Prediction Markets are Attention Labor Markets

The New York Times this week profiled traders making millions on Kalshi and Polymarket. The article frames prediction market trading as one of those era-defining occupations, like being a Wall Street trader in the 1980s, a dot-com founder in the 1990s, or an influencer in the 2010s.

Influencers got paid to be noticed. Prediction market traders get paid to notice.

Traders stood outside Rockefeller Center after an SNL rehearsal canvassing attendees to find out whether Elon Musk had said “DOGE.” A market existed on whether a single word would be spoken on live television, liquid enough that someone calculated the expected value of standing in the cold asking strangers what they heard and decided it was worth doing.

The edge didn’t go to someone with insider access. It went to whoever was willing to physically show up and ask people walking out of a building what happened inside.

Joel Holsinger made hundreds of dollars betting Trump wouldn’t say “stuffing” at the turkey pardon by studying word frequencies in transcripts of past speeches.

Jonathan Zubkoff built what he calls “the Bloomberg terminal for Rotten Tomatoes information,” a custom dashboard tracking entertainment newsfeeds so he can model weekly review scores before aggregation.

The NYT frames all this as quantitative aptitude and above-average information-processing skills. But reading through what these traders actually do with their days, almost none of it involves knowing something hidden.

Holsinger studying old transcripts. Zubkoff refreshing a dashboard. Traders standing outside buildings. The information sits there, available to anyone who cares to look.

The edge is the caring. The willingness to do tedious systematic work that anyone could do and almost no one will.

The work itself has a particular texture.

Non-English Telegram channels for war markets. Embassy appointment calendars for geopolitical signals. Calling to check emergency passport availability because a discrepancy between what the website says and what the operator tells you sometimes means something is about to happen. Tracking flights into airports before a VP announcement because one scheduled from Alaska might mean something.

This is the work. Paying attention in languages and mediums and channels most traders won’t bother with, not because they can’t access them but because they won’t do the labor of monitoring them.

A bond mule locks up capital on markets resolving soon for small premiums, betting against outcomes with extremely low probability of occurring. On the other side of these trades sit people who share three characteristics: undercapitalized, leverage-starved, obsessed with convex payouts. They buy “yes” shares at one or two cents on moonshots, not because they’ve calculated the probability but because the potential payout excites them, because the idea of turning two cents into a dollar feels like winning the lottery even if the odds make it a terrible bet. They don’t think about who’s on the other side stacking the other 98 cents to entertain their fantasy.

One side doing attention labor. The other side buying lottery tickets and calling it trading.

The influencer economy rewarded being noticed. Prediction markets reward noticing.

The influencer economy made it possible for anyone to monetize attention by becoming the thing people pay attention to. Build an audience, become your own media company, convert followers into dollars through ads and sponsorships and products. You had to be noticed to get paid.

Prediction markets flip this. You can monetize the labor of paying attention itself. The work of discovering early, of caring about tedious things before anyone else does. You don’t need an audience. You don’t need to be famous. You do the work of noticing and settle directly in cash.

Music pages on Instagram post Spotify listener growth formatted like trading P&L. Fakemink at 57,000 monthly listeners when they found him, 8.7 million now.

Nobody told them to present discovery this way. The format emerged on its own across thousands of accounts that don’t know each other because something about the underlying economy made it feel natural, because “I found this at 57k” wants to be expressed the same way as “I bought at 5 cents.”

Early attention as a position. Growth as returns. Receipts as proof of work performed.

The structure emerged because the economy underneath is the same economy, just settling in followers instead of dollars. Music pages with great taste still need to convert that into followers, then into sponsorships or affiliate deals or whatever comes next. The Polymarket sharp just gets paid. The prediction market traders didn’t invent attention labor.

They’re doing the same job the internet has rewarded for years. The difference is the settlement layer.

Over 170 third-party tools now exist around Polymarket across 19 categories.

Whale tracking lets you see who’s paying attention and to what, the financialized version of following tastemaker accounts. Copy trading lets you automatically mirror positions of successful traders, which is paying to skip the attention labor entirely. Alert systems notify you when someone whose attention you trust makes a move.

None of these tools help you predict better. They help you track, copy, or automate attention labor.

Top traders now use secondary and tertiary wallets because their mains get copy-traded immediately. They don’t have private intelligence about whether Trump will say “stuffing.” They’re protecting their attention labor from being free-ridden. The same resentment a music discovery account feels when bigger accounts repost their finds without credit.

The LLMs everyone thought would provide edge have become table stakes. If you offload all of your cognitive workload to a language model, you’re indistinguishable from anyone else with the same $15 subscription. The edge is in the work that can’t be automated.

Two distinct user types are separating as prediction markets mature.

Pro users rely on aggregators and analytics, increasingly competing with AI agents that arbitrage faster than humans can, pushing markets toward efficiency where the easy edges disappear and the work moves to harder layers.

Social users optimize for signaling, not profit. They’re betting on the Super Bowl or the election because they want skin in the game on outcomes they care about, treating prediction markets like the influencer economy, where the point is having a position to talk about rather than maximizing returns.

The 0.04 percent of Polymarket addresses that account for 70 percent of profits are the ones willing to do the work. As Domer put it, “If I made $2.5 million last year, someone else lost that.”

The fish aren’t just undercapitalized. They’re emotional.

Politics and sports are where even traders with phenomenal systems throw out large chunks of their P&L by trading on feelings rather than work.

This is what attention labor looks like when it professionalizes.

Holsinger told the Times he’s “so terminally online, I don’t think I even recognize my own neighborhood. I spend maybe 16 hours a day on the computer.” Domer described sitting down to dinner with his wife and having to run upstairs because Eric Adams just tweeted. During the Israeli election his sleep schedule was based around Israel’s time zone.

These aren’t side effects of the job. They are the job. The edge goes to whoever is paying attention when everyone else is eating dinner or sleeping or living in their neighborhood.

When the settlement happens in dollars instead of followers, the returns justify giving your entire life over to the work. The music discovery accounts can close the app and go outside. The Polymarket sharps can’t.

If prediction markets are really attention labor markets, the next step isn’t better predictions.

It’s dropping the pretense entirely and trading attention directly.

Noise does something like this. Instead of asking whether something will happen, it asks how relevant something is right now and where that relevance is going.

Prices derived from aggregated mentions, interactions, growth metrics across platforms. No resolution event. No expiration date. Just the thing the internet is paying attention to, made tradeable.

Which starts to sound less like a prediction market and more like what the music discovery accounts have been doing all along, except with financial settlement built in.

The first era of the attention economy let anyone become the object of attention and monetize that through audiences. The second era lets anyone monetize the labor of paying attention without ever being noticed themselves.

The infrastructure is finally starting to catch up to the work.