Alpha Raccoon Had a 22-for-23 Record. That Was the Confession.

Alpha Raccoon

9:17

Michele Spagnuolo is 36, Italian, lives in Switzerland, and works on Google's security team. That is, as someone noted online, an excellent dating profile. It is also, as it turns out, the beginning of a federal rap sheet.

In late 2024, Spagnuolo allegedly logged into an internal Google tool — one available to all 180,000 employees — that tracked what people all over the world were searching for in real time. He wanted to know who would top Google's annual Year in Search rankings. Then he opened a Polymarket account under the username Alpha Raccoon, found a bet on Google's most-searched person of 2025, noticed that a singer named David was listed at near-zero probability, and placed his wager. It wasn't a risky bet. He already knew the answer. He waited for Google to announce the results in December, collected $1.2 million, deleted the Alpha Raccoon account, moved the crypto to a different wallet, and presumably went back to work on Google's security team — the irony of which cannot be overstated. He did not get away with it.

How You Catch a Raccoon

The FBI didn't catch Alpha Raccoon. The internet did. Polymarket's own community account tweeted on December 4th: "$150,000,000 profit in 24 hours trading Google Search markets. Who is Alpha Raccoon?" Traders were openly following his bets. When someone asked what to put money on, other users replied: check Alpha Raccoon's account. Random people on the internet had started mapping the pattern before any federal agency got involved. Alpha Raccoon went 22 for 23 on Google-related bets. He placed 16 wagers over two months, risked $2.7 million total, and won with a frequency that is statistically incompatible with luck. A 22-for-23 record on highly specific, low-probability outcomes isn't a hot streak. It's a confession.

A Meta engineer flagged the wallet publicly on social media. The community called it. The FBI caught up later. Spagnuolo now faces charges of commodities fraud, wire fraud, and money laundering. His bond is $250,000. Google put him on administrative leave and said it cooperated with investigators. He is unlikely to be working on security from a federal prison.

This Is Not an Isolated Case

In April, a U.S. Army Special Forces master sergeant named Gannon Ken Van Dyke was charged with using classified intelligence about the timing of the military operation that captured Nicolás Maduro to place bets on Polymarket. He made over $400,000. The charges: wire fraud, commodities fraud, misuse of non-public government information. Two federal insider trading cases on prediction markets in the same year. Two very different institutional contexts — a tech campus in Zurich, a special forces unit with access to classified intelligence. Same basic logic: someone with asymmetric information realized they could convert that information into money with minimal friction, and did.

The cautionary tale framing is tempting. But the more honest read is that these cases are not aberrations. They're early signals of a structural problem that nobody has solved yet.

The System Nobody Built

Here is what's notable about the Spagnuolo case beyond the personal hubris: Google said the tool he used wasn't a secret. It wasn't some classified internal system accessible only to senior leadership. It was available to every employee in the company. All 180,000 of them.

Which means the asset being monetized wasn't a security breach. It was a standard employee benefit — access to internal data — being arbitraged against a platform with no mechanism to detect or prevent exactly this kind of trade. Polymarket doesn't ask whether you work at the company you're betting on. There's no conflict of interest disclosure. No verification layer. No system that checks whether the person placing a near-zero probability bet on a Google search outcome happens to have a Google employee badge. You create an account, pick a username, and start trading.

The Senate has already banned its own members from prediction market platforms. Arizona has charged 20 people over related betting activity. But those are enforcement actions against individuals. They don't address the structural gap. The structural gap is this: prediction markets have scaled to roughly $1 billion in daily trading volume and nearly $24 billion in monthly volume as of April 2026. They are now large enough to be financially material — large enough that a single well-placed bet can generate seven figures in a single event. And the regulatory framework governing them still treats information asymmetry the way the old insider trading rulebook treats it: as a problem to be prosecuted after the fact, not prevented at the point of entry.

There is no mechanism on the platform side that would flag a pattern like Alpha Raccoon's before it became a $1.2 million payout. There is no disclosure requirement that would have surfaced Spagnuolo's employment relationship to the market he was trading. There is no system at Google — or most companies — that monitors whether employees are placing bets on outcomes their internal access would materially inform. The incentive structure exists. The tooling to exploit it exists. The guardrails don't.

What Prediction Markets Actually Are

Polymarket and platforms like it are fundamentally information markets. The price of any outcome at any moment represents the aggregate belief of all participants about the probability of that outcome. When a bet trades at near-zero, it means the market collectively believes the event is nearly impossible. That collective belief is only as good as the information available to the participants.

When one participant has access to information that would resolve the uncertainty entirely — when they're not estimating probability but reading an answer from an internal dashboard — they're not trading against the market. They're extracting rent from it. Everyone else in that market is providing liquidity to someone who is, in effect, robbing them.

That's not a niche edge case. It's the core vulnerability of any prediction market where outcomes are knowable in advance by someone with institutional access. And the more prediction markets expand into domains that touch real organizations — corporate events, government actions, military operations, policy decisions — the larger the pool of people with potentially decisive insider knowledge becomes.

Soldiers know about raids. Engineers know about product launches. Analysts know about earnings. Executives know about mergers. The list of people who work adjacent to knowable outcomes is very long.

The Enduring Human Talent for Finding Loopholes

There's a version of this story that ends with tighter regulations and a few more prosecutions, and things proceed roughly as before. That version is probably what happens in the short term. The more interesting version is that prediction markets represent something genuinely new: the first mass-scale infrastructure for monetizing information asymmetry that is accessible to ordinary people. You used to be a Google employee sitting on trend data, and the worst thing you could do with it was mention something interesting at dinner. The information had no direct exchange value outside your employment context.

Now it does. The exchange mechanism exists, it's liquid, and it's global. The payout for being right about something almost nobody else knows can be enormous. And the compliance infrastructure — inside companies, inside platforms, inside regulatory agencies — was built for a world where that exchange mechanism didn't exist. Alpha Raccoon understood this. So did the master sergeant who bet on a classified military operation. They were, in their way, perfectly rational actors operating in a system that hadn't yet priced in their rationality.

The raccoon got smoked out this time. But 180,000 Google employees still have access to that tool. Prediction markets are still growing. And the gap between the information asymmetries that exist in the world and the systems designed to govern them has never been larger.

Someone will find the loophole again. Probably with a better username.

---

Rich Washburn is a technologist and strategist working at the intersection of AI, cybersecurity, and capital. He is Managing Partner and Chief AI Officer at Eliakim Capital, and CIO of Data Power Supply.