Regulation, Insider Information, and Essence: The Story Behind Kalshi's $20 Billion Valuation

Video Author: John Collison

Translation: Peggy, BlockBeats

Editor’s Note: Over the past few years, prediction markets have evolved from a relatively fringe financial experiment to a central topic in discussions of technology, finance, and public policy.

Their widespread attention is not only because of the appeal of “betting on the future” itself but also because, against the backdrop of social media amplifying noise, polls repeatedly missing the mark, and declining trust in traditional information systems, a more fundamental question has emerged: Can market prices serve as a signal mechanism that is closer to reality than opinions, emotions, and narratives?

This conversation centers around that question. Participants include Stripe co-founder John Collison, Paradigm co-founder Matt Huang, and Kalshi co-founders Tarek Mansour and Luana Lopes Lara.

Kalshi’s two co-founders Tarek Mansour (right) and Luana Lopes Lara (left)

As one of the most representative compliant prediction markets in the U.S., Kalshi gained rapid prominence during the 2024 U.S. election. Before this surge, it had spent years negotiating with the Commodity Futures Trading Commission (CFTC), ultimately opening the door to legalization of prediction markets in the U.S. through a key lawsuit.

The first part of the conversation focuses on Kalshi’s path to creation: why the founders chose “compliance first, growth later” instead of the common Silicon Valley approach of “build first, ask for permission later”; why they endured long approval processes, layoffs, and external skepticism to secure the “election market”; and how the lawsuit against the CFTC became a turning point for the company’s growth.

The second part delves into the operational logic of prediction markets. Tarek and Luana explain the fundamental difference between Kalshi and traditional online entertainment platforms: it is not a “house model” that profits from user losses, but a trading exchange centered on fees that encourages liquidity and information entry. They also highlight a counterintuitive reality: Kalshi’s liquidity does not mainly come from traditional large market makers but from a large number of dispersed individual traders, “super forecasters,” and small teams. In a sense, prediction markets are not just financial products but mechanisms that directly convert dispersed cognition into price signals.

In the latter half, the discussion extends to the future boundaries of prediction markets: from elections and sports to AI, GPU computing power, macro variables, and policy pathways—can more and more real-world uncertainties be broken down into tradable, feedback-driven, decision-supporting markets? Meanwhile, several unavoidable controversies surface—how to define insider trading, whether sports contracts will amplify online gambling risks, and how platforms and regulators can establish new balances between innovation, transparency, and user protection.

Because of this, the significance of this conversation extends beyond Kalshi itself. It aims to answer: Will prediction markets become the next-generation financial markets or the next-generation information infrastructure?

Below is the original content (reorganized for clarity):

TL;DR

· Kalshi chose an unusual path: regulation first, growth later: spent 3 years obtaining licenses, sued the CFTC to open the election market. The core judgment is that whether prediction markets can exist legally is more important than growth.

· The essence of prediction markets is incentivizing truthful information with money: compared to polls and social media, markets filter information through profit and loss mechanisms, serving as signals closer to truth.

· Ordinary people, not institutions, form the core liquidity: over 95% of matches come from dispersed users and super forecasters, not traditional market makers.

· Kalshi emphasizes itself as an exchange, not an online entertainment platform: revenue from fees, not user losses; encourages expert participation rather than restricting winners like entertainment industries.

· Elections are the holy grail, but future markets will go far beyond: from sports and macro variables to AI and computing power, the team aims to build a derivative system that prices everything.

· Prediction markets are becoming a new information infrastructure: users not only trade but also consume probabilities; 80% of users mainly use it to understand the world rather than to bet.

· Behind its rise is distrust in traditional information systems: polarized social media and inaccurate polls push people toward price-based judgment mechanisms.

· The long-term goal is to improve societal decision-making efficiency, not just to run a trading platform: through continuous pricing and feedback, political and economic fields can reach genuine consensus faster.

Interview Summary

John Collison (Stripe co-founder & host):

Tarek Mansour and Luana Lopes Lara are co-founders of Kalshi. Kalshi is an emerging prediction market company that quickly gained fame during the 2024 U.S. election. To establish the first compliant prediction market in the U.S., they spent four years negotiating with regulators and seeking approval before launching. Now, Kalshi’s monthly trading volume exceeds $10 billion.

