Consider that a monetary theory can fail in a way its holder cannot detect.
On October 5, 2026, Li Lu โ the investor Charlie Munger publicly called his best external hire, the operator behind Himalaya Capital and the BYD position โ stood at the 36th Graham & Dodd Annual Luncheon at Columbia Business School and called cryptocurrency "fundamentally bogus," predicting its eventual disappearance. The claim reached me the way most claims reach me now: a Wu Blockchain post on X dated October 21, quoting a talk for which I have found no transcript, no video, no timestamped recording. A paraphrase of a paraphrase of a speech, relayed by a crypto-native account, reported by a crypto-native outlet, containing zero protocol architecture, zero token mechanics, zero code, zero on-chain data, and exactly one quantitative figure about spot ETF inflows โ with no source attached to it.
I am not raising that as a gotcha. I am raising it because provenance is the job. The first move in any audit is never to argue with the contract. It is to establish what you actually have in front of you. What I had was a low-fidelity transmission of a high-confidence opinion.
And buried under the parts that cannot be tested, there is one proposition that can be. That is the only part worth the work.
The tool determines what the eye can see.
Li Lu's intellectual lineage runs Graham to Dodd to Buffett to Munger to him, and that lineage is not a vibe. It is a specific valuation instrument: discounted cash flow, owner earnings, margin of safety, circle of competence. Every one of those primitives requires a cash flow to discount or a book value to mark against. An asset with no issuer, no receivable, no coupon, and no terminal liquidation value does not score badly on that instrument โ it fails to enter the instrument at all. The output is not "overvalued." The output is undefined behavior.
This matters because it reframes what "fundamentally bogus" means. It is not an empirical finding about Bitcoin's properties. It is a report about the boundary conditions of a tool. When your only measuring device is a scale, everything that is not mass appears weightless. The statement is true about the scale. It is silent about the object.
So I want to separate two claims that the coverage collapsed into one. The first is a prediction: crypto will disappear. Predictions are falsifiable, and this one is currently being tested in public. The second is a normative wish โ that it would be better if crypto were banned outright. That is a value judgment, not falsifiable, and it should be quarantined from the first. The reporting treated them as a single utterance. They are not the same epistemic object, and the difference is where the entire analysis lives.
Between those two sits the actual argument. And the argument has a nameable structure.
The enforcement chain, premise by premise.
Strip the rhetoric and the syllogism is this: money requires an enforcement mechanism; that mechanism can only be supplied by a state or a society; cryptocurrency was launched by a small number of private individuals; therefore cryptocurrency cannot be money. Four steps. The first is uncontroversial. The second is where the structure fails, and it fails in a way that is invisible if you read quickly.
"Enforcement mechanism" is doing two incompatible jobs in that sentence.
Legal enforcement is permission-based. It works because the state says so, continuously, and it stops working the moment the state stops saying so. Tax obligation, legal tender status, monopoly on legitimate violence. It is a standing political fact, renewed daily.
Economic enforcement is prohibition-based. It works because violating the rule costs more than obeying it. Bitcoin's version is thermodynamic: to rewrite a settled block you must re-accumulate the proof-of-work that secures it, and that cost scales with the depth of the rewrite and the hashrate of the network. There is no permission required and no authority to petition. There is only a cost function that gets steeper the further back you try to reach.
These are not two flavors of the same thing. They are opposites on the axis that matters. One is revocable by a signature. The other is revocable only by spending joules.
Once you split the term, the conclusion collapses. "No private actor can supply an enforcement mechanism" is false โ a private actor supplied a thermodynamic one, and it has held for seventeen years across multiple jurisdictional shocks, two major exchange collapses, and one civil war over block size. The conclusion survives only if you define enforcement as state enforcement. And if you do that, the syllogism becomes: money requires state enforcement; crypto has no state enforcement; therefore crypto is not money. That is a tautology. It contains no information. It cannot be wrong, which means it cannot be right either.
