothing will happen," and walks out — the boy sleeps.4 Weeks later, mid-orgy, he's handed a telephone for an earache and simply commands the pain away; fifteen minutes later it's gone, and asked how, he says "I told him to."4 That's a man who has been confirmed so many times he can no longer find the seam between his own word and a result the world happened to produce on its own.
By 1916, war did what nothing else could: it shrank the category of things a man could visibly fix down to nothing, against a front that wouldn't hold and bread that wouldn't arrive.1 The clearest single moment in the whole record is an old woman with arthritic hands, come for the laying-on that used to be automatic. Rasputin looks at her and says, "I am sorry, babushka, I cannot do it. The Lord has taken my power from me."1 That's not modesty. It's a man checking an instrument he's trusted his whole life and finding the needle dead — and saying so to someone who has no power to punish him for the admission, which is exactly why the sentence can be trusted. Within months the collapse has a full clinical signature: the sacred and the appetite, which used to stay in their separate lanes, start fusing into fever-babble — "God, prayer, wine," repeated, unsortable — and by autumn he's telling a journalist that if the nobles burn him, Russia burns with him.1 Watch the shape of that last line carefully. He hasn't gone quiet. He's gone from claiming he can save the empire to claiming he's load-bearing in its fall. Same scale of self-importance, same intensity — the sign just flipped.
A mechanism this clean deserves a hard test, and the vault has two genuinely different collapse cases on file to test it against — neither one built with Rasputin in mind.
The meritocracy-decay report examined four historical systems where capable rule failed to reproduce itself, and found four different mechanisms doing the killing: the Mongols needed terror to make an advancement ceiling credible, and terror couldn't survive Genghis Khan's death because it depended on his personal, demonstrated willingness to follow through; the Kakatiya court died to a purely external military mismatch that had nothing to do with how it selected administrators; the Yuan dynasty's openness was an imported input that decayed as the wider Mongol world fragmented; the Mauryans produced their best three generations by violent contest, not by the elaborate selection doctrine built to prevent exactly that.5 None of those four is, on its face, an efficacy-collapse story — they're about succession, not about a single man's confirmable wins running dry.
But look closer at the Mongol case specifically and the resemblance sharpens into something real. Khan's terror-backed authority worked the same way Rasputin's healing worked: it required continuous, personal, visible follow-through to stay credible, and it could not be inherited, because Ögedei could run the machinery without understanding why it worked, and the machinery's psychological credibility depended on a demonstrated willingness only Khan himself had ever shown.6 That's the identical structural claim the Rasputin page makes about confirmation-dependent charisma: a system built on a specific person's continuously renewed demonstration is not transferable, and it does not degrade gracefully — it holds until the demonstrations stop, then it breaks all at once. The difference is what breaks in the gap. Khan's authority needed to be tested by officers probing whether the threat was still real; Rasputin's needed to be tested by a war that made healing impossible to fake. Two different tests, one identical failure architecture: a personal, non-transferable, continuously-renewed demonstration, broken the moment the demonstrations stop and nothing forces a sober recalibration in between.
Now check the mechanism against the opposite kind of case — the Shivaji legend the culture-warrior report traces, and the destiny-script Jijabai installed in the boy who became that legend.7 Neither of those constructs runs on efficacy at all. Purandare's legend survives centuries specifically because it never has to clear a real-time test: a killing gets overlaid with Narasimha executing a demon, and the overlay doesn't need Shivaji to keep performing miracles, because it's a retrospective story about a dead man, immune to a war arriving next Tuesday and refusing to cooperate.8 The vault's own psychology material names this directly, from the opposite tradition: a holy cause functions best when it is unfalsifiable, because a doctrine that can be checked against reality is a doctrine that can be falsified, and falsified doctrine generates the doubt that dissolves conviction.9 Rasputin's charisma was the fragile, high-maintenance kind precisely because it staked itself on the one thing a legend never has to risk — a specific, checkable, real-time result. That's the sharpest thing the comparison produces: it isn't that charisma in general is fragile. It's that testable charisma is fragile in a way that narrative charisma structurally is not, and the two are not the same phenomenon wearing different clothes — they're built on opposite design principles, one betting everything on a falsifiable claim renewed daily, the other betting everything on never having to make one.
Here's where the model has to earn its keep rather than just look elegant, because there are two documented facts about the actual Rasputin collapse that don't sit comfortably inside it.
