Every argument about prediction markets this year has been about who is allowed to run one. New York suing for billions. A dozen state gaming regulators issuing orders. Three California tribes at the Ninth Circuit. A bill to ban sports contracts outright.

Almost none of it touches the question that decides whether these things work at all: how does the market know who won?

Until this week there were two answers shipping in production. Kalshi settles with a rulebook, applied by a licensed operator with a regulator behind it. Polymarket settles with an optimistic oracle, where an outcome is proposed, challenged and, if contested, voted on by people holding a governance token.

On September 9, World shipped a third. Chainlink data feeds settle the contract automatically the moment the game ends. No panel. No vote. No dispute window. Nobody to appeal to, because the settlement is a program that already ran.

It launched to a waitlist of over a million people and took its own website down.

Worth understanding what that third model buys and what it gives up, because it is not obviously better or worse than the other two. It is differently broken, which is the only honest thing anyone can say about resolution mechanisms.

What World actually is

None of the individual pieces is novel. The combination is.

World launched inside the Phantom wallet during the summer and opened a standalone site at world.xyz on September 9, extending access to a waitlist exceeding one million users. More than 150,000 markets have been created since the Phantom integration went live, spanning sports, crypto, politics, finance, economics, and culture.

The opening lineup is broad: every NFL regular-season game, seven soccer leagues, Formula 1 races, binary contracts on the 2026 United States midterm elections, and contracts on Federal Reserve policy decisions. Equity, commodity, and weather markets are planned.

Mechanically, each market issues yes and no contracts priced between zero and one dollar, and the verified outcome settles one side at a dollar. That is the standard binary event contract structure our guide to the instrument covers.

Three structural properties distinguish it. It is non-custodial: the protocol holds no customer funds, and assets move only when a user enters a market. Orders route to liquidity providers on Solana, not through an off-chain order book. And settlement is in CASH, the dollar-backed stablecoin used inside Phantom, with winning positions redeemed automatically in the wallet.

No brokerage account is required. No exchange registration is required. Users pay network fees to open and close positions.

The three resolution models

Nobody has put the three side by side, so here they are.

Kalshi: rulebook resolution by a regulated operator. Contracts settle according to criteria published in advance, applied by an exchange holding designated contract market status. A named entity makes the determination, a federal regulator supervises it, and participants have a complaints path. Our guide to that licence sets out the obligations. The operator can correct errors, and the operator exercises discretion, which are the same property viewed from two angles.

Polymarket: optimistic oracle with token-holder voting. An outcome is proposed, a challenge window opens, disputes escalate to a vote by holders of the oracle’s governance token, and the result finalises on chain. Our guide to that mechanism explains the stages. Nobody can unilaterally decide an outcome, and nobody can correct one after finality, which is again the same property from two angles.

World: automated data feeds with no dispute stage. Chainlink Data Streams supply the market data and the Chainlink Runtime Environment executes settlement once a game concludes or a defined event reaches its deadline. The published description is explicit that there is no human panel, no token-holder vote, and no dispute window delaying payouts.

The tradeoff runs consistently across all three and is worth stating as a principle. Every mechanism that removes discretion also removes correction. Kalshi can fix a mistake and can also make a discretionary call you dislike. Polymarket cannot make an arbitrary call and cannot fix one either, which is precisely the tension our guide to delisted and voided markets examines. World removes the most discretion of the three and therefore removes the most correction.

What automated settlement is good at

Inside a specific range this design is clearly better than the alternatives, and the case deserves its strongest form.

Objective outcomes settle instantly. A football match ends with a score. A Bitcoin price at a stated timestamp is a number. A Federal Reserve rate decision is a published figure. For contracts resolving on unambiguous, machine-readable data, a dispute window is pure latency, delaying payout to accommodate an argument nobody will make. Automated resolution pays immediately, which is a real user benefit and the clearest competitive advantage the design has.

It removes the failure mode that has damaged the category most. Contested resolutions on the oracle-based venue have produced the most reputational damage prediction markets have suffered, because a market resolving against what most observers believed happened is the single thing that destroys confidence in a forecasting instrument. Removing the discretionary stage removes that possibility for contracts where the data is unambiguous.

And it scales. A resolution process requiring human attention constrains how many markets can exist. More than 150,000 markets created since the summer is a number that would be operationally impossible under rulebook administration, and it is achievable precisely because resolution costs nothing per market.

Polymarket has moved in the same direction for price-based markets, adopting oracle-based settlement where the data permits, which is confirmation that the design is correct for that category, not a criticism of World for adopting it.

What automated settlement is bad at

The limits are just as specific, and they sit in exactly the markets people care about most.

Ambiguous events have no data feed. A contract on whether an official will resign, whether a conflict qualifies as a ceasefire, or whether a statement constitutes an endorsement cannot be settled by a price feed, because the disputed element is the definition and not the measurement. These are also disproportionately the markets people care about, which is why the oracle-based venue’s most contested resolutions have involved exactly this category.