I have made this mistake myself in the other direction. In 2017 I spent one hundred and twenty hours inside the Uniswap V1 core contracts during the ICO froth, and I found an integer overflow in the price calculation path that could have drained pools. I submitted it to GitHub before mainnet. The lesson I took was not "code is safe." The lesson was that I had been auditing arithmetic while assuming the trust model. The overflow was a bug. The trust model was a choice. Li Lu is making the mirror error: he is evaluating a trust model while assuming the arithmetic doesn't count.
The premise nobody checks: is state enforcement durable?
The argument's hidden load-bearing assumption is that legal enforcement is permanent. That assumption is falsifiable, and it has been falsified repeatedly within living memory.
Executive Order 6102, 1933. The United States compelled private holders of gold coin, bullion, and certificates to surrender them at a fixed price. Not a tax. Not a disclosure requirement. Compulsory conversion of a private store of value into state paper, enforced under penalty of law.
Cyprus, 2013. Depositor balances above the insured threshold were haircut to recapitalize banks. The enforcement mechanism that guaranteed deposits was suspended by the same authority that guaranteed it.
Canada, 2022. Emergency powers were invoked to freeze accounts associated with a protest movement, without a court order per account. The mechanism worked exactly as designed. That is the point. It worked.
In each case, the legal enforcement mechanism did not fail. It was exercised. What failed was the implicit assumption that the terms of holding would not change. That is the dimension Bitcoin was engineered against โ not inflation, which is the marketing story, but revision. The property being optimized is not purchasing power. It is the non-revisability of the ledger.
Here is where the selective application becomes visible. Run Li Lu's standard against gold. Gold has no cash flow, no issuer, no national backing, no enforcement mechanism of the legal type. By the letter of the argument, gold fails every criterion. Yet the Graham-Dodd tradition holds gold as a legitimate store of value and has done so for a century. The standard is not applied uniformly. It is applied to the new asset and suspended for the old one.
I want to be fair to the strongest version of the position, though, because the weak version is easy to beat. And the strongest version is not in the speech. It is in the mechanism.
What actually enforces the ledger, and why it is weaker than maximalists claim.
The honest answer is that proof-of-work is the outer shell and social consensus is the core. The 21 million cap is not enforced by mathematics. It is enforced by the fact that changing it requires coordinated adoption across miners, node operators, exchanges, and holders, and that coordination is expensive. We have a controlled experiment on this. The 2015 to 2017 block size war ended with the network refusing to change the base rule and a minority forking away as Bitcoin Cash. The change cost was demonstrated to be extremely high. That is the enforcement mechanism working.
But note what the experiment also demonstrated: when consensus fractures, the enforcement mechanism does not resolve the dispute. It duplicates the asset. Both chains continued. Holders received both. The ledger's immutability held at the individual chain level and dissolved at the network level.
This is the part of Li Lu's argument that survives contact with evidence, and I will not pretend otherwise. The enforcement mechanism is real, and it is social before it is mathematical. Its strength is a function of how many independent actors would have to be simultaneously persuaded, and that number is a sociological quantity, not a cryptographic one. It is high today. It is not a constant.
This is also why the debate between the two camps is permanent. You cannot refute a prior with a benchmark. Li Lu's objection is not addressable by higher throughput, lower energy, or better decentralization โ those are answers to questions he did not ask. The disagreement is not about Bitcoin's properties. It is about whether trust can be relocated from an institution to a protocol plus a social layer. Everything else is downstream.
Where the argument is simply correct.
Three monetary functions. Store of value, medium of exchange, unit of account. Bitcoin's penetration in the second and third is approximately zero, and this is not a temporary condition awaiting better UX. Unit of account requires price stickiness, which requires low volatility, which is incompatible with an asset whose price discovery is driven by marginal belief. Medium of exchange requires that the asset not be expected to appreciate, because a appreciating medium of exchange is a medium nobody spends. The functions are in tension with each other, and Bitcoin chose the first one.
So the value proposition rests entirely on store of value. And store of value for a non-yielding asset is reflexive โ its value depends on the expectation that others will value it, which depends on the expectation that others will expect others to value it. Price and belief are coupled in a positive feedback loop with no obvious damping term.
