Insider Trading Is Good, Actually

Prediction Markets are the closest instantiation of a real-time truth machine, so why are we getting mad that it’s working?

In 1990, a DC-based Domino's owner named Frank Meeks noticed something odd about his business.

Every time an order of more than twenty pizzas came from one of the government departments in one night, something terrible happened the next morning.

In 1983, pizza deliveries to the Pentagon spiked the evening before the U.S. invaded Grenada. In 1989, same thing the night before Panama. On August 1st, 1990, the CIA placed a single-night record of twenty-one pizza orders. The next day, Iraq invaded Kuwait.

When Meeks reported this to CNN, its Pentagon correspondent Wolf Blitzer famously responded with, "bottom line for journalists: always monitor the pizzas."

Blitzer was joking.

But there was some truth in that.

https://www.pizzint.watch/" class="w-full rounded-lg" />

Three decades later, the Pentagon Pizza Index is no longer a joke, it's a real OSINT (open source intelligence) tool with a dedicated community, its own X account, and a track record that, if you ask me, makes most intelligence analysts look kind of silly.

In June 2025, the Pentagon pizza report tracked a huge surge in activity at pizza shops within two miles of the Pentagon at 7 PM.

By 8 PM, Israel had launched a bombing campaign against Iran.

The account also noted abnormally low traffic at a bar near the Pentagon. I guess the insinuation is that when you're monitoring a war halfway across the world, you skip happy hour.

On June 22nd, a Papa John's near the Pentagon lit up at 10:38 PM. One hour later, Trump announced strikes on Iranian nuclear facilities.

February 2026: orders spiked at Pizzato Pizza at 1:30 AM. Hours later, US strikes killed Iran’s supreme leader Ayatollah Ali Khamenei.

The Israeli Air Force, in response to the Pizza Index theory gaining traction, issued formal directives telling soldiers to stop ordering food delivery to military headquarters in Tel Aviv.

Defense Secretary Pete Hegseth joked, "I've thought about just ordering lots of pizza on random nights just to throw everybody off."

I love this.

There is something deeply humorous about the most powerful military in the world being unable to keep a secret from Domino's.

But let’s not get carried away because the humor is hiding something important underneath.

Something I think most prediction markets critics haven't fully absorbed yet.

You see, information is physical - like water. It has mass. It leaks through the people who carry it, via their Uber Eats orders, their overtime, their cancelled dinner plans, their empty bar stools. The Department of War could classify every briefing, and stamp every document ‘top secret’ but the information will still seep out through pizza.

On February 28th, 2026, a Polymarket user called "Magamyman" placed $32,000 on the possibility that the United States would strike Iran that day. The market at the time had this priced at 17%. In other words, almost nobody thought it was happening.

By 5 AM, the probability had jumped to 43%. Five other brand-new accounts, "Planktonbet," "Dicedicedice," and my personal favorite, "nothingeverhappens911" placed coordinated bets totaling $1.2M that strikes would come on the 28th specifically.

The first missile struck at 9:14 AM ET.

Magamyman cleared $431,146 across all his bets. The six accounts collectively made millions.

Senator Chris Murphy called it "insane" and "disgusting." Congress even introduced legislation banning government officials from trading on prediction markets. Consumer watchdog Public Citizen demanded a CFTC investigation.

People are furious and I get it. I think it’s genuinely grotesque that someone made almost half a million dollars because they knew bombs were about to fall somewhere. Imagine if people in Iran were able to find this out before their homes were destroyed?

Don’t hate me for being that person, but I am going to be that person and say it: the market was right.

Not right in the casino sense that the bet paid off. Right in the much more interesting sense that four hours before any news organization, any government statement, or any intelligence briefing confirmed what was happening, the prediction market knew and told you.

If you were watching Polymarket in the wee morning hours of February 28th, you knew what CNN wouldn't be able to report until after 9:15 AM.

I can’t stop but think how prediction markets do exactly what the Pentagon Pizza Index does but for everything, everywhere. The markets give private information that is locked inside the heads of a small number of people a venue and an incentive, turning it into a public signal.

In fact, I would argue that prediction markets go much further than pizza such that the pizza index only gives us a fuzzy "something is happening" signal.

Prediction markets deliver a probability, a timestamp, and most importantly, a price.

I asked Claude to come up with an analogy to describe the difference and here’s what it said: ”The difference between the pentagon pizza index and prediction markets is akin to the difference between trying to identify the location of a fire by seeing smoke in the distance vs reading a fire report with GPS coordinates.

It’s important to remember though that both do the same thing: surfacing private knowledge.

Feature, not a bug

Santiago Roel Santos wrote a blog recently arguing that “insider trading threatens the ability of prediction markets to scale” and that “institutions won't participate and retail will flee if the game feels rigged.”

This is the latest in what appears a public meltdown around insider trading on these venues.

I argue that the current conversation about insider trading in prediction markets is misplaced. Popular discourse treats insider trading as a bug, a scandal, even a growth constraint.

But hear me out:

The insiders have been there way before the prediction markets.

The same inner circle of people always know what is coming. They take advantage of that knowledge via defense contractor stock trades, through group chats with hedge fund buddies and maybe a thousand other invisible channels. There is a history of the SEC investigating clusters of suspicious pre-strike trades in defense stocks going back decades. The investigations mostly go nowhere because the trades are scattered across brokerage accounts and jurisdictions.

Now look at what happened on Polymarket.