First: the social register of his charisma didn't die on the same schedule as the medical one. At the Villa Rhode, officers drew swords on him and he fixed them with a stare — "You've seen that my power won. You don't have any power anymore" — and they went home, in the same general period the medical loop was already failing.1 Moynahan's own gloss on the scene notes the power here "was not magical" and did not, ultimately, save him from a beating or the knife that killed him.1 So intimidation-charisma and healing-charisma are running on different clocks inside the same man, and a model built entirely around confirmable-wins-in-healing doesn't automatically explain why the intimidation register held on longer. The likeliest answer — that pure social nerve is self-confirming in a way medicine isn't, because the audience's own fear supplies the "win" without any external test — is plausible, but it's an extension the page itself doesn't fully build, and it should be flagged as an extension rather than smuggled in as already-proven.
Second, and sharper: if the efficacy loop had genuinely collapsed by late 1916 — if Rasputin really was, by his own admission, a spent instrument — then Yusupov and his fellow conspirators had no rational reason to consider him dangerous enough to murder. Yet they killed him anyway, in a plot built on the premise that his continued influence over the throne was an active, urgent threat to Russia.1 Something kept functioning that the collapse model, read narrowly, says should have already burned out. The likeliest resolution is that Rasputin was never running one charisma loop — he was running at least two, and they depended on different fuel entirely. The healer loop needed confirmable medical wins and it died exactly when the page says it died. The court-broker loop — his hold over Alexandra specifically, and through her, over ministerial appointments and the front — never depended on confirmable wins at all, because her need for a holy intercessor was already fully formed before hemophilia, before Rasputin, even before the earlier faith-healer Philippe.10 A need that precedes its object doesn't require the object to keep performing; it only requires the object to remain occupying the slot. That's why the political-broker charisma survived the healer charisma's collapse intact enough to get him killed for it — the two were never the same transaction, and the vault's own efficacy-collapse model, taken alone, quietly assumes they were.
If the mechanism is real for the testable variety of charisma, it should cash out as something you could watch for in real time, not just diagnose afterward from a biography. Four questions do the job, each one lifted directly off a documented moment in the collapse.
Is the win getting harder to fake? Track whether the figure's claimed victories are drifting from ambiguous-and-recoverable (a nosebleed that was always going to stop) toward unambiguous-and-irreversible (a front line, a famine). The drift itself is the leading indicator, well before any confession follows it.
Is there a private confession to someone with nothing to gain from it? The babushka test: watch for the moment the figure admits a limit to an audience that cannot punish or profit from the admission. A confession delivered to someone with skin in the game is still performance. A confession delivered to someone who simply wanted a cure and didn't get one is the instrument reading itself.
Are the categories that used to stay separate starting to blur? Rasputin's sacred register and his appetite register ran in parallel for years — heal the heir, then carouse, two distinct activities — and fused into unsortable fever-babble only once confirmation stopped sorting them.1 A figure who can no longer tell his discipline from his hunger has already lost the internal audit that let the performance look effortless.
Has the scale of self-importance flipped sign without shrinking? Grandiosity doesn't fade into modesty when a testable-efficacy loop breaks — it inverts, preserving its own scale. Watch for the exact moment a figure stops claiming what they can save and starts claiming what their fall will destroy. That inversion, not any external event, is the diagnostic that the loop has already finished breaking.
Put the whole comparison together and the honest verdict is narrower than "charisma is a countdown," but sharper for being narrower. The mechanism the vault names for Rasputin — a feedback loop that requires continuously renewed, checkable, personal proof — is real, and it explains the Mongol succession collapse with almost no modification, because both are cases of authority staked on a specific person's non-transferable, repeatedly-demonstrated follow-through. It does not explain the Shivaji legend or the doctrine Hoffer describes, because those systems are built, on purpose or not, to never make a testable claim in the first place — and a claim that's never tested can't fail a test. The generalization isn't "all charisma breaks the same way." It's that charisma splits cleanly into two architectures with opposite failure modes: the kind staked on a renewable, checkable win, which breaks suddenly and completely the instant the checks stop passing, and the kind staked on an unfalsifiable story, which doesn't break from efficacy failure at all because it never risked anything a fact could touch. And even inside one man, in one court, in one specific year, those two architectures can be running simultaneously on separate clocks — which is exactly why the same collapsed miracle-worker still had to be murdered.
Threads: → What Breaks a Charismatic Hold on Power — Thread Map