So, how do you two divide your roles? But more than that, what are your different ways of approaching problems?

Luana (Kalshi co-founder & COO):
Actually, our backgrounds are almost identical. We both studied math and computer science at MIT, interned similarly, with little difference. But I’m very optimistic, adventurous, always believing things will turn out fine; he’s very cautious, even somewhat pessimistic. So I think that creates a good balance. Looking back, beyond daily roles, our true complement is this difference.

Tarek (Kalshi co-founder & CEO):
I’ll add a bit of background. I originally planned to be a trader—that was almost my career path. If you’ve met traders, you’ll understand—they always seem to have a calculator for expected returns in their heads.

Matt Huang (Paradigm co-founder):
A very typical trader.

John Collison:
Right, but—

Tarek:
If you’re really a trader, you’re constantly thinking about tail risks, worst-case scenarios. She doesn’t usually think that way. I think it’s precisely this difference that leads to good results.

Why Kalshi Chose the Hard Path: Compliance First, Growth Later

John Collison:

I want to ask about this. Your starting point is interesting—Kalshi took years before it could operate properly, only after getting CFTC approval. Most companies don’t start that way. On the other hand, Silicon Valley often follows a pattern—PayPal, Uber in their early days—“build first, then seek permission”—get things going, then patch the structure and licenses afterward. It’s build first, then ask for forgiveness, not ask first.

Can you tell us how you started? How did the approval process go? And do you think this path applies to other companies?

Luana:
From the beginning, we knew very clearly that if you’re doing financial services or healthcare, you can’t just build first and ask later. In finance, once user funds are involved, the cost of problems is huge—FTX is a typical example; healthcare is even more sensitive, with many disastrous precedents. We wanted to do it the right way. More importantly, when we looked at this market, the core issue wasn’t growth but whether it’s legal to operate in the U.S. So we decided to address the biggest legal hurdle first, then move forward. For a long time, many thought this was the wrong strategy.

Before we won the lawsuit on election contracts, many believed offshore markets were doing better and growing faster. But after we won that case, proving our legal understanding was correct and that we could operate legally in the U.S. as we envisioned, things really took off.

John Collison:
What was the timeline? When did you start? When did you win the election contract lawsuit?

Luana:
We founded the company in 2019, got into YC that year. It took us three years to get regulatory approval and launch, around 2022. Then, at the end of 2024, we won the election contract lawsuit, and that’s when the company really started accelerating.

Tarek:
There are two levels here. First, practical considerations. We believed that to gain mainstream and institutional adoption, the key question was whether this could operate within a regulated, trustworthy, and safe framework. After all, it’s a complex market involving user funds. We had to solve this hardest problem first—that’s the path to success.

Second, it’s a matter of principle. When we first drafted that one-page document on Google Docs, we listed questions: Why do we want to build this company? Why are we excited? Our answer was, we want to create the next New York Stock Exchange. We want a trusted, regulated financial market in the U.S. We’re not excited about offshore setups that mimic that. The key is: what kind of company do you want to build? Why do you want to do this? There are many paths to success, but the one we don’t want is offshore. We want this to happen here, in the U.S.

John Collison:
You’re the first to get CFTC approval and reach a certain scale.

Tarek:
Yes, exactly.

John Collison:
And even today, each contract still needs individual approval, right?

Luana:
Yes. Every contract we submit to the CFTC can be vetoed within 24 hours.

John Collison:
So they almost receive your contract info stream in real time?

Luana:
Yes, that’s correct.

Tarek:
Exactly. To reach today’s level of contract processing, it’s been a very long process. Imagine the first time we walked into the CFTC building, our minds were full of this concept, and the regulators had to operate at high speed. Because we’re discussing a product without traditional underlying assets, with dozens or hundreds of contracts weekly, and no precedent for this kind of regulation. Now we do much more, but initially, this regulatory model wasn’t designed for such scenarios.

So, it’s like iterating on a product—except you’re not serving customers but working with regulators to figure out how to regulate this. What concerns do they have? What can we do to address those concerns?

Luana:
In a way, it’s about finding a regulatory product-market fit.

