A futures contract that never expires is not a futures contract. It is a bet wearing a legal wrapper. Last week, MNX announced it would list futures tied to the valuations of OpenAI and Anthropic โ two private AI companies โ and packaged the whole thing as 'democratizing access to AI valuations.' I read the announcement twice. Then I looked for the oracle. Then I looked for the settlement clause. Then I looked for the audit. None of them were there. The code does not lie; only the founders do, and here the founders have said almost nothing at all.
What MNX is selling is not a derivative in any functional sense. A derivative derives its value from an underlying that can be observed, traded, and โ critically โ hedged. MNX's underlying cannot be observed continuously, cannot be traded by the counterparty taking the other side, and cannot be hedged by anyone. That is not a derivative. That is a prediction market with a futures label stapled to the front, and the label is doing all the work.
The tell is the vocabulary. 'Democratization' is a word that appears when a product's mechanism cannot survive plain description. Nobody needs to democratize Treasury bills, because the mechanism is legible. When a team reaches for 'democratize' and 'reshape investment strategy' in the same paragraph, they are not describing a product. They are describing a marketing thesis, and the mechanism is somewhere behind it, unbuilt.
I have audited contracts that laundered risk through six layers of indirection. I have watched founders explain away a reentrancy hole as 'an edge case.' This is different. This is a case where the core mechanism โ the thing that makes the product a product โ was never specified in public at all. And in security work, an unspecified mechanism is not a neutral fact. It is the single loudest alarm in the building.
Context: How We Got Here
To understand why MNX's announcement is structurally fragile, you have to understand the market it claims to open. Private company equity โ the stock of Anthropic, of OpenAI, of every venture-backed firm before its IPO โ has always been illiquid by design. That illiquidity is not an accident. It is a feature of the venture model. Employees hold restricted stock units they cannot sell. Early investors hold positions they can only move in structured secondary rounds. The company controls who buys its shares and at what price, because uncontrolled ownership transfer would let competitors, hostile actors, or short-sellers into the cap table.
A secondary market for that equity exists anyway. Forge Global, EquityZen, Carta's secondary product, and a handful of broker-dealers run it. They operate under securities law. They require accredited investors. They settle in weeks, not seconds. They charge spreads that would make a DeFi trader wince โ often 5% to 15% on either side. And they do it this way because the underlying asset is genuinely hard to price, genuinely hard to transfer, and genuinely regulated.
The crypto industry has tried to shortcut this twice before, in ways worth remembering. The first attempt was the ICO era, when tokens were sold as 'future equity' in companies that never existed. Most of those went to zero, and the SEC spent years afterward cleaning up the wreckage. The second attempt was more sophisticated: FTX's Pre-IPO perpetual futures, launched in 2021, which let users speculate on the valuations of companies like Coinbase and SpaceX before their listings. Those contracts worked fine โ right up until the exchange holding the collateral imploded in November 2022, and every position on the platform was settled at zero simultaneously. Not because the underlying failed. Because the venue did.
That is the precedent MNX is walking into. Not a fresh frontier. A graveyard with a fresh coat of paint. The AI narrative is the paint. The paint is good โ it is probably the best narrative available in this cycle โ but paint does not change what is underneath.
And the AI valuation narrative itself deserves a cold look. Since 2023, private AI companies have been marked up on the strength of compute spending, model capability claims, and the assumption that a handful of firms will capture the value of an entire technological transition. Some of that is real. Some of it is a self-referential loop, where the same funds that mark up the private valuations are the ones whose public portfolios benefit from the mark. The valuations are not prices. They are negotiated numbers, set in rounds that happen once or twice a year, disclosed selectively, and defended by everyone with an interest in the number staying high. That is the 'asset' MNX wants to build a continuous market around.
I do not trust the audit; I trust the gas fees. Here there are no gas fees to trust, because there is no on-chain footprint disclosed, and there is no settlement logic published. There is a press release. That is the whole asset base of this analysis, and the absence of more is itself the finding.
Core: The Systematic Teardown
The Oracle Problem: You Cannot Quote a Price That Does Not Exist
Every derivatives market rests on a price feed. For BTC perpetuals, the feed is a composite of spot prices from dozens of venues, each independently arbitraged, each continuously quoted. If one exchange's feed drifts, arbitrageurs close the gap in milliseconds. The price is real because it is contestable โ anyone can trade against it and profit from the correction.
