Over the past 30 days, the loudest number in prediction markets has not been volume. It has been the incentive budget — the points, the rebates, and the liquidity-mining multipliers that platforms pay users for showing up and quoting both sides of a binary contract. According to a headline moving through aggregator feeds, that budget is now the subject of a US Commodity Futures Trading Commission inquiry. The same feed, in the same block of text, reported that the President had renamed "AI" to "SI."
One of those items is a signal. The other is noise wearing a signal's clothes, and in a market this thin, telling them apart is the entire job.
I want to be precise about what I can and cannot verify. The aggregator that carried both items cited no primary source. Its body text repeated its own headline. No link to a CFTC filing, no Wells notice, no named respondent, no docket number. What I have is a rumor about a regulator and a rumor about a word. But the first is directionally consistent with a two-year pattern anyone who has read the CFTC's event-contract docket would recognize. The second is a stress test of the reader's skepticism — and most readers failed it. It was forwarded, screenshotted, and traded on.
So let me do what I always do: throw out the headline, keep the mechanism, rebuild from the code up.
Prediction markets are older than most of the tokens that now fund them. The lineage runs from Intrade — which the CFTC effectively shut out of US retail access in 2012 — through the first generation of on-chain attempts: Augur, Gnosis, and the early market makers who learned the hard way that a prediction market is a liquidity problem disguised as an information problem. Then came the second generation: Polymarket on Polygon, Kalshi as a designated contract market, and a dozen smaller venues splitting the long tail of "will X happen by Y."
For most of that history, the regulatory question was settled by geography. Offshore venue, no US users, no CFTC jurisdiction. That settlement collapsed in 2022, when Polymarket paid a $1.4 million penalty and agreed to block US persons — a quiet admission that event contracts are derivatives and that derivatives have a regulator. Then the 2024 election cycle arrived and turned prediction markets into a mainstream growth story. Volume spiked into the billions. Every venture fund in the space published a thesis. And every thesis quietly assumed the growth would continue without a legal reckoning.
It will not. The reckoning is exactly what the aggregator headline, however poorly sourced, is gesturing at. Because the growth was never organic. It was subsidized.
I have a specific bias here, and I should name it. In 2017, at 25, I spent four months dissecting the tokenomics of EOS and Tron while everyone else watched the price. The output was a 40-page comparative study on centralization risk in delegated proof of stake. What it taught me was that a token emission is rarely a reward for security — it is a subsidy for behavior the protocol cannot attract organically. That habit — reading the emission schedule before the chart — is why I do not treat "prediction market growth" as a single variable. It is at least two: organic demand, and the price of renting it.
Here is the mechanism nobody wants to say out loud: the prediction market's product is not the market. The product is the incentive program, and the market is the thing the incentive program buys.
Strip away the marketing and a modern prediction venue runs on three economic layers. The first is the order book or the AMM that quotes the binary contract. The second is the oracle that resolves it. The third — the layer that actually determines whether anyone shows up — is the subsidy that pays market makers to quote when the natural spread would otherwise be too wide to attract takers. On a thin market, the maker's edge is negative; the incentive is what turns a losing quote into a break-even one. Remove the subsidy and the spread blows out, the depth evaporates, and the "market" becomes a ghost town with a nice landing page.
This is the same structure I pulled apart in 2017. The lesson then was that token emission was not a reward for security — it was a subsidy for a specific behavior the protocol needed but could not organically attract. The lesson now is identical, one layer up the stack. A prediction market's incentive program is not a marketing expense. It is an unregistered distribution channel. That is the sentence the CFTC cares about.
Think about what the incentive actually is. A user is promised points, or tokens, or a rebate, contingent on performing a specific economic action — providing liquidity to, or taking a position in, an event contract that pays out based on a real-world outcome. If that action constitutes trading a derivative, and if the reward for performing it is a transfer of value that depends on the efforts of a promoter, then the reward itself starts to look like an investment contract. The regulator does not need to ban the market. It only needs to say that the instrument you are using to grow the market is itself a security or an unregistered derivatives promotion. Pull that thread and the whole growth model unravels.
I have watched this exact unraveling from the inside. Based on my audit experience, the most dangerous document in any DeFi treasury is not the smart contract — it is the reward logic spreadsheet. In 2023 I reviewed the incentive contracts of a mid-cap perpetual DEX. Not the trading engine; the emission logic. The team had modeled their rewards as "community incentives" and their legal memo treated them as a cost center. Neither framing was right. The emissions were a continuous, discretionary transfer of value to a defined class of actors, for performing a defined economic act, administered by a centralized team with unilateral control over the multiplier. That is not a cost center. That is the anatomy of an offering, minus the paperwork. The prediction market incentive program is the same animal with a more legally exposed underlying instrument.
Prediction markets are structurally worse, because the underlying is an event contract, over which the CFTC has already asserted jurisdiction. When Kalshi fought the agency over election contracts in 2024 and won at the district level before the appellate picture muddied, the win was about whether the CFTC could block a licensed venue from listing a lawful product. It was not a ruling that unlicensed venues could do as they please. The market keeps collapsing that distinction, and the collapse is expensive.
Now layer in the token. Most on-chain prediction venues either have one or are pre-token, and the incentive program is the bridge between the two — points today, an airdrop tomorrow. This is where my long-standing discomfort with the Layer 2 sector sharpens into something more general. The L2 thesis was that fragmentation was a feature: more venues, more capacity, more users. What happened instead was that dozens of chains competed for the same small set of users, slicing already-scarce liquidity into ever-thinner fragments. Prediction markets are repeating the error at the application layer. Every new venue launches a points program, the programs compete for the same mercenary liquidity, and the "growth" is a rotation of the same wallets chasing the next multiplier. The dashboard goes up. The user base does not.
