The Cost of Being Wrong
In 1796 a Federalist named Joab Griswold bet "more pounds than I am worth pence" that his man would win a congressional seat by a thousand votes. His man lost by eighteen. Griswold was, by his own arithmetic, ruined. It's a funny story until you notice the thing underneath it: Griswold's wager was more honest than any pundit's prediction that year, and for one reason only — he had to pay for being wrong. That single feature, the price of error, is the quiet engine under prediction markets, under proof-of-work, and under the entire question of whose version of the truth you should ever believe.
For more than a century, Americans found out who was winning an election the way you'd find out the price of wheat: they looked at a market. Specialized bookmakers set odds from Wall Street offices. By 1896, the money wagered on the presidential race is believed to have exceeded the volume of trading on the New York Stock Exchange. In 1916, more was bet on the election than both campaigns spent. The major papers printed the odds daily, as news, because the odds were the news — the single best estimate anyone had of what was about to happen.
Then, in 1936, George Gallup called the Roosevelt landslide with a new thing called scientific polling, and within eight years the betting markets were gone. The papers swapped the odds for the pollster's percentages. And almost nobody noticed what had actually been traded away.
What we swapped, and what it cost
We replaced a market with an authority.
The old signal was decentralized and self-funding: thousands of people, each with money on the line, each punished instantly and personally for being wrong, aggregated into a single price. The new signal was centralized and credentialed: one expert, with a methodology, telling you the answer. The crowd that paid to be right gave way to the specialist who got paid regardless.
And if that shape sounds familiar, it should, because it's the exact move that happened to money. Decentralized, market-discovered money — gold, competing notes, value the market chose — was replaced by a central authority that issues the truth about money by decree. Same substitution, same century, same logic: a distributed price signal retired in favor of an institution that promises to know better.
Which is why the thing happening right now should interest you more than the midterm it's attached to. The betting markets are back — Polymarket, Kalshi, more than a billion dollars already wagered on next week's races, ten billion projected for 2028 — and they are clawing the truth-telling function back from the pollster, exactly as hard money is trying to claw it back from the central bank. Prediction markets are to the pollster what Bitcoin is to the central banker. The market is coming to repossess the truth the authority borrowed.
Why a bet beats a poll
Here's the mechanism, and it's the whole point, so slow down for it.
Why would a betting line be more honest than a survey? Because a bet is a costly signal and a poll is cheap talk. When a pollster calls you, lying or guessing or performing costs you nothing — you hang up and get on with your day, and the pollster who calls it wrong loses a little face and no money. But when you put a wager down, your belief acquires a price. You will only bet against the obvious if you can afford to be wrong, which means the only people moving the line are the ones who've weighed their confidence against their wallet. The cost is the filter. The cost is what squeezes the performance and the wishful thinking out, and leaves something closer to what people actually believe will happen — as opposed to what they'd like to say.
Now look at what else runs on exactly this principle. Bitcoin's ledger is secured not by an authority vouching for it but by the fact that writing a lie into it is expensive — you'd have to burn real energy, at real cost, to forge the record, and the system is arranged so that telling the truth is always cheaper than faking it. That's proof-of-work. And a prediction market is proof-of-work for beliefs: truth you have to pay to falsify. The honest line and the honest ledger are the same invention wearing different clothes. In both, the integrity doesn't come from trusting anyone. It comes from making the lie cost more than the truth.
What they both stand against is truth-by-decree — the pollster, the central banker, the credentialed expert who emits the official number at no personal cost and bears no personal loss when it's wrong. That's the deepest thing this publication keeps circling, and prediction markets state it cleanly: the question to ask of any claimed truth is not who said it but who pays if it's false. A signal nobody bleeds for is just weather.
It even closes the loop on the feed. The algorithmic feed is the purest cheap-signal machine ever built — costless to post a lie, and the lie spreads because it's free, selected for contagion rather than correctness. A prediction market does the one thing the feed deleted: it puts a price back on being wrong. It is, structurally, the feed's antidote — the consequence, reinstalled. The honest edge
Keep the blade out, because "the market always knows" is itself a seductive, spreadable, flattering meme, and swallowing it whole is the exact error this essay warns against. Three cuts.
First, market-truth was never pure, and the history proves it in its own words. In the early 1900s the Tammany Hall machine kept a dedicated war chest to manipulate the odds — a historian called it "indirect propaganda" — and in 1904 the New York Times accused Republicans of rigging the line to make Teddy Roosevelt look inevitable. Whales push Polymarket the same way today. Cost raises the price of lying; it does not make it infinite. A market is less gatekept than an authority, not un-gatekept, and anyone who sells you the frictionless version is doing the pundit's trick in a new hat.
