The Feedback Loop
When prediction market prices enter the information infrastructure, they stop being measurements and start being interventions.
In December 2025, CNN's Harry Enten reported that Democrats' chances of retaking the House had fallen from 83 percent to 63 percent. The chyron read: “A TWIST IN THE BATTLE FOR HOUSE CONTROL.” The graphics were identical to a polling segment. The tone of authority was identical. The audience received what looked like an objective reading of political reality.
The numbers were Kalshi betting prices.
No clear disclosure. No caveat that this was a thinly traded derivatives market, not an aggregation of survey data. Democratic donors saw falling odds. Republican organizers saw rising odds. Media outlets cited the price movement as evidence of “momentum,” generating additional coverage that further depressed Democratic odds. The price did not discover a truth about the House race. It generated a narrative that became partially self-fulfilling.
This is not an edge case. In the span of three months, prediction market prices were embedded in every major layer of the American information stack. The feedback loop that the inaugural issue of this newsletter described as a structural feature of prediction markets has now been scaled to mass communication. It is no longer a theoretical concern. It is the defining feature of 2026 information markets.
The Integration Map
Between November 2025 and January 2026:
Google Search and Google Finance began displaying Kalshi and Polymarket probabilities in response to natural language queries. Google handles over 8.5 billion search queries per day. When someone types “Will the U.S. enter a recession?” and sees a market-implied probability alongside a time-series chart, the price has been encoded as fact by the most trusted information intermediary on earth.
CNN signed Kalshi as its exclusive prediction market partner in December 2025. Kalshi data runs across television, digital, and social programming. Money changed hands. Enten presents betting odds in the same segment format as polling analysis.
CNBC struck a multi-year exclusive with Kalshi days later. An on-screen Kalshi-branded ticker displays market-implied probabilities on Squawk Box and Fast Money. Kalshi hosts a CNBC-branded page on its platform where viewers trade directly on questions highlighted in coverage. The editorial-to-exchange pipeline is seamless.
Dow Jones — parent of the Wall Street Journal, Barron's, MarketWatch, and Investor's Business Daily — announced an exclusive partnership with Polymarket in January 2026. Dedicated data modules on homepages. An earnings calendar based on “market-implied expectations.” CEO Almar Latour framed it as “a rapidly growing source of real-time insight into collective beliefs about future events.”
Yahoo Finance tapped Polymarket as its exclusive prediction market provider in November 2025.
The Golden Globes on CBS in January 2026 featured Polymarket as its prediction market partner — the first major awards show to formally integrate a prediction market into a live broadcast. Hosts checked odds in real time. Announcers promoted betting throughout the three-hour show. Roughly $2.5 million was wagered across 30 categories.
Google Search, CNN, CNBC, the Wall Street Journal, Yahoo Finance. This is the information infrastructure through which Americans receive data about politics, economics, and markets. Prediction market prices are no longer one input among many. They have been embedded in the architecture of how reality is described.
The Mechanism, Quantified
The feedback loop has a structural asymmetry that makes it dangerous.
Kalshi's daily volume runs roughly $135 million — approximately 0.02 percent of daily U.S. equity volume. These are thin markets. Thin markets are movable. A few million dollars of directed capital can shift a political contract by 10 to 15 points. This is not speculation; during the 2024 election, single large trades on Polymarket moved presidential contracts by multiple percentage points within hours, and those movements were reported as news by outlets that had not yet formalized their partnerships.
Now the partnerships are formalized. The price is formed in a thin market by a self-selected group of participants. It is then broadcast by CNN, CNBC, the Wall Street Journal, and Google to hundreds of millions of people, presented in the same visual format as polling data and economic indicators. The audience — voters, donors, investors, policymakers — adjusts behavior. That behavioral adjustment feeds back into the market price. The cycle repeats. Each iteration amplifies.
The cost structure makes this especially potent. Moving a prediction market price is cheap relative to the distribution it receives. A wealthy actor places a large bet predicting a candidate will win. The price shifts. CNN broadcasts the shift as analysis — that is what the partnership pays for. The cost per impression of narrative manipulation through this channel is orders of magnitude lower than traditional political advertising. The price becomes the intervention.