Crypto analytics firm Bubblemaps identified six suspected insiders within hours. They verified the accounts were newly created and had no prior trading history.

Not just that, the following things happened within a few hours:

• Israeli police opened an investigation.
• Congress introduced legislation.
• CFTC asserted jurisdiction.

Think about this, insider trading on classified information has been going on for decades. But this time, there's a paper trail for regulators, journalists, and investigators to follow.

What I’m really trying to say is: prediction markets didn't create the insider.

These markets just gave them a venue, a place to come outside from the shadows they usually lurk in.

Before February 28th, people with advance knowledge of the strikes almost certainly traded on that knowledge through traditional financial channels. There was no way to find and investigate them.

On Polymarket, the same information leakage produces a public, timestamped, traceable record that triggered investigations on three continents within a week.

Magamyman’s trades may be the most visible, most traceable, most investigable insider trade in the history of financial markets. I think that’s a feature, not a bug.

It's an evidence of the resilience and significance of prediction markets that when they work, they can work so well it makes the brokenness of our systems visible for the first time.

Wrong instincts

A few months ago, Polymarket announced a partnership with Palantir, the Peter Thiel-backed AI surveillance company to build what they're calling "a next-generation sports integrity platform." Palantir’s Vergence AI engine will be used to conduct real-time trade monitoring, anomaly detection, and restricted participant screening.

The same Palantir that boasts the U.S. military and the broader global intelligence apparatus as their most prominent clients.

I understand why Polymarket feels pressured to do this but I want you to hold the irony that I am seeing for a second.

The prediction market that just demonstrated it can surface private information faster than any intelligence agency on earth hired an intelligence company to prevent private information from entering the market.

And look, I get it. Polymarket needs to show adults-in-the-room energy. Palantir is about as adult-in-the-room as it gets.

But I want to hold two ideas simultaneously and I think most people are only holding one:

One: People should not profit from advance knowledge of bombing campaigns. Financial incentives around military violence are genuinely dangerous. Markets that reward assassination or terrorism are obviously terrible.

Two: The information-surfacing function of prediction markets and their ability to compress dispersed private knowledge into a public price signal is the single most valuable thing prediction markets have produced. In fact I argue they’re even more valuable than the trading fees, or the new generation structured financial products.

The Palantir partnership is Polymarket’s response to their Pentagon Pizza Index. It's trying to stop the information from leaking. It optimizes for idea one at the expense of idea two.

Sensors, not casinos

As someone who fell in love with prediction markets due to my love for and past life experience as an investigative journalist, it bothers me to see them marketed and categorized as betting platforms. They sit in the same regulatory bucket as DraftKings. The UX is designed around placing wagers. The business model runs on volume. The policy debate is "should this be legal" and "is this gambling."

What if that's the wrong framing entirely?

What if prediction markets are sensors, not casinos. I have been regularly describing prediction markets as truth and measurement infrastructure countering the popular ‘new financial instruments’ perspective. Sure they are instruments, but of a different kind. They detect the state of the world in real time, using financial incentives as the sensing mechanism.

If this sounds like a semantic trick, just hear me out.

A casino worries about insider trading because it's unfair to other bettors. A sensor worries about insider trading because it needs to separate informational value from moral hazard and those are completely different design problems requiring completely different solutions.

A casino hires Palantir to catch cheaters. A sensor hires Palantir to understand why the price moved, then publishes the analysis and build tools to help disseminate the information as fast and as efficiently as possibly.

When we recategorize prediction markets as truth infrastructure rather than betting infrastructure, the entire design space shifts. The question stops being "how do we keep insiders out?" and becomes "how do we capture the informational value of insider knowledge while eliminating the moral hazard of insider profit?"

I don't have the full answer. But I think articulating the question clearly is itself useful because we’re currently trying to answer the wrong question.

I built something to show what I mean.

The Daily Gondor

I made a prototype of a new media news platform and I call it The Daily Gondor.

The twist is that every news story is generated by live Polymarket data.

The focus of each individual story is why markets moved, how the movements influenced real-world events and vice versa. The Daily Gondor’s mandate is to contextualize markets, market movements and trader activity within the real-world.

It's a provocation, obviously.

Because if you squint, you can see a future where the most honest newsroom in the world isn't staffed by journalists at all. It's a prediction market with an editorial layer where every claim about the future has a price, where being wrong costs real money, and where the "reporter's" incentive is accuracy rather than attention.

put the pizza on the front page

We keep building systems that try to stop information from leaking. From role-based access, to classification systems, NDAs, surveillance technology. All of this is equivalent to the Israeli Air Force banning food delivery to prevent the Pizza Index from working.

It never works. Information is physical. It leaks through human behavior.

Prediction markets are the first technology that says: fine, let it leak. In fact, pay people to leak it. Build a system where private information flows into a public price, visible to everyone, traceable by anyone.

The result is messy and morally complicated and sometimes genuinely ugly. But it's also honest. More honest than the world where the same insiders trade the same information through channels nobody can see.

The Pentagon figured out that pizza orders leak classified information. Their response was to ban the pizza.

Let’s not ban pizza, please.

I say let’s put the pizza on the front page. Let the price and movements be the headline. And let’s figure out together, how to make a system where truth-surfacing and accountability aren't in tension with each other such that the sensor works and the receipt is public and the incentives eventually point somewhere good.

I hope Kalshi and Polymarket realize this and understand the actual problem they’re working with. It’s not a compliance problem.