Second, and nastier for markets than for polls: reflexivity. The odds can cause the outcome they claim only to measure — the bandwagon, the manufactured scent of inevitability that pulls the undecided toward the favorite. It's a thermometer that changes the temperature. When there's real money behind the number, "indirect propaganda" isn't an abuse of the system; it's a available strategy inside it, and the signal starts eating its own tail.
Third — the deepest one, and it's carved into law. New York amended its constitution in 1874 to bar anyone who had bet on an election from voting in it, and the clause is still on the books. That law understands something the cheerleaders forget: skin in the game cuts both ways. A financial position makes you honest about what you believe and simultaneously hands you a motive to rig what happens. The instant the same wallet is used for telling the truth and for making the outcome, the truth-machine has a conflict of interest welded into its heart. (Proof-of-stake has precisely this problem, incidentally: the people securing the ledger are the people who profit from it. Proof-of-work keeps the truth-teller and the beneficiary at arm's length by making the truth cost something external — energy — that can't simply be voted into your own pocket.)
The coda that should unsettle you
And then there's the twist that turns the whole thing reflexive at the species level. A recent paper found that the best way to train an AI to forecast real events is to train it on resolved prediction-market questions — and the trained models went on to out-predict every frontier system tested, at a fraction of the cost. Which means the market's honest, costly, hard-won signal becomes the training data for a machine that can then forecast the outcome before the costly crowd gets paid to price it.
The real purpose of the web, it turned out, was to train machines to think. The real purpose of prediction markets may turn out to be training them to predict. A truthmachine, mining itself out of a job — and leaving open the question of what an honest price even means once the answer is known before the betting opens.
The one line to keep
We spent a century finding the truth in a market, then handed the job to an authority who pays nothing to be wrong, and we're only now reclaiming it — in our elections and in our money at the same time, for the same reason. A poll is what people say. A bet is what they're willing to lose. And only one of those costs enough to be trusted.
The question to carry out of here isn't who told you the truth. It never was. It's who pays if it's false — because a signal no one bleeds for has never, in the whole history of money or markets or men betting more pounds than they're worth, been worth the paper the odds were printed on.
In 1796 a Federalist named Joab Griswold bet "more pounds than I am worth pence" that his man would win a congressional seat by a thousand votes. His man lost by eighteen. Griswold was, by his own arithmetic, ruined. It's a funny story until you notice the thing underneath it: Griswold's wager was more honest than any pundit's prediction that year, and for one reason only — he had to pay for being wrong. That single feature, the price of error, is the quiet engine under prediction markets, under proof-of-work, and under the entire question of whose version of the truth you should ever believe.
For more than a century, Americans found out who was winning an election the way you'd find out the price of wheat: they looked at a market. Specialized bookmakers set odds from Wall Street offices. By 1896, the money wagered on the presidential race is believed to have exceeded the volume of trading on the New York Stock Exchange. In 1916, more was bet on the election than both campaigns spent. The major papers printed the odds daily, as news, because the odds were the news — the single best estimate anyone had of what was about to happen.
Then, in 1936, George Gallup called the Roosevelt landslide with a new thing called scientific polling, and within eight years the betting markets were gone. The papers swapped the odds for the pollster's percentages. And almost nobody noticed what had actually been traded away.
What we swapped, and what it cost
We replaced a market with an authority.
The old signal was decentralized and self-funding: thousands of people, each with money on the line, each punished instantly and personally for being wrong, aggregated into a single price. The new signal was centralized and credentialed: one expert, with a methodology, telling you the answer. The crowd that paid to be right gave way to the specialist who got paid regardless.
And if that shape sounds familiar, it should, because it's the exact move that happened to money. Decentralized, market-discovered money — gold, competing notes, value the market chose — was replaced by a central authority that issues the truth about money by decree. Same substitution, same century, same logic: a distributed price signal retired in favor of an institution that promises to know better.
Which is why the thing happening right now should interest you more than the midterm it's attached to. The betting markets are back — Polymarket, Kalshi, more than a billion dollars already wagered on next week's races, ten billion projected for 2028 — and they are clawing the truth-telling function back from the pollster, exactly as hard money is trying to claw it back from the central bank. Prediction markets are to the pollster what Bitcoin is to the central banker. The market is coming to repossess the truth the authority borrowed.
Why a bet beats a poll
Here's the mechanism, and it's the whole point, so slow down for it.