This is the reflexive dynamic from the inaugural issue, running through the broadest communication channels ever constructed. The feedback loop between prices and outcomes that Soros identified in currency markets and sovereign debt now operates at mass scale. When Google puts a probability in a search result, it is not reporting a finding. It is distributing an input that reshapes the event it describes.
The Institutional Silence and Its Exceptions
The major media organizations that signed these deals have published no internal analysis of the feedback effects. The platforms that sell the data have no incentive to question its authority. The venture investors who fund the platforms need the prices to be treated as truth, because truth machines justify the $20 billion in combined valuations.
A few voices have named the problem.
Slate, December 2025: “The potential for abuse isn't the only reason it's foolish to present prediction market odds as some sort of real political science.” Their argument — that election gambling is not a proven predictive tool and that presenting it in the visual language of polling is misleading regardless of manipulation risk — identifies the epistemological problem that precedes the integrity problem.
The Intercept, late December: “Most news organizations did not include prediction market odds in their coverage prior to striking what are likely lucrative deals.” The article documented the contract spectrum's outer edge — markets on deportation numbers, markets on famine — which exist because platforms need contract volume and media partnerships create incentive to present the prices as authoritative.
Popular Information (Judd Legum) named the asymmetric cost structure directly: moving a prediction market is cheap. Having CNN broadcast the movement as analysis is free. The combined cost of narrative manipulation through this channel undercuts any alternative.
Axios, February 2026, reported that both Polymarket and Kalshi distribute “affiliate badges” to social media influencers, including accounts that spread false or satirical claims under the veneer of market credibility. The amplification pipeline does not distinguish between signal and noise.
These critiques share a structural observation: the partnership economics ensure amplification runs in one direction. An industry-funded lobbying coalition (the Coalition for Prediction Markets, launched December 2025) pushes for treating prices as authoritative. Paid media integrations push for treating prices as authoritative. Platform-sponsored PR campaigns push for treating prices as authoritative. There is no equivalent institutional force pushing for skepticism. No partnership pays CNN to caveat the numbers.
The Intellectual Gap
The case for prediction markets — the one justifying the media deals, the valuations, the lobbying — rests on Hayek's 1945 argument that prices aggregate dispersed information more efficiently than any centralized authority. a16z has published the most sophisticated version: six major pieces across 2024 and 2025, invoking Hayek, Tetlock, Hanson, and Buterin. The framework is rigorous within its assumptions.
The assumption it never examines: that media integration is neutral. That distributing the price to hundreds of millions of people does not change the price's relationship to the outcome it measures.
In the Hayekian view, putting a prediction market price on CNN is an unambiguous good — more information to more people. In the reflexive view, it is the moment the feedback loop scales past a qualitative threshold. More people seeing the price means more people acting on the price means the price having more influence on the outcome. The instrument stops measuring and starts moving.
a16z never mentions Soros across six pieces, multiple podcasts, and two annual outlooks. This is not an oversight. Reflexivity complicates the commercial thesis. If prediction market prices reshape the events they measure, then the media integrations that a16z's portfolio companies are signing do not just distribute useful information. They amplify a feedback loop whose effects are unmodeled, unregulated, and — at current scale — unprecedented.
What This Means for Practitioners
The feedback loop is structural. The media integrations are multi-year deals. The question is not whether the loop exists but whether anyone builds the tools to navigate it.
The price is treated as an endpoint: a probability to display, a number to report. It is a midpoint. The price is produced, distributed, internalized, and fed back into the system. For anyone trading these markets, the analytical question is not “what does the price say?” but “what is the price doing to the thing it measures?”
Answering that requires empirical work that does not yet exist: tracking cross-platform price divergences after media amplification events, measuring the velocity of narrative feedback, mapping the lag between displayed probability shifts and downstream behavioral response. It requires identifying where the Hayekian model holds — thick markets, well-defined questions, organic liquidity — and where the Sorosian model dominates — thin markets, politically salient questions, media-amplified prices.
The platforms have no incentive to do this work. The media partners have no incentive. The venture investors have no incentive. The incentives all point toward treating the price as fact.
The price is not a fact. It is a feedback loop. And in 2026, the loop has been wired into the infrastructure of American information.
Dawson Smith writes Reflexivity, a newsletter on prediction markets as reflexive systems.