This is the part of the critique that has real force, and it deserves to be stated precisely rather than dismissed as generational incomprehension. Reflexivity is not fraud. There is no counterparty, no pool, no promised return, no promoter obligated to pay early participants with late participants' money. The distinction matters and it is structural, not semantic. But the absence of a promise does not eliminate the risk. It relocates it โ from counterparty risk to consensus risk. You have not removed trust from the system. You have changed its object, and you have made its failure mode correlated rather than idiosyncratic.
That is the trade. It is a real trade. It is not free.
The empirical claim, measured.
The one assertion in the speech that is testable is that crypto facilitates crime. The article supplies no data for it. So I checked the order of magnitude against public chain analytics and against the traditional-finance baseline, because an unquantified claim is not an argument, it is a mood.
Illicit addresses as a share of on-chain transaction volume has generally registered in the low single digits, frequently well under one percent depending on methodology and year. Money laundering through the traditional financial system, by UNODC estimate, runs at two to five percent of global GDP. Directionally, the claim is weak. It is also not zero, and pretending otherwise is the kind of advocacy that gets researchers ignored. Ransomware settlement, darknet markets, and sanctions evasion are domains where the asset's properties โ permissionless transfer, censorship resistance, settlement finality without an intermediary โ are genuinely efficiency advantages for the adversary. The same properties that make it a hedge against account freezes make it a tool for evading legitimate freezes. You cannot engineer one without the other. The property is singular.
Composability is a double-edged sword, and so is censorship resistance. Anyone who tells you otherwise is selling something.
The trace that gives the whole thing away.
There is one passage in the reporting that I keep returning to, because it is the only place the speaker's own framework visibly strains. Li Lu defines wealth as a share of global purchasing power. If applied consistently, that definition implies that an asset whose share of global purchasing power rises over time is creating wealth. Which is, almost verbatim, the bull case.
The reporting notes that he did not connect the two. That gap is the most informative artifact in the entire transcript. It is the visible seam where the framework refuses to close. And it tells you the conclusion preceded the analysis.
I have seen this pattern before, from the other side. In 2021 I audited fifty popular ERC-721 contracts for a Singapore fund and found that roughly eighty percent of the top mints lacked meaningful access controls on the mint function โ open to griefing, gas wars, and supply manipulation. The artistic narrative was decoupled from the code. But the more useful finding was not the eighty percent. It was that the audit reports circulating at the time were written to confirm what buyers already believed. The conclusion preceded the analysis there too. It just pointed the other way. Architects build, auditors break. Nobody in that market wanted the breaking.
The structural inversion nobody priced.
The most consequential fact in the source material is not a quote. It is an editorial juxtaposition. The article places "Li Lu is bearish" next to "spot Bitcoin ETFs have absorbed billions since 2024" โ and does not notice that the two sentences describe a decoupling.
Here is what changed, and it is a change in market microstructure, not sentiment.
Before the ETF wrapper, the marginal buyer of Bitcoin was a self-selected, self-custodial participant who had formed a thesis. That buyer read the arguments. That buyer was persuadable, and therefore that buyer's behavior was coupled to the discourse. Opinion leaders had price impact because they had audience impact, and audience impact converted into orders.
After the wrapper, the marginal buyer is a model-driven allocator. The decision is made by a portfolio construction rule โ a target weight, a rebalancing band, a risk budget โ executed by an advisor or an index methodology. That buyer does not read Li Lu. That buyer does not read Matthew Sigel either. That buyer does not participate in the question of whether Bitcoin is money. The allocation is a covariance input.
Which produces the genuinely novel structural fact: the philosophical dispute and the price have been severed. The bear case can win every argument and lose every flow. The bull case can lose every argument and keep the flows. Narrative influence and capital allocation now run on separate rails, and the second rail is the one that clears.
And here is the second-order effect, which I think is the real story and which almost nobody is pricing.
