Hook
A White House teleprompter operator—someone whose job is to ensure the President reads the right words—turned those words into a $100,000+ profit on Kalshi before they were spoken. The system claims to be a market for information, but the information was already owned. The code is law, but the humans are the bug.
The event is trivial in scale—a single operator betting on predictable lines from a Trump speech—but its implications are tectonic for the entire “information finance” stack. It reveals that the weakest link in a prediction market is not the smart contract, but the soul of the operator who holds the script.
Context
Prediction markets like Kalshi (CFTC-regulated central limit order book) and Polymarket (on-chain, using UMA’s dispute resolution) are designed to aggregate dispersed knowledge into a probability. Their value proposition hinges on the assumption that no participant possesses material non-public information about the outcome—or if they do, they are prevented from trading on it.
Kalshi, in particular, prides itself on being a compliant, audited venue. It requires KYC, tracks trades, and submits to CFTC oversight. Yet here we have a user—Perez—who was not only employed by the White House but worked directly on the President’s teleprompter, giving him access to exact phrases that would be used in public addresses. He bet on those very phrases. The platform’s surveillance systems did not flag him. The CFTC is now investigating, and bipartisans in the Senate are demanding the same scrutiny be applied to Polymarket.
Core
The core failure is not technical—it is trust-minimization failure at the human layer.
Kalshi’s trust model is entirely centralized around a single fact-finder: the CFTC-approved referee that decides whether an event occurred. In this case, the “referee” was never needed because the outcome was unambiguous—the President used the words Perez predicted. But the damage was done before the referee could act. The platform’s internal controls lacked the ability to correlate a user’s employment background (White House teleprompter) with the specific contracts he was actively trading (presidential speech key terms).
Polymarket faces an even deeper paradox. Its fact-resolution relies on UMA’s optimistic oracle, where token holders can challenge a proposed outcome within a dispute window. The assumption is that economic incentives will prevent falsehoods. But an insider with non-public information can execute a trade, and if the market resolves correctly (since the insider’s information is true), no one will challenge the resolution—the insider profits perfectly within the system. The oracle is designed to catch lies, not truths that were leaked.
I have audited over 400,000 lines of Curve’s governance simulation data, and one pattern repeats: decentralization is often a myth when the information asymmetry is structural. In prediction markets, the asymmetry is not about capital—it is about access to the future. Perez had access not because he was a better trader, but because he held a key to the script. No cryptographic primitive can patch that.
Contrarian
The conventional wisdom says this scandal kills trust in regulated prediction markets. I see the opposite. The fact that Perez was caught—that his trades were traced back to a White House background, that the CFTC is negotiating a settlement, that the White House quickly severed ties—proves that Kalshi’s compliance apparatus can be effective after the fact. In a fully pseudonymous, unregulated market like Polymarket, Perez could have used a fresh wallet, moved funds through a mixer, and vanished. The absence of a regulator would make the crime invisible.
The real danger is that this event will strengthen the case for mandatory surveillance—and that might destroy the very principle of permissionless prediction. If every platform must now enforce strict insider-trading policies, including real-time monitoring of user employment data, the cost of compliance becomes prohibitive for smaller protocols. The result could be a bifurcation: a few “whitelisted” markets for approved participants, and everything else left to the dark corners of the internet.
The contrarian take: this incident is a feature, not a bug, of the Kalshi model. It exposes a weakness, but the weakness is correctable through better data integration (linking government employee databases to trading accounts). Polymarket cannot make that correction without becoming a surveillance-state itself. In the void, we found our own gravity.
Takeaway
Prediction markets were supposed to democratize truth. Instead, they have become a mirror reflecting our deepest social fractures: those with privileged access to tomorrow can arbitrage the ignorance of today. The only durable solution is not more code—it is a rethinking of how we define material non-public information in the age of real-time digital speech.
Silence is the only consensus that never forks. But the market will not go silent. The question is whether we will design for the honest majority or for the few who hold the script.