Matt Huang:
So now you’re used to this rhythm—sending out contracts unless explicitly blocked. Have they ever vetoed anything recently?

Luana:
Not recently. The biggest rejection was the election contract, which led us to sue them. They blocked us for two years over that. But now, we’ve worked with them for so long that both sides understand the boundaries. They trust us, knowing we’re a self-regulatory entity that understands what can and cannot be done. For example, markets on war or assassination—those we don’t do. As long as we stay within established boundaries, the process speeds up significantly.

John Collison:
So, just to confirm, the core of the election lawsuit was that they’re generally willing to approve various contracts but refuse to approve contracts about who will win the election, which is the most popular type, especially during U.S. presidential elections. So you sued the CFTC.

Tarek:
Yes. It’s actually their own rules—

John Collison:
And generally, suing your regulator isn’t considered best practice.

Tarek:
Indeed. It started at the end of 2021 when we pushed for the election market, began communicating with policymakers, Congress, and regulators. Everyone said it sounded good. But they never advanced it, and we started feeling something was wrong. By late 2022, they effectively delayed approval until after the election—pocket veto. That period was very tough; we had to lay off many people. More painfully, the team, investors, and most investors started losing faith in this path.

John Collison:
It’s not that they don’t believe in the idea, but they don’t believe in this strategy anymore.

Tarek:
Exactly. They no longer believe in the strategy, even question the idea itself. People thought things were unhealthy—should we do something else? Clearly, this path seemed impossible. But we couldn’t force ourselves to do something else. We just kept trying.

Imagine the morale was at rock bottom, everyone waiting for a new strategy. Many left, many were laid off, because we had to downsize. Then, at the next stand-up, we told everyone: 2023’s plan is—let’s try again.

John Collison:
So, you’re doing the same thing, just hoping it will succeed this time.

Tarek:
Exactly. This time, it will. Despite almost all evidence pointing the other way. I have to say, a lot of that was her pushing. I really wanted this to succeed, but my rational mind kept telling me—listen to these people, this path won’t work. But she was more determined. So we tried again. By late 2023, they blocked it again. I was almost thinking—

John Collison:
Well, prediction markets just can’t be done.

Tarek:
Yes, I really felt that way. Then she said, among all options, the only thing left is to sue the government. My first reaction was: that’s crazy. We brought it to the board—Alfred, Michael, and Seibel from my side were there.

John Collison:
That’s Alfred Lin and Michael Seibel.

Tarek:
Yes. I remember those board meetings—initially, it was basically, “You need to understand, this is a bad idea.” There are many reasons: your opponents are regulators; you’re only twenty-something; the government really wants to shut you down, revoke your license, and can do it. And it’s not just theoretical risk. Even if you win, you might be dragged through the process.

I also remember, before the official board discussion, we had an internal meeting. The night before we contacted lawyers and prepared to sue, I suddenly hesitated. I said, maybe we should just do a clearinghouse, or focus more on financial products—don’t go all in on this. The exact words I don’t remember, but roughly it was, “Are you joking?”

Luana:
That definitely sounds like something I would say.

Tarek:
I realized then that I couldn’t win that argument. But part of me also knew we had to do it. Later, we discussed with the board—they basically said it was an anti-pattern, a bad idea. But many great companies are built on anti-patterns; something abnormal always happens—maybe this is your abnormality.

John Collison:
That’s a good point. Every company eventually finds a new, unusual way to succeed—that might be yours. So, what was the legal basis for winning the election lawsuit? Any interesting policy angles?

Luana:
The core is simple: the government cannot arbitrarily ban a contract unless it’s deemed against public interest, and such bans must fall into specific categories like war, terrorism, assassination, etc. At the time, the CFTC tried to shoehorn election contracts into these categories. They argued that elections might be illegal under some state laws, even citing a state’s bucket shop law to block it.

But our legal position was clear: elections have economic impact, and as long as they do, they should be tradable on futures or derivatives exchanges. The lawsuit ultimately told the CFTC: you can’t do whatever you want.

John Collison:
So, the so-called banned categories must truly fall into those explicitly prohibited, and elections clearly do not.

Luana:
Exactly.