Now remove the venue. Remove the continuous quote. Remove the arbitrageur. What is left for a private company like Anthropic?
What is left is a valuation event โ a funding round โ that happens perhaps once a year, sometimes less. Between rounds, the 'price' of Anthropic equity is not a number that exists in the world. It is a number that people remember from the last round, adjusted by rumor. There is no aggregator, because there is no aggregation to do. There are no competing market makers quoting a two-sided market, because no one can settle a trade without the company's consent, and the company is under no obligation to consent.
So what does MNX anchor to? The announcement does not say. That silence is the most important line in the entire story. If the anchor is the last funding round, then the feed updates annually and is stale by construction โ a 'futures' price that lags reality by up to twelve months. If the anchor is secondary-market prints, then the feed depends on sporadic, private, non-binding transactions that no one is obligated to report. If the anchor is a proprietary index, then the platform is the oracle, and the platform is also the counterparty, and the platform is also the settlement authority. That is not a market. That is a casino that owns the roulette wheel, the table, and the definition of red.
I have spent years stress-testing oracle designs. The rule is simple: an oracle is only as strong as the number of independent parties who profit from correcting it. Chainlink works because dozens of node operators are slashed for lying and paid for accuracy. A private valuation feed has exactly one interested party โ the platform โ and that party is on the other side of the trade. This is the highest-severity technical risk in the entire product, and it is not a bug that can be patched. It is the architecture.
The Settlement Problem: Futures That Cannot Converge
A futures contract has a settlement date. On that date, the contract resolves against a reference price, and the position closes. This is not a formality. It is what forces the futures price to converge toward the spot price as expiry approaches. Without convergence, a futures price is just a number floating free of any anchor, and the entire economic logic collapses.
Ask the obvious question: what happens if Anthropic never IPOs? What happens if OpenAI stays private for a decade, raising money in quiet rounds, never listing? What is the settlement price of a contract whose underlying never produces a public, verifiable price?
The announcement says 'futures contracts.' It does not say when they settle, against what, or by whose authority. If they settle at the next funding round, then the contract's entire payoff depends on an event the company controls and can time โ which means the company, not the market, decides when your position resolves. If they settle at an IPO that may never come, then the contract is a perpetual with no convergence mechanism, which is mathematically incoherent as a futures contract. If they settle at a platform-determined mark, then the platform decides your payoff directly.
There is a reason traditional futures have delivery months. It is the same reason bonds have maturity dates. Time-to-settlement is the mechanism that ties a derivative to reality. Strip it out, and you are not trading a derivative. You are trading a sentiment.
This is the second structural failure, and it compounds the first. A bad oracle corrupts the price. A missing settlement mechanism corrupts the exit. Together they mean the trader has no reliable entry price and no reliable exit. In a market with neither, the only party with an edge is the one who defines both. That party is MNX.
The Market Maker Paradox: No One Can Hedge the Other Side
Here is where the product's economics become impossible rather than merely risky.
In a healthy derivatives market, the market maker takes the opposite side of customer flow and hedges the resulting exposure in the underlying or in a correlated instrument. When a customer buys a BTC call, the market maker buys spot BTC or a correlated hedge, and the residual risk is small and manageable. The market maker earns the spread and the funding, and the system stays solvent because the hedge exists.
Now ask the market maker to take the other side of a bet on Anthropic's valuation. To hedge, the market maker would need to buy or sell Anthropic equity โ which it cannot do, at scale, on demand, without the company's permission. So the market maker has two choices: refuse to quote, or quote a price with a risk premium so large that the product becomes unusable.
Do the arithmetic. If a market maker cannot hedge, it must reserve capital against the worst case. A valuation can move 50% in a single round โ that is normal for late-stage private AI companies, in both directions. To be safe against a 50% adverse move, the market maker needs to hold reserves of roughly that magnitude against every unit of exposure. That reserve cost translates directly into spread. You end up with bid-ask spreads that swallow the trade. A trader who buys at the offer and sells at the bid is down double digits before the position moves at all.
This is not a solvable problem. It is a structural property of trading an unhedgeable underlying. The only ways around it are to subsidize the market maker with token incentives โ which means the 'market' is funded by the platform, not by demand โ or to let the platform itself be the counterparty, which concentrates all directional risk in a single balance sheet with no hedge.