Data feed failure has no remedy. If a feed reports incorrectly, reports late, or reports a value that does not reflect what happened, an automated system settles on it. With no dispute window there is no stage at which anyone can say the input was wrong before money moves. The integrity of the entire system rests on the integrity of the data source, and the participant has no mechanism to contest it.

Edge cases resolve mechanically. A postponed match, an abandoned race, a rescheduled announcement, a data source that stops publishing. Rulebooks handle these with voiding provisions. An automated system handles them according to whatever the contract specified in advance, and contracts cannot anticipate everything.

And there is nobody to appeal to. This is the practical consequence of the architecture. A participant who believes a market settled wrongly on Kalshi can complain to an exchange and to a regulator. On the oracle-based venue they can, in principle, participate in a dispute. On World, the settlement is the output of a program that already ran.

None of this makes the design wrong. It makes it correct for a specific class of contract and unsuitable for another, and the honest question is which class dominates the platform’s 150,000 markets.

The regulatory position

Here is the part nobody covering the launch has touched.

World lists contracts on every NFL regular-season game, the 2026 United States midterm elections, and Federal Reserve policy decisions. Those are precisely the contract categories currently under attack in the United States. Our status page on the sector maps the fights: state gaming regulators contending that sports event contracts are wagers requiring state licensing, California tribes litigating under federal Indian gaming law, and a bipartisan bill that would prohibit sports contracts on regulated exchanges outright.

The venues fighting those battles hold federal licences. Kalshi is a designated contract market. Polymarket operates domestically through an exchange it acquired. Both submitted to registration, and both are being sued anyway.

World requires no brokerage account and no exchange registration, holds no customer funds, and routes orders to liquidity providers on a public blockchain. That is a structurally different posture, and it raises the question the coverage has not: what happens when a non-custodial protocol lists the same contracts the licensed venues are being sued over.

Two readings are available. The optimistic one is that a non-custodial protocol with no operator holding funds is genuinely outside the frameworks being applied to exchanges, which are built around intermediaries. The sceptical one is that the same argument was made by offshore venues before 2022 and produced a settlement and a geoblock, and that regulators reach operators of protocols when they can find them.

Nothing in the launch materials addresses geographic restriction, and that absence is the most significant unexamined fact about this launch.

Why Solana, and why now

Solana is not an accident and neither is the timing.

Solana spent the year as the dominant venue for memecoin activity, and the network has been pushing into prediction markets as the next consumer application category. The Solana Foundation’s head of decentralised finance framed World as introducing a new asset class while keeping liquidity fully on chain, which is the strategic pitch: a network that captured speculative trading volume wants the next category of speculative trading volume.

Phantom’s role is the distribution mechanism. It is among the most used wallets on Solana, World operated inside it before going standalone, and settlement in CASH ties the product tightly to Phantom’s ecosystem. A million-person waitlist is what wallet-native distribution produces, and it is a channel neither licensed competitor has.

Chainlink’s position is the infrastructure play. The same Data Streams and Runtime Environment combination was adopted by another prediction market earlier in the year for automated creation, resolution, and settlement of crypto price markets, which suggests a standardising pattern instead of a bespoke integration.

So the launch is the intersection of three strategies: a network seeking its next consumer category, a wallet monetising distribution, and an oracle provider becoming the settlement layer for a market type. None of those three is primarily a bet on prediction markets being legal in the United States.

The market that breaks it

Pick a real example and the limits stop being theoretical.

Take a contract on whether a ceasefire holds. The data feed needs a number, and there is no number. Somebody has to decide what counts as a violation, whether a single incident breaks it, whether a disputed report is credible. Kalshi’s rulebook answers this in advance, badly or well, and a person applies it. Polymarket’s oracle answers it through a challenge and a vote, slowly and sometimes contentiously. An automated feed cannot answer it at all, because the thing in dispute is the definition and a feed only measures.

Now take a contract on whether an official resigns by a date. Clean, until the official announces an intention to resign effective later, or is removed, or resigns and then withdraws it. Every one of those has happened in politics and each produces a different answer depending on wording nobody wrote carefully enough.

These are not edge cases. They are the markets that make prediction markets interesting, and they are also the markets that have generated every reputational disaster the category has suffered. The oracle-based venue’s worst moments have all involved exactly this kind of contract.

So World has two options with its 150,000 markets. Either the contested-definition contracts are a small fraction of the book, in which case the design fits and the reputational risk sits mostly with the competitors. Or they are not, and the first genuinely disputed settlement arrives with no mechanism at all for handling it, which is worse than either alternative, not better.

Which one is true is checkable from the market list, and it is the single most useful piece of due diligence available on this platform.

What the incumbents should be worried about

Not the technology. The distribution.

Kalshi and Polymarket have both spent heavily on advertising, sports partnerships and corporate deals to acquire users. That is the normal cost of building a consumer financial product and it is enormous.

