In May 2026, federal regulators charged a Google employee with insider trading on Polymarket. Michele Spagnuolo — known on the platform as AlphaRaccoon, according to court filings — allegedly had access to Google's internal search trend data. According to charging documents, he used it to bet on prediction markets priced on those same trends. Prosecutors say he won virtually every position he took. Prosecutors allege he made over $1 million.
The CFTC called it fraud.
But buried inside the AlphaRaccoon story is a question nobody asked: what would the legal version of that look like?
I spent the last week building it.
The Information Advantage Problem
Regulators allege he had access to information the market didn't. Internal Google data is non-public. Using non-public information to trade is generally prohibited under commodities law — which is precisely what regulators allege occurred here.
But here's what's interesting: the market was wrong not because it lacked sophisticated participants. It was wrong because it lacked connected participants. People who could see across multiple public data streams simultaneously and recognize when they all pointed in the same direction.
That connection is legal. It's just hard to build.
The SEC requires corporate insiders to disclose stock purchases within two business days. The filings are public — Form 4, available on EDGAR, updated every ten minutes. Congress members must disclose their stock trades within 45 days under the STOCK Act. Those filings are public too, buried on the House Clerk's website. Prediction markets publish their prices in real time via public APIs.
Three separate streams of public information. Nobody connecting them automatically.
Until now.
What I Built
Over the course of one week, working from a $20/month DreamHost VPS in Grand Forks, North Dakota, I built a compound signal engine that monitors all three streams simultaneously.
The EDGAR monitor pulls Form 4 filings every ten minutes, scores them by insider seniority, purchase size, and pattern of repeated buying, and fires an alert when the score exceeds a threshold. A CEO buying $5 million of his own company's stock scores 165. A Co-Chairman buying $5.76 million across six transactions in a single day scores 590.
The congressional monitor scrapes STOCK Act disclosures from the House Clerk's website, parses them from PDF format, and cross-references them against a database of committee assignments. When a member of the Armed Services Committee buys a defense contractor, that's a signal.
The compound signal engine runs every time new EDGAR data arrives — five times a day on weekdays. It looks for one thing: tickers where both corporate insiders and members of Congress are buying simultaneously.
When it finds one, it fires an SMS to my phone.
Update: The Legal Reckoning
Since the original AlphaRaccoon story broke in December 2025, the case has escalated significantly. In May 2026, both the CFTC and the New York Department of Justice filed criminal and civil charges against Michele Spagnuolo, identified in court filings as the operator of the AlphaRaccoon account. Prosecutors say he went 22 for 23 on his Google Year in Search bets — including correctly picking d4vd, a relatively obscure artist, to top the global rankings at what prosecutors described as shockingly unlikely odds. Federal prosecutors allege he made over $1 million using nonpublic Google search data.
The case has broader implications. Intercontinental Exchange invested $2 billion in Polymarket, valuing the platform at $8 billion, partly on the premise that prediction markets provide reliable, manipulation-resistant price discovery. As Forbes noted, institutional clients have zero tolerance for markets they perceive as rigged. If insiders can systematically exploit prediction markets using privileged data, the informational value of those markets collapses entirely.
This is exactly why the legal version matters. The compound signal engine described below doesn't require access to anyone's internal servers. It doesn't disappear when someone gets fired or charged. It's built on SEC filings and congressional disclosures that exist precisely because the public has a right to know. The edge isn't stolen — it's earned, by connecting data that's been public all along.
The First Real Signal
After loading 23,209 historical Form 4 filings from Q3 and Q4 2025, the compound engine fired 28 signals on its first real run.
Top signal: TSLA. Compound score 3,555. $4 billion in total insider buying over six months, held simultaneously by multiple members of Congress.
That's not inside information. That's public information, connected.
The difference between what regulators allege AlphaRaccoon did and what this system does is simple: he allegedly used data that wasn't his to use. This system uses data the SEC and Congress are legally required to make public. The edge comes not from knowing something secret, but from processing public information faster and more systematically than the market has.
The Prediction Market Layer
Here's where it gets interesting for 2028.
Prediction markets — Polymarket, Kalshi — price political and regulatory outcomes in real time. When a defense contract award is being debated in committee, there's often a Kalshi market for it. When a pharmaceutical approval is pending, there's a Polymarket contract.
If corporate insiders and the relevant committee members are both buying the affected company's stock while the prediction market is pricing the regulatory outcome at 30 cents — that's the gap. That's where the edge lives.
Not because anyone broke the law. Because three public data streams haven't been connected yet.
After analyzing 28,798 resolved Polymarket markets totaling $7.9 billion in volume, one thing is clear: the crowd consistently misprices outcomes that sophisticated, connected observers would price differently. 74.4% of all markets resolve NO. The crowd is wrong about the future nearly three times out of four. We explore this in depth in Research Note #003: The Mirror We Built Without Knowing It.
The compound signal is the legal, systematic version of having better information. It doesn't require access to Google's servers. It requires a Python script and a willingness to read public filings.
What This Means for 2028
The 2028 presidential election will generate an estimated $10 to $50 billion in prediction market volume. That's the largest collective bet in human history. And the infrastructure to analyze it — the congressional trading data, the EDGAR filings, the prediction market APIs — is already public.
The AlphaRaccoon case focused public attention on information advantages in prediction markets. The question it raised is real: who has better information, and what are they doing with it?
The answer, it turns out, is hiding in plain sight. In SEC filings. In congressional disclosures. In public APIs updated every ten minutes.
You just have to connect the dots.
The Infrastructure
The compound signal engine is live at api.greenwood.financial. The full research is documented in Collective Conviction: What $8 Billion in Prediction Markets Reveals About American Democracy, available for pre-order on Amazon for October 6, 2026.
The data is public. The edge is real. The only question is who builds the infrastructure first.
Want to go deeper?
Get the free research note — insider signals, congressional trades, and prediction market data from the Greenwood platform.