I spent a Tuesday evening last month staring at a CoinGlass tab for a product that, by any honest reading of financial engineering, should not exist.
It was a perpetual futures contract tracking Anthropic — the AI lab that filed a confidential draft registration statement with the SEC back in June. Roughly $79 million in open interest. More than $20 million in 24-hour volume. Binance carrying close to 40% of the activity, with the remainder scattered across smaller offshore venues.
Then I noticed the columns that were empty. No circulating supply. No spot market. No underlying asset of any kind.
This is not a tokenized share. It is not a security. It is a cash-settled perpetual future — a number that rises and falls because two groups of traders are betting against one another about the eventual value of a company that had still never acknowledged the contract's existence when I ran my numbers.
The quote on the screen read $2,147. I have read every Binance Research note I could find, and I still cannot tell you what unit that number represents — per share, indexed, or derived from some undisclosed valuation assumption. That opacity is not a footnote. It is the product.
To understand why this matters, you have to understand what a perpetual future actually is. When I spent four months in 2017 auditing EtherTrust's smart contracts for a reentrancy bug that could have drained $4.2 million, the lesson was never really about Solidity. It was that a financial instrument is only as honest as its settlement mechanism. Perpetuals were built for assets that trade somewhere else. A funding rate — that periodic payment flowing between longs and shorts — exists for exactly one purpose: to tether the contract price to a real spot market. When the contract drifts above spot, longs pay shorts, and arbitrageurs sell the contract and buy the underlying until the gap closes. That is the mechanism. That is the whole machine.
Strip out the spot market, and you have removed the engine of it.
There is no Anthropic share to buy. Anthropic is still private. So there is no spot leg to arbitrage against, no delivery mechanism, no forced convergence, no natural seller to answer an inflated bid. The funding rate on this product has nothing to anchor to — it is a signal with no receiver. Polymarket-style event contracts at least resolve against an external fact. A perpetual with no spot resolves against nothing at all. You can push this price to any number the two sides agree to, and nothing in the structure will pull it back.
I want to be careful here, because the reflexive response is "so what — it is just a casino." Casinos are honest about being casinos. This product is being narrated as price discovery. CryptoSlate, in its coverage, framed the market as establishing a reference price for Anthropic ahead of Wall Street. That framing is doing a lot of quiet work, and it deserves scrutiny.
Consider the arithmetic. Anthropic's rumored valuation sits around $2 trillion. The entire open interest in the Anthropic contract is roughly $79 million. That is four-hundredths of one percent of the valuation it is supposedly discovering. It is not the tail wagging the dog; it is a flea on the tail wagging the dog. The number on the screen is not a price — it is the current midpoint of a leveraged argument.
And the argument is loud. Combined open interest across Anthropic and OpenAI perpetuals has crossed $160 million. The narrative runs that open interest jumped from about $1 million in April to over $160 million now. Except the same reporting describes that move as a 179% increase. One million to one hundred sixty million is a 15,900% increase. Those two figures cannot both be true, and no one issuing them appears to have checked. When a market's foundational statistic is internally contradictory, you are not watching price discovery. You are watching storytelling with a spreadsheet.
I have seen this shape before. In 2022, after the exchange collapses, I read forty whitepapers from failed projects and documented the recurring pattern. It was never the technology. It was the gap between what a product claimed to do and what its mechanism could actually enforce. Trust is earned, not mined — and it is certainly not manufactured by an exchange's research desk publishing its own open interest figures. Binance Research is not an independent oracle. It is a marketing arm of the venue collecting the fees. Every data point it publishes about its own product should be discounted accordingly.
The bull market makes all of this harder to see, not easier. When everything is green, the instinct is to read participation as conviction. It is not. Open interest growth is a measure of engagement and leverage, not of directionality — for every long betting Anthropic moons, there is a short betting it does not, and the exchange collects from both sides of the bet. This is the one piece of genuinely honest data interpretation in the entire conversation, and it deserves to be said louder than it is.
Here is where I part ways with the more comfortable reading of this market.
The optimistic case is that even a flawed synthetic instrument produces useful information — that the crowd's aggregate bet on Anthropic carries signal traditional investors lack. I do not buy it. A price that cannot be arbitraged toward anything real is not information; it is noise with a number attached. When you remove the spot leg, you do not get a purer expression of belief. You get a purer expression of leverage. And leverage, in a narrative-driven market, amplifies whatever story is trending that week — not whatever is true about the company.
Watch what actually moves this contract. It rose when OpenAI shipped its Astra model. It dipped when IPO timing slipped. None of those are Anthropic earnings. None are revenue, margins, or burn rate. They are headlines. A market priced by headlines is a market that will reprice violently the moment the headlines stop. Soul in the machine requires a machine — and this one has no engine.
The deeper problem is regulatory, and it is not subtle. Run the contract through the four prongs of the Howey test and it fails all four: money invested, common enterprise, expectation of profit, reliance on the efforts of others. This is, by structure, an unregistered securities derivative on a security that is not yet public. Every prior crackdown on offshore venues offering similar products ended the same way — not with a debate, but with a delisting and open interest collapsing to zero in an afternoon.
The window here is short and visible. Anthropic's public listing was expected in October and has reportedly slipped to November. If that date slips again, or the offering is shelved, the narrative that sustains this market evaporates, and there is no fundamental floor underneath it to catch the fall. I should note, too, that the model spreads. Nothing about this construction is specific to Anthropic. SpaceX, Stripe, any hot private company with a compelling story can be wrapped into the same synthetic shell tomorrow. That scalability is precisely what makes it worth understanding now, rather than after the first venue gets a subpoena.
I do not write this as someone who thinks crypto should stay small. I spent 2024 building a curriculum to help institutional investors read blockchain projects honestly, and I did it precisely because I believe this technology can carry serious capital. But that belief comes with an obligation. DeFi must mature — and maturing means not dressing up a leveraged bet as a price signal, and not letting a bull market convince us that a number going up is the same as a truth being found.
The question worth sitting with is not whether Anthropic's perpetual trades at $2,147 or $1,147 next week. It is whether we, as an industry, are going to keep calling noise "discovery" because the noise is profitable — or whether we are finally ready to build instruments that anchor to something real. Conscience over consensus. The consensus right now is that this is innovation. The conscience knows better.