World got a million-person waitlist by existing inside a wallet people already had open. Phantom is among the most used wallets on Solana, World ran inside it before going standalone, and settlement happens in Phantom’s own stablecoin with winning positions landing back in the wallet automatically. No app to download, no account to open, no deposit to make, no identity check.

That is a distribution channel neither licensed competitor can replicate, and it did not cost a marketing budget. It cost an integration.

The uncomfortable part for the incumbents is that the thing making World’s distribution cheap is the same thing making its regulatory position ambiguous. No brokerage account means no onboarding friction and no registered intermediary. No identity check means faster signup and no way to screen prohibited participants, which is the entire surveillance apparatus the licensed venues built at considerable expense after the insider trading scandals.

So the competitive question is not whether automated settlement beats rulebook settlement. It is whether a product that skips registration, custody and identity can out-distribute products that did not skip them, and whether regulators reach it before the answer becomes obvious.

The category has run this experiment before, offshore, and it ended in a settlement and a geoblock. What is different this time is that the thing being regulated is a protocol on a public chain rather than a company with a bank account, and nobody has tested whether the old tools reach the new structure.

Chainlink is the real winner here

Follow the infrastructure and a different story appears.

World runs on Chainlink Data Streams and the Chainlink Runtime Environment. Earlier this year another prediction market adopted the same combination to automate creation, resolution and settlement of crypto price markets, with stated plans to extend into stocks, commodities and other real-world assets. That is two venues on one stack in a single year, which is how a standard forms.

The position is worth understanding. An oracle provider that supplies price feeds is a utility, paid per call, substitutable if someone builds a cheaper one. An oracle provider that supplies settlement for an entire market category is something else: it becomes the arbiter of outcomes for every contract built on it, and switching costs rise with every market that depends on its determinations.

Polymarket runs its own oracle mechanism with a governance token attached. Kalshi has an exchange rulebook and a regulator. Both built their resolution layer in-house because resolution is the product. World rented it, which is faster and cheaper and means the most consequential function in the business belongs to someone else.

There is a market-structure question buried in that which nobody has asked. If prediction markets standardise on one settlement provider, the failure mode of the entire category becomes correlated. A rulebook venue and an oracle venue fail independently, because their resolution mechanisms have nothing in common. Two venues on the same data infrastructure do not.

None of which is a criticism of the technology, which by all accounts works. It is an observation about concentration, and it is the kind of thing that looks like efficiency right up until the moment it looks like systemic risk.

What this means if you are actually trading on it

Practical, because the architecture changes what you should check before entering a position.

Read the resolution source, not the market title. On an automated venue this matters more than anywhere else, because there is no stage at which a human reconciles the title with the feed. If the market says one thing and the data source measures something slightly different, the data source wins and nobody reviews it.

Understand that settlement is final the moment it happens. No challenge window means no window. On a rulebook venue you can complain. On an oracle venue you can, in principle, dispute. Here the transaction has cleared and the funds have moved before anyone has formed an opinion about whether it was right.

Check what happens to postponed and abandoned events. Sports contracts are the bulk of the book and postponements are routine. A rulebook handles this with voiding provisions written by people who have seen it happen. An automated system does whatever the contract specified in advance, and you want to know what that is before a rain delay decides your position.

Size for the absence of recourse. This is the practical version of everything above. On a venue with no operator to appeal to and no regulator supervising the outcome, the correct position size is smaller than on a venue that has both. That is not a criticism of the design, it is what the design implies.

And know which entity you are dealing with. The protocol is non-custodial and holds no funds, which is good for counterparty risk. It also means there is no counterparty, and no counterparty means nobody to make you whole if something goes wrong that is not covered by the code.

The trade-off is the same one that runs through all of decentralised finance. You give up recourse and you get access, speed and no permission required. Whether that is a good trade depends entirely on how much you were going to need the recourse, and most people find out the answer at the worst possible time.

What to watch

The composition of the 150,000 markets. How many resolve on unambiguous machine-readable data and how many on contested definitions. That ratio determines whether the resolution model fits the product.

The first contested settlement. Every prediction market eventually produces a resolution a large number of participants believe is wrong. With no dispute window, what happens next is the question the architecture has not been tested on.

Whether geographic restriction appears. Nothing in the launch materials addresses it, and the contracts listed are the ones under active litigation in the United States.

Whether the licensed venues adopt the same model. Polymarket has already moved toward oracle-based settlement for price markets. If rulebook resolution retreats to only the contracts that require judgment, the category will have converged on a hybrid.

Volume against markets created. More than 150,000 markets is a supply figure. How much of it trades is the demand figure, and only the second one matters.

Disclaimer: This article is for information and educational purposes only and does not constitute financial, investment, or legal advice. The legal status of event contracts varies by jurisdiction and is subject to active litigation, and availability of any platform depends on where you are. Nothing here is a recommendation to use any service. Information is accurate as of September 10, 2026.