Why would a betting line be more honest than a survey? Because a bet is a costly signal and a poll is cheap talk. When a pollster calls you, lying or guessing or performing costs you nothing — you hang up and get on with your day, and the pollster who calls it wrong loses a little face and no money. But when you put a wager down, your belief acquires a price. You will only bet against the obvious if you can afford to be wrong, which means the only people moving the line are the ones who've weighed their confidence against their wallet. The cost is the filter. The cost is what squeezes the performance and the wishful thinking out, and leaves something closer to what people actually believe will happen — as opposed to what they'd like to say.
Now look at what else runs on exactly this principle. Bitcoin's ledger is secured not by an authority vouching for it but by the fact that writing a lie into it is expensive — you'd have to burn real energy, at real cost, to forge the record, and the system is arranged so that telling the truth is always cheaper than faking it. That's proof-of-work. And a prediction market is proof-of-work for beliefs: truth you have to pay to falsify. The honest line and the honest ledger are the same invention wearing different clothes. In both, the integrity doesn't come from trusting anyone. It comes from making the lie cost more than the truth.
What they both stand against is truth-by-decree — the pollster, the central banker, the credentialed expert who emits the official number at no personal cost and bears no personal loss when it's wrong. That's the deepest thing this publication keeps circling, and prediction markets state it cleanly: the question to ask of any claimed truth is not who said it but who pays if it's false. A signal nobody bleeds for is just weather.
It even closes the loop on the feed. The algorithmic feed is the purest cheap-signal machine ever built — costless to post a lie, and the lie spreads because it's free, selected for contagion rather than correctness. A prediction market does the one thing the feed deleted: it puts a price back on being wrong. It is, structurally, the feed's antidote — the consequence, reinstalled. The honest edge
Keep the blade out, because "the market always knows" is itself a seductive, spreadable, flattering meme, and swallowing it whole is the exact error this essay warns against. Three cuts.
First, market-truth was never pure, and the history proves it in its own words. In the early 1900s the Tammany Hall machine kept a dedicated war chest to manipulate the odds — a historian called it "indirect propaganda" — and in 1904 the New York Times accused Republicans of rigging the line to make Teddy Roosevelt look inevitable. Whales push Polymarket the same way today. Cost raises the price of lying; it does not make it infinite. A market is less gatekept than an authority, not un-gatekept, and anyone who sells you the frictionless version is doing the pundit's trick in a new hat.
Second, and nastier for markets than for polls: reflexivity. The odds can cause the outcome they claim only to measure — the bandwagon, the manufactured scent of inevitability that pulls the undecided toward the favorite. It's a thermometer that changes the temperature. When there's real money behind the number, "indirect propaganda" isn't an abuse of the system; it's a available strategy inside it, and the signal starts eating its own tail.
Third — the deepest one, and it's carved into law. New York amended its constitution in 1874 to bar anyone who had bet on an election from voting in it, and the clause is still on the books. That law understands something the cheerleaders forget: skin in the game cuts both ways. A financial position makes you honest about what you believe and simultaneously hands you a motive to rig what happens. The instant the same wallet is used for telling the truth and for making the outcome, the truth-machine has a conflict of interest welded into its heart. (Proof-of-stake has precisely this problem, incidentally: the people securing the ledger are the people who profit from it. Proof-of-work keeps the truth-teller and the beneficiary at arm's length by making the truth cost something external — energy — that can't simply be voted into your own pocket.)
The coda that should unsettle you
And then there's the twist that turns the whole thing reflexive at the species level. A recent paper found that the best way to train an AI to forecast real events is to train it on resolved prediction-market questions — and the trained models went on to out-predict every frontier system tested, at a fraction of the cost. Which means the market's honest, costly, hard-won signal becomes the training data for a machine that can then forecast the outcome before the costly crowd gets paid to price it.
The real purpose of the web, it turned out, was to train machines to think. The real purpose of prediction markets may turn out to be training them to predict. A truthmachine, mining itself out of a job — and leaving open the question of what an honest price even means once the answer is known before the betting opens.
The one line to keep
We spent a century finding the truth in a market, then handed the job to an authority who pays nothing to be wrong, and we're only now reclaiming it — in our elections and in our money at the same time, for the same reason. A poll is what people say. A bet is what they're willing to lose. And only one of those costs enough to be trusted.
The question to carry out of here isn't who told you the truth. It never was. It's who pays if it's false — because a signal no one bleeds for has never, in the whole history of money or markets or men betting more pounds than they're worth, been worth the paper the odds were printed on.
3🤮1
