You're not losing to the market. You're losing to yourself.
© 2026 Kodex Academy · Education, not financial advice. The Kodex crew are fictional personas.
An option's price is not mainly a guess about where the stock is going. It is arithmetic about what it costs to hold that stock until the option expires, and a desk in Chicago runs that arithmetic every time it quotes one: it borrows the cash, carries the shares, collects the dividends along the way and pays the financing. The number that falls out of that is the forward price, and every option listed on a US share is built on top of it. You never have to work it out yourself. Rates are published, dividends are announced, and the cost of borrowing a share has a market of its own with prices in it.
None of those numbers belong to the exchange selling you the option.
On 17 September at 20:00 UTC, Bybit lists contracts where that machinery is not present. The product is called Perp Options, and the first two are written on SpaceX and Nvidia. They trade around the clock and settle in USDT, which is why you can buy a fraction of one instead of the hundred shares a listed contract controls. Tesla, the Invesco QQQ ETF, SOXL and Micron are queued behind them.
Tao is the bridge between structure and instinct at Kodex, and he arrives at this one relaxed. An option on Nvidia is an option on Nvidia. He has priced a thousand of them. He opens the contract page and reads down it the way he reads everything. He is looking for one line, the one that names what the thing settles against. He stops on the second.
Tao backs up to a contract he does trust, an ordinary Nvidia call listed in Chicago, and takes it apart in front of you. It all begins at spot, the last print on the share. What sits on top of that is the cost of carrying the position to expiry, starting with the cash you borrow to hold it, priced by a money market with thousands of participants in it. Some of that cost comes back to you, because Nvidia announces a dividend on a date it publishes in advance. Going the other way charges you again. A lender sets a price to lend you the share so you can be short it, and that price moves when the share gets hard to find.
Put those four together and you have the forward. It is the price at which you could contract today to own that share on the day the option expires. The premium sits not on the spot price you see quoted, but on the forward.
"The forward is the part you never see quoted," Tao says. "It is also the only part that is not an opinion."
Everything feeding that number is produced by somebody with no position in your trade. The exchange listing the option does not choose the financing rate or set the dividend. It quotes a price that reflects them, and when the quote drifts away from what they imply, somebody arbitrages the gap and it closes. Your premium is checkable by anyone willing to look up four public numbers.
Now take the expiry date away.
A perpetual future never settles, so there is no date to price toward and nothing to build a forward curve out of. What holds the contract near the share price instead is a payment. Longs pay shorts, or shorts pay longs, at intervals the venue sets, and the size of that payment is the funding rate. Kodex has already taken that apart as a bilateral payment in perpetual futures vs swaps. What matters here is narrower, and it sits one layer up.
The premium part of that rate measures how far the contract has drifted above or below the venue's own index, and the interest part prices what holding that drift costs. MetaMask's description of the general mechanism is that funding is derived from the perp's premium to the index plus an interest component, subject to platform-specific caps. Exchanges set their own intervals, formulas and caps.
So the carry inside a listed option comes from a money market. The carry inside a perpetual comes from a formula the venue wrote.
Ackerer, Hugonnier and Jermann formalised exactly that in Perpetual Futures Pricing, published in Mathematical Finance in 2026. Their central result is that a perpetual's price is the risk-neutral expectation of the spot sampled at a random time, and the randomness in that timing is the intensity of the anchoring itself. Read that slowly, because it puts the funding specification inside the price rather than alongside it.
Tao reads it twice. "The tether is not a safety rail bolted on afterwards," he says. "It is in the number."
An option written on that contract inherits it.
That inheritance runs one level further down, to the line Tao stopped at. Finance Magnates reported it plainly: the options do not settle against SpaceX or Nvidia shares directly. Their underlying is Bybit's own perpetual futures contracts, SPCXUSDT and NVDAUSDT. Not the share. The venue's contract on the share.
Which raises the question the page does not answer. A venue quotes several prices for the same contract. There is a last traded price, an index price assembled from outside references, and a mark price used to decide liquidations. Those three are not the same number at the same moment. The distance between them is the distance between a position that survives a wick and one that does not. Finance Magnates also notes that Bybit has not published the contract specifications, settlement methodology, margin requirements or jurisdictional restrictions for Perp Options. The clause is not hidden. It has not been written down yet.
"I can price the volatility," Tao says. "I cannot price the clause that is missing."
A settlement reference is normally a separate question from a contract's price. Which index a contract settles to is a piece of work on its own. Here the two questions collapse into one. Strip out the forward curve and the settlement reference does both jobs. It anchors the price on the way in. It decides the payoff on the way out.