I can put a rough shape on the churn. A mercenary LP in a subsidized market is not loyal to a venue; it is loyal to the highest multiplier net of gas and impermanent loss. When a competing venue raises its emission, the LP migrates within a settlement window. What remains behind is the subsidized spread, which collapses the moment the incentive stops. The result is a market whose depth is a function of its emission rate — a curve that looks healthy until you plot it against the emission schedule, at which point it looks like what it is. I did a version of this exercise in 2021, when I pulled on-chain data from roughly 12,000 Art Blocks mints to show that secondary-market volume had decoupled from creator royalties. The same decoupling is happening now: reported volume is decoupling from real, unsubsidized demand. The metric and the reality have come apart, and the dashboard is the last place you will see it.
The information-pollution layer is the part of the story the aggregator accidentally made more interesting than the CFTC rumor itself. A feed that can carry a credible regulatory signal and an obviously false political claim in the same paragraph, with no source for either, has told you something about the state of crypto information: the signal-to-noise ratio has inverted. The "AI renamed to SI" item is not a mistake I can explain charitably. AI is a general-purpose technology, not a registrable asset with a ticker that a decree can rewrite. There is no plausible mechanism by which a word denoting a class of systems becomes "SI" by fiat. The item is a translation failure, a misread of a speech, or a fabrication — and it survived editorial review at whatever outlet originated it. That it then propagated is the real data point.
Here is why that matters for your portfolio, not just your reading list. In a bear market, the cost of acting on bad information is asymmetric. The upside of a good trade is capped by thin liquidity; the downside of a bad trade is not. Every forwarded headline is a latent order. When the headline is noise dressed as signal, the order is a loss waiting to settle. In this market, information hygiene is a risk-control function, not a hobby.
Let me be careful about the one thing I can anchor. The CFTC has jurisdiction over event contracts. It has exercised that jurisdiction repeatedly since 2022. It has a documented interest in how retail users are induced to trade derivatives. An inquiry into an incentive program fits the agency's established pattern precisely — which is why I weight it as a real signal even without a primary source. The "AI to SI" item fits no pattern, which is why I weight it at zero. Same feed, same block, opposite credibility. The only variable that separates them is whether the claim maps onto a mechanism that already exists.
There is a forward-looking wrinkle worth flagging, because it is the one most readers will miss. In 2025 I published a speculative framework on what I called the DAO of Algorithms — a system in which autonomous AI agents trade compute and settle with each other through smart contracts, largely without human intermediation. Prediction markets are the natural habitat for that. An agent does not need a points program to be induced to quote a contract; it needs a positive expected value net of latency. That is, in a real sense, the incentive program's endgame: the subsidy gets replaced by an agent that computes its own edge and provides liquidity because the math clears, not because a team is paying it to. If that happens, the regulatory question does not disappear — it sharpens. An autonomous agent taking a position in an event contract is still a participant in a derivatives market, and someone, somewhere, is still the promoter. Automation removes the human from the trade; it does not remove the human from the liability.
And that is where the macro frame closes the loop. In 2024, when the spot Bitcoin ETF was approved, I wrote a report called "The Liquidity Premium," arguing that the approval shifted Bitcoin from speculative tech to institutional asset class and changed its volatility profile in ways traditional-finance data could model. The same channel is now opening for prediction markets, and it looks identical: an activity that was legally ambiguous, wrapped in a structure a regulated venue can carry, sold to an audience that wants the exposure without the ambiguity. The offshore, token-incentivized venue is the pre-ETF asset. The licensed venue is the post-ETF wrapper. History rhymes, but the code doesn't — and this time the code is a compliance engine, not a liquidity pool.
The consensus reading of a CFTC probe is bearish for prediction markets. I think the consensus is wrong in the way that matters most.
The bearish case says regulation suppresses growth, growth was the thesis, therefore the token bleeds. But the growth was subsidized, which means it was never the thesis — it was a rental. What the probe actually does is accelerate a migration already underway: from offshore, token-incentivized venues toward licensed, fee-based ones. That is not the death of prediction markets. It is the re-rating of them from degen casino with a data layer to regulated information derivative. The venues that survive will look less like a points farm and more like an exchange. Thinner margins, lower multiples, real revenue.
Here is the uncomfortable part for anyone holding the narrative tokens. The prediction market's problem was never the technology. It was that the incentives were doing the work a product should do. I applied this logic to gaming NFTs in 2021: the blocker was never the tech, it was that the business model could not survive players owning their assets. The same shape holds here. A licensed venue does not need a token to attract liquidity; it needs a spread that clears. An unlicensed venue cannot build that spread without a subsidy, and the subsidy is what is under scrutiny. Remove it and you find out which venues have a product and which have a Ponzi with a resolution oracle bolted on.
There is a second turn, and it should worry the on-chain maximalists. The migration to licensed venues is the same story RWA has told for three years, and it ends the same way: the regulated venue does not need your public chain. Kalshi does not settle on Polygon. It settles where the law is legible. Every step toward compliance is a step away from the permissionless rail, because that rail is precisely what makes the activity ambiguous. If prediction markets win their regulatory fight, they may win it by leaving.
Watch three things, in this order. First, whether the CFTC rumor resolves into a primary document — a filing, a docket, a Wells notice — or evaporates back into the aggregator feeds it came from; the resolution tells you whether the signal was real. Second, whether incentive budgets at the top prediction venues shrink over the next two quarters, because a shrinking budget is the first measurable sign that the growth was rented. Third, whether liquidity migrates from offshore venues to licensed ones, because that migration is the whole ballgame and it is already visible to anyone reading the volume tables instead of the headlines.
The prediction market does not need to be rescued. It needs to be priced honestly. The venues that figure that out first will be the ones still quoting when the subsidy is gone. Which is, quietly, better.