Consider what that last option implies. If retail traders are net long AI valuations โ and they will be, because the narrative is bullish โ then MNX is net short AI valuations. Every dollar of customer optimism is a dollar of MNX liability. If AI valuations keep climbing, as the narrative promises, MNX's book bleeds. The platform is structurally positioned against the very thesis it is selling. That is a business model that requires the product's narrative to be wrong in order to stay solvent. Or it requires the platform to control the settlement mark, so that 'wrong' never gets recorded. Pick your poison.
The Precedent: FTX Pre-IPO Futures Settled at Zero
I want to be precise about the history, because it is the cleanest available evidence.
FTX listed Pre-IPO perpetual futures in 2021. The contracts tracked the valuations of companies expected to IPO โ Coinbase, SpaceX, and others. The mechanism was internally consistent. The exchange had a functioning order book, real users, and enough liquidity to quote tight markets. On its own terms, the product worked.
It settled at zero anyway. Not because the underlying collapsed. Because the exchange holding the collateral collapsed, and when a venue dies, every position on it dies with it. The users who were 'long Coinbase pre-IPO' did not get Coinbase stock. They got a claim in a bankruptcy estate, ranked behind secured creditors, payable in years, if ever.
The lesson is not that Pre-IPO derivatives are inherently fraudulent. The lesson is that the venue is the real risk, and in a product where the venue is also the oracle, the settlement authority, and the counterparty, the venue risk is not one risk among many. It is the only risk that matters. The rug was pulled before the mint even finished โ in FTX's case, the collateral was gone before most users understood the contracts were even live.
MNX inherits every one of these venue dependencies and adds a worse oracle on top. FTX at least had public equities to reference for some contracts. MNX has private valuations controlled by the companies themselves. The failure mode is not hypothetical. It has a date, a name, and a bankruptcy docket.
The Regulatory Layering: Shadow Securities and the Howey Trap
Run the Howey test against the contract itself, not against any token. Investment of money: yes, the trader posts capital. Common enterprise: yes, the trader's outcome is pooled with the platform's and the counterparty's. Expectation of profit: yes, the entire product is a bet on valuation appreciation. Efforts of others: yes, the payoff depends on the platform's valuation feed and the company's performance.
Four for four. The contract reads as a security under US law. And not a simple one. Because the underlying is itself a security โ private company equity โ the derivative is a security-based swap or a security future, which pulls it into joint SEC and CFTC jurisdiction with registration requirements that no offshore platform can casually satisfy.
There is a deeper problem. The underlying equity in Anthropic and OpenAI is restricted. Shareholders sign agreements that limit transfer, and those agreements typically prohibit creating derivatives on the shares without company consent. So MNX may be constructing exposure to shares that the holders cannot legally let it reference. That is not a gray area. That is a contract-breach vector that gives the target companies a direct legal claim against the platform, independent of any regulator.
Now layer the geography. Anthropic and OpenAI are Delaware-incorporated. Their equity is governed by US law. Their shareholder agreements are enforceable in US courts. The natural way for an offshore platform to 'solve' this is to block US users and register in a permissive jurisdiction. But US enforcement does not stop at a screen door. The SEC has repeatedly reached offshore venues serving US persons, and the CFTC has done the same with offshore derivatives platforms. A geographic block is a speed bump, not a defense.
The realistic outcome is not that MNX operates a compliant US product. The realistic outcome is that it serves users in regulatory gray zones, until it attracts enough attention to trigger enforcement, at which point the venues it depends on โ payment processors, stablecoin rails, exchange partners โ cut it off. And when the rails are cut, the settlement mechanism, whatever it is, does not get to run. The position just freezes.
The Missing Information: What the Announcement Does Not Say
Let me list what a serious derivatives launch would disclose, and mark each against what MNX disclosed.
Valuation methodology: not disclosed. Settlement rules: not disclosed. Oracle design and data sources: not disclosed. Audit reports: not disclosed. Team identities: not disclosed. Jurisdiction and legal structure: not disclosed. Token economics, if any: not disclosed. Liquidity commitments or market maker agreements: not disclosed.
The announcement disclosed one thing: that contracts tied to OpenAI and Anthropic valuations would exist, and that this constitutes 'democratization.'
In my experience, the information that is missing from a launch announcement is not random. Teams disclose what makes them look strong. They omit what they cannot defend. A derivatives product that omits its oracle and its settlement rules has omitted the two things that define whether it is a derivatives product at all. That is not an oversight. It is a signal.
