Kalshi bans athletes from trading their own games. Nobody has written the rule for a company that can make it rain.

In August 2026 the CFTC fined former congressman George Santos $17,500 and made him disgorge $17,569.98 in profits, with a three-year trading ban, over Kalshi contracts tied to the 2026 State of the Union. Small money, ordinary enforcement, and a completely standard theory of the case: someone positioned to influence an outcome took a position on it.

Every event market has this rule. Kalshi bars athletes, coaches and officials from trading their own sports, and bars political candidates from trading their own campaigns. The CFTC's prediction markets advisory puts an "independent duty" on the platforms themselves to run surveillance and enforce it.

The rule works because of an assumption nobody has had to state. The set of people who can move a given outcome is small, nameable in advance, and leaves a record when they act. A quarterback is on the roster. A candidate is on the ballot. Santos gave a speech and placed trades, and both are documents.

Weather has always sat outside that assumption for the simplest possible reason.

The assumption has a funded counterexample now

Rainmaker Technology Corporation was founded in 2023 in El Segundo, California, by Augustus Doricko, who took a Thiel Fellowship in early 2024. It seeds clouds with silver iodide from drones to increase precipitation over a target area, and sells that to water districts and states. It raised $6.3 million in seed funding in May 2024 and a $25 million Series A in May 2025 led by Lowercarbon Capital, Starship Ventures and Long Journey Ventures.

Then in April 2026 the company announced that it had become the first private operator to measure and validate the output of a cloud-seeding operation using satellite remote sensing — 143 million gallons delivered for clients in Oregon and Utah.

Read that announcement as what it is: a company press release, not a peer-reviewed result. But read the claim itself carefully, because the interesting part is not "we made rain." It is "we can now prove which rain was ours." That is the thing that had never been done, and it is the thing this entire question turns on.

What the science actually supports, stated plainly

The honest summary is narrower than either the marketing or the panic.

The SNOWIE project — Seeded and Natural Orographic Wintertime Clouds — produced the first clean quantitative evidence that seeding generates snowfall, tracking ice crystals from injection to ground. Estimates put seasonal precipitation gains from a well-run program at roughly 10%. Before that, a 2003 National Research Council review concluded there was still no convincing proof that intentional weather modification worked at all.

The limits are the whole story. Seeding cannot make a cloud; it needs supercooled clouds that are already there, under the right winds. It shifts a seasonal total in a basin. And natural precipitation is so variable that pulling the seeded signal out of the noise is genuinely hard — which is why randomised trials produced mixed results for fifty years, and why validating a single operation was news in 2026.

Nobody can raise the high temperature at LaGuardia by two degrees next Tuesday. That is not a capability that exists, and anyone telling you otherwise is selling something — in either direction. Hold onto that, because the structural problem does not need it.

The market settles where the intervention is invisible

Kalshi's weather contracts settle on a specific number from a specific instrument. Daily high and low temperature markets resolve against the final climate report from the National Weather Service for a named station, published the next morning. Precipitation, snowfall and severe-weather contracts work the same way — each tied to a designated NOAA source.

That precision is a feature. It is what makes the contract adversarially defensible: one station, one day, one reading, no argument.

It is also exactly the resolution at which a seeding operation cannot be seen.

Those two boxes are the argument. A daily station contract is blind to a seasonal basin-scale effect, and a seasonal basin-scale effect cannot be aimed at a daily station contract. Today that gap is protective — it is the reason this is a structural observation and not an allegation.

But notice what closing the gap requires. It does not require seeding to get more powerful. It requires detection to get more precise — which is the exact thing Rainmaker just announced it had achieved, and the exact thing every operator in the industry is racing toward, for entirely legitimate commercial reasons. You cannot sell water you cannot prove you delivered.

The enforcement model breaks before the capability arrives

Santos was catchable because his influence and his positions both left documents. Kalshi can enforce its athlete rule because "played in that game" is a fact you can look up. Every participant ban in every event market rests on an observable predicate: was this person in a position to move it.

Now write that rule for weather.

You would need to establish that an operation occurred, that it affected a specific station, on a specific day, by an amount that crossed a specific threshold — against a counterfactual sky that does not exist and cannot be sampled. The measurement problem that has dogged this field since 1946 is not a scientific inconvenience here. It is the reason the rule is unwriteable.

And the failure runs symmetrically, which is the part most people miss. Over the 2025 Fourth of July weekend, flash floods in central Texas killed more than 100 people. Rainmaker had seeded two small clouds near Runge, Texas, on July 2 — about 150 miles from Kerr County, where the worst flooding hit. Meteorologists at Texas A&M and UT said there was zero evidence of any connection, and there wasn't.

It did not matter. The company became the center of a national conspiracy, and Texas passed SB 1154 restricting government weather modification, effective September 1, 2025.

You cannot prove you made it rain. You also cannot prove you didn't. An industry that cannot be convicted also cannot be exonerated, and a market that cannot detect interference also cannot certify its absence.

How big is the exposure, honestly

Small, today. That belongs in the argument, not in a footnote.

The larger money is in the older market. CME Group's weather futures and options are traded largely as blocks through ClearPort, and CME's own account of the market names hedge funds and insurance companies as the typical risk-taking counterparties to utilities and agricultural producers — with volumes reported up 23% in 2025 as more speculative capital entered. That market is decades old, deeper than the event exchanges, and settles on the same public instruments.

Which points at the thing worth watching. The retail-facing daily temperature contract is the visible surface. The exposure that would actually matter is a seasonal degree-day position — sized in real money, held by a fund, settling on a regional aggregate over a whole season.

That is the resolution seeding operates at.

What would make this wrong

Seeding may simply never get aimable. It needs pre-existing supercooled clouds, the right winds, the right terrain. If the effect stays a 10% seasonal nudge over a basin, it can never be pointed at a contract, and this stays a structural curiosity.

The validation claim is a press release. Rainmaker's satellite-attribution result has not been through peer review, and the field's history is littered with confident measurements that did not replicate. If it doesn't hold up, the detection side of the gap is not closing at all.

Nobody has alleged this. There is no enforcement action, no complaint, and no reporting suggesting a weather-modification operator has traded weather contracts. This piece describes a gap in a rulebook, not a thing that happened, and treating it as more than that is how the Texas conspiracy started.

And prohibition may outrun capability. States are already legislating against weather modification for reasons that have nothing to do with markets. A ban is a blunt answer to a measurement problem, but it does resolve it.

What we take from this

Every market has an oracle. Somewhere at the bottom of any contract is a source of truth — a station reading, an API, a report — and the integrity of the whole structure rests on one question: can anyone with a position also touch the source?

That question is usually easy, which is why it is usually unasked. It gets hard exactly when a new capability makes the source reachable by someone who was never on the list of people to watch.

We build systems that settle things — what counts as a delivered send, an attributed lead, a closed ticket, a paid invoice. Every one of those numbers is an oracle for somebody's money, and the useful discipline is not accuracy. It is asking who can reach the instrument, and whether you would be able to tell if they had.

Most of the time the answer is fine. The failure is never in the measurement. It is in never having asked.

If you want to know what your numbers actually rest on, come talk to us.