"Line them up," Tao says. "Three objects, one question each."
| What supplies the carry | What the payoff resolves against | Who chooses that number | |
|---|---|---|---|
| Listed Nvidia option | a forward built from published rates, dividends and borrow | the share price on the listing venue | the market, through observable inputs |
| Bybit perp option | the perpetual's funding rate | Bybit's own perpetual contract | the venue |
| Everlasting option (XPO) | a streaming premium, no expiry at all | the underlying pool | the protocol's fee model |
That last question is the one that separates them. A market answers it in the first case, a single venue in the second, a fee schedule in the third. None of that is visible in the product's name, and all three are sold as options.
That third row has a name of its own, and it happens to be the same name.
A perpetual option, written XPO, is an option with no expiry date at all. You exercise it whenever you like and it never runs out. That is what the term has meant since long before this product existed, and it is what the reference literature still means by it. The Ackerer paper prices those too, in closed form, and calls them everlasting options.
Bybit's contracts do not clear the first line of that definition, because they have expiry dates. Both outlets that covered the launch reported the same detail: new expiries will be added on an ongoing basis. That is only a sentence you write about a product that expires. What is perpetual here is the underlying. Not the option.
Several outlets still wrote the launch up as an equity perpetual option. Bybit's own name for it is Perp Options. That is the more accurate label and the less familiar one. The shorter phrase drifted onto the longer one because they sound alike.
There is a one-question test that survives any label. Does the contract have an expiry date? If it does, you are holding a dated option, whatever is underneath it, and time decay works the way you already know. If it does not, you are holding a different instrument entirely, and what you pay is a stream rather than a premium.
"The name is the least reliable field on the page," Tao says.
Kodex has a walkthrough on how settlement currency changes an option's payoff. That is the adjacent question and a different one. Not what you get paid in, but what the payoff is measured against.
Count the objects standing between you and Nvidia.
You hold an option. The option is written on a perpetual. The perpetual tracks the share. The share sits at the far end as a reference. Finance Magnates described those perpetuals as tracking the stocks' prices without Bybit holding or delivering any actual shares. Settlement is in USDT and the lots are fractional. Nothing in the published material describes a share moving to anybody at any point.
That is an absence in what has been published so far, not a denial Bybit has issued, and the distinction is worth keeping.
It also does not make the product defective. Cash-settled contracts that never touch the underlying are ordinary, and index options have worked exactly that way for decades without anybody losing sleep. What changes is the shape of your claim. You are not owed a share anywhere in that chain. You are owed a payment, computed by the venue, from a number the venue publishes.
Tao came in to check the volatility surface. He has three questions written down instead, none of which the page answers: which price settles it, who publishes that price, and what the funding term does to a premium carried across an expiry.
What happens to that count when the object at the far end barely trades? SpaceX is listed, and the slice of it that actually trades is small. Kodex put that number at roughly five percent of shares outstanding while working through the supply cliff behind the SpaceX short interest. Five percent is the float. It is the part loose enough to move on a given day.
Against that, The Block reported tokenized-equity perpetual volume rising from about $85 billion in January to roughly $470 billion in June. The SpaceX perpetual alone cleared $66 billion in June. Those are its own figures from its own prior reporting, and Bybit disclosed none of them.
Putting those two facts beside each other is not the same as comparing them. One is a percentage of a share count, the other a notional traded across venues, and the arithmetic that would make them commensurate is not available here. The narrower claim is the one that holds and the one that matters: the reference at the far end of the SpaceX chain is a thin market, and thin references move on smaller flows than deep ones do.
Nvidia is the opposite case at that end of the chain and the same case at this one. It prints on Nasdaq all day in enormous size, and there is nothing thin about the reference. The problem there is not the depth of the reference, but the hours it keeps. Nasdaq's cash session runs 9:30 to 16:00 Eastern, five days a week, while the perpetual underneath your option runs continuously. An expiry landing outside those hours resolves against a number the share market was not producing at the time. There is no print to check it against. None exists.
"Same product, two different reasons to read the settlement clause," Tao says. "That usually means the clause is doing more work than the marketing."
The habit that comes out of this is small, and it transfers to anything. You count the wrappers between you and the asset, then ask who publishes the number at the end of the chain and whether that party is also the one selling you the wrapper. For a listed Nvidia call the answers are one wrapper and no. For a Bybit perp option, on everything published so far, they are two and yes.
That is not a verdict on the product. There are real reasons to want an equity option that trades on a Sunday night. Reasons to want one in a size you can afford, without a broker's calendar in the way. A single listed contract controls a hundred shares, which has kept those options out of reach of small accounts for decades, and this walks around that wall.
What you are holding is a price with one more author in it than the version you are used to. On 17 September the contracts list, and the specification either arrives alongside them or it does not. Read that page before you read the volatility surface. The surface is an opinion about the future. The settlement clause is a fact about who pays you.
You now have a question that works on anything with a wrapper on it. The cheapest place to make it a habit is somewhere the answer is boring. Take a tokenized stock position in the Kodex simulator on the $5,000 paper balance. Before you size it, say what your profit and loss will resolve against, and who publishes that number. Learn the question where it is easy to answer, so you still ask it when it is not.