I have seen this pattern in audits. When a project's documentation is thin in exactly the places where the mechanism should be, it is because the mechanism does not exist yet, or exists in a form the team is unwilling to publish. Both possibilities are disqualifying for a user deciding whether to post collateral.
The Narrative Machine: 'Democratization' as a Tell
Strip the press release to its verbs. 'Democratize.' 'Reshape.' 'Access.' These are not technical terms. They are persuasion terms. They do the work that a specification would do, if a specification existed.
Ask who actually benefits from democratization in this framing. Traditional Pre-IPO access is restricted to accredited investors not because the elite hoard it, but because the asset is illiquid, opaque, and legally restricted, and the restrictions exist to protect people who cannot absorb the loss. 'Democratizing' access means giving that risk to people less equipped to hold it, in a wrapper that hides the illiquidity and the legal exposure behind the word 'futures.'
The direction of the transfer is the tell. In a genuine innovation, the new mechanism expands what is possible. Here, the new mechanism expands who absorbs the downside. The upside โ the spread, the funding, the settlement discretion โ stays with the platform. The downside โ the unhedgeable, unpriceable, legally fraught exposure โ is distributed to retail.
The AI narrative is doing the emotional work that the mechanism cannot. AI is the hottest story in this cycle, and anything with 'AI valuation' in the title gets attention. That attention is the product. The contracts are the excuse.
Contrarian: What the Bulls Got Right
I am not going to pretend the demand is fake. It is not. The case for a private-market derivatives venue is stronger than the crypto crowd admits, and I will give it its due before I explain why this particular implementation cannot carry it.
The traditional Pre-IPO secondary market is genuinely broken. Forge, EquityZen, and Carta operate with spreads of 5% to 15%, settlement times measured in weeks, and access gates that exclude most people who want exposure to private AI. For a retail investor who believes Anthropic is undervalued and wants a liquid way to express that view, there is no good option. The bull case is that a continuous, low-spread, always-on market for private valuations would serve real demand that the incumbents cannot serve. That is a legitimate thesis, and it is not stupid.
The bull case is also right that the AI valuations themselves are the defining asset class of this decade. Whether or not MNX's contracts work, the underlying interest in AI equity is real and durable. The mistake is not the ambition. The mistake is believing the ambition can be met with a mechanism that ignores the two hard constraints: you cannot continuously price an asset that has no continuous price, and you cannot settle a contract against an event that may never occur.
The bulls also have a point about the incumbents' opacity. The private secondary market is clubby, relationship-driven, and slow. A crypto-native venue that solved the oracle and settlement problems โ using, say, legally structured SPVs with periodic audited marks, or cash-settled contracts with explicit event triggers โ could genuinely serve demand the incumbents leave on the table. The problem is that solving those problems requires legal engineering and honest disclosure, which is slow, expensive, and unglamorous. It does not fit a launch announcement. So instead of solving them, MNX described a product that presupposes they are solved.
Here is the part the bulls miss. The reason no one has built a clean, liquid, retail-accessible market for private AI valuations is not that nobody thought of it. It is that the constraints are real, and every serious attempt runs into them. FTX ran into them and died. The incumbent brokers run into them and stay slow and expensive on purpose, because that is the only way to stay solvent. When you see a product that appears to ignore constraints that killed everyone who came before, you are not looking at an innovation. You are looking at an omission. Reentrancy is not a bug; it is a feature of trust. Here the trust is in a valuation feed nobody can verify.
Takeaway
MNX's contracts will be judged by one metric, and it is not the narrative. It is the spread. If the bid-ask on an AI valuation future is wider than 5%, the product is functionally dead on arrival โ no rational trader pays that round-trip cost to express a view. If the spread is tight, ask who is subsidizing it, because no unhedgeable market can be tight without a subsidy, and the subsidy comes from somewhere: token incentives, platform capital, or the settlement mark itself. Follow the subsidy, and you find the risk.
The code does not lie; only the founders do. And here there is no code to read, only a press release that describes a derivative without a price, a settlement without a date, and a market without a hedger. Until MNX publishes its oracle design, its settlement rules, and an independent audit, the only defensible position is no position. The AI narrative will keep selling. The question is who is left holding the contract when it finally has to settle โ and against what.