The asset was designed to eliminate the trusted third party at settlement. The institutional adoption path reinserted one at exactly the layer where the monetary claim is weakest. An ETF share is not a bearer asset. It is a claim on a custodian, mediated by an authorized participant, administered by a transfer agent, reconciled by a service provider, and settled on a ledger that is not the chain. When you hold the wrapper, your enforcement mechanism is not thermodynamic. It is legal. It is the same class of mechanism Li Lu says is the only legitimate one โ which means the institutional bid is validating his premise while falsifying his prediction.
That is the custodial reversion. The thing that made the disappearance prediction look wrong is the same thing that hollows out the property he was wrong about. And it happened without a single vote being cast by anyone who understood what they were voting on.
I spent eight months in 2022 reverse-engineering the Groth16 proof generation circuit in zkSync Era and found a constraint-system bottleneck costing roughly fifteen percent of transaction finality. That was a fixable engineering problem with a measurable cost. The custodial reversion is not that. It is an architectural drift with no owner and no bug bounty. Silence is the ultimate verification, and the silence here is the sound of a property being transferred out of the asset and into a service-level agreement.
The reporting problem is the same problem.
One more observation, and it is the one that makes this piece worth writing rather than just reading.
A speech arguing that an asset is illegitimate because it lacks a verifiable enforcement mechanism reached the public as a chain of trusted intermediaries: a speaker, an audience, a paraphrase, a poster, a reporter, a reader. Six hops. No transcript. No timestamp. One unsourced number.
The symmetry is almost too neat to be accidental. The critique of an unverifiable trust-minimized system was itself unverifiable and trust-maximized. And the inflow figure โ the single quantitative claim in the piece, the one that actually bears on whether the prediction is being falsified โ is the one with a source field reading empty.
I am not arguing the speech didn't happen. The context is self-consistent, the venue is real, the lineage is documented. I am arguing that the epistemic status of the claim and the epistemic status of the report are the same, and neither is high. Trust is math, not magic โ and there was no math anywhere in that transmission chain.
The contrarian read: stop arguing with the bear case, watch the wrapper.
The consensus response from the digital asset side is that Li Lu doesn't get it, that VanEck's Matthew Sigel is right that the critics keep losing the argument, that a value investor applying a DCF to a non-cash-flowing asset is a category error. All of that is true and all of it is boring. It is the same exchange that has run since 2013 and it has never once changed a mind.
Here is the less comfortable reading. Both sides hold unfalsifiable cores. The bear core is "money requires state enforcement." The bull core is "sound money wins eventually." Neither is updatable by evidence. That is why the debate is ritual rather than inquiry โ and why the Graham & Dodd luncheon is the correct venue for it. A speech at a lineage ceremony is a statement of identity, not an output of research. It is aimed inward.
The regulatory tail risk is not a ban. A ban has no political path โ the ETF exists, the major jurisdictions moved from prohibition to licensing between 2023 and 2025, and the tax and custody infrastructure is now too entangled to unwind. The tail risk is absorption: bank custody thresholds, stablecoin yield restrictions, criminalization of privacy tooling, attestation regimes that make self-custody operationally expensive without ever outlawing it. Bans are legible. Restrictions are not. Nobody protests a reporting requirement.
And the bear case that actually bites is not "the government kills it." It is "the government wraps it, and the wrapper becomes the asset."
Takeaway.
The prediction to watch is not whether Bitcoin disappears. It is whether the property that makes it interesting survives its own distribution.
Three things I would instrument if I were building the dashboard. First, the ratio of ETF-wrapped supply to self-custodied supply, and whether the wrapped share crosses the point where the marginal vote on consensus is held by entities whose enforcement mechanism is legal rather than thermodynamic. Second, the cadence and independence of custodian attestations โ a quarterly PDF from a counterparty is not a proof, and the gap between attestation and proof is where the next failure will live. Third, the next consensus-scale dispute: quantum migration key schemes, or fee-market sustainability as the subsidy decays. That will be the real test of whether social consensus is an enforcement mechanism or a shared assumption.
Li Lu is wrong that crypto has no enforcement mechanism. He may be right that the one it has is thinner than its holders believe.
Which of those two errors turns out to be expensive depends entirely on a question neither camp is asking: when the enforcement mechanism is social consensus and the marginal holder never reads the argument, who exactly is doing the enforcing?