Tarek:
This is very important. We often say laws constrain companies, but laws also constrain governments.

John Collison:
Right. Matt, you mentioned the two or three years of challenging the government?

Matt Huang:
Yes. I think in crypto and prediction markets, suing the government seems especially common, but I’ve realized it’s actually more common than in traditional Silicon Valley thinking. Coinbase sued its main regulator; in GovTech, SpaceX, Anduril, Palantir have all sued the government for various reasons. So I’m curious—since you’ve had so much interaction with the government—what advice would you give to others wanting to do similar things? When is challenging the government the right move?

Tarek:
I think only when there’s no other choice. It’s still very painful.

John Collison:
But are you really out of options? Without the election market, can’t you just keep going? Of course, election markets are very attractive, but I guess they’re not your main source of contracts now?

Luana:
I think it’s just too important. Maybe it sounds obsessive, but it’s the holy grail market. It best demonstrates how data from prediction markets can be used, and shows their value. Take the 2024 election—polls were wildly wrong, but the market did a much better job of aggregating information. It’s the most shining example, proving why prediction markets are beneficial and why the U.S. needs to have them within a regulated framework. No other market has such a strong demonstration.

Core Logic of Prediction Markets: Using Real Money to Produce Information

Matt Huang:
John mentioned PayPal and Uber’s build-first approach. Actually, other prediction markets already operated offshore and showed real demand. So I wonder—did that help you in your lawsuit? For example, does it show that election markets don’t conflict with public interest—since people are already doing them?

Tarek:
I’m not sure. But from a legal perspective, the focus is more on the text of the law itself. We’re discussing the Commodity Exchange Act, one of the core financial regulations; and the Securities Exchange Act. The key is to read and interpret these laws carefully and judge whether the regulator is overstepping.

From our perspective, offshore markets help because they provide external data points in a pre-regulation, product-first approach. We couldn’t learn directly from our own product at first because we insisted on licensing first. So, external data and evidence can help us make decisions. They also help more people understand what prediction markets are and how to use them. But in policy terms, do offshore players help us? Probably not much.

John Collison:
If Kalshi had appeared ten or fifteen years ago, would it have been impossible? Is it because the current CFTC is more open, or because certain tech like stablecoins had to mature first?

Luana:
I think part of it is related to crypto. Early prediction markets like Augur already existed. Their existence probably made the CFTC realize they need a legitimate, regulated alternative. Before, they could just say no. That played a role, maybe 5-10%, but not more.

Tarek:
More broadly, I think the public’s interest in prediction markets has been around since the 1950s. People have long known it’s a better signal source than many other mechanisms. But ten or fifteen years ago, there wasn’t such a pressing practical need. Recently, that need has become real. Society is more divided, the world more fractured. Polarized social media fragments information streams into camps, clickbait is rampant, and most content—whether traditional news or social media—has incentives to attract eyeballs. Because of that, more urgent problems have emerged, leading to the adoption of prediction markets. I don’t think fifteen years ago we’d see today’s situation because the problems weren’t as severe then.

Luana:
Most of our users aren’t even trading—they’re consuming information. About 80% use it to understand the world. For example, yesterday’s Texas primary—people look at the odds, see polls say a tie, but the market does a better job of aggregating information. It’s a very strong case for why prediction markets are beneficial and why the U.S. needs them in a regulated environment. No other market is as demonstrative.

Core Logic of Prediction Markets: Using Real Money to Produce Information

Matt Huang:
John mentioned PayPal and Uber’s build-first approach. Actually, other offshore prediction markets already proved demand. Did that help in your lawsuit? For example, does it show prediction markets don’t conflict with public interest—since people are already doing them?

Tarek:
I’m not sure. But legally, the focus is on the law’s text. We’re analyzing the Commodity Exchange Act and the Securities Exchange Act. The key is to interpret these laws carefully and judge whether the regulator is overreaching.

Offshore markets do help because they provide external data points in a pre-approval, product-first approach. We couldn’t learn from our own product initially because we insisted on licensing first. So external data and evidence can inform our decisions. They also help more people understand prediction markets and how to use them. But do offshore players help us policy-wise? Probably not much.

John Collison:
If Kalshi had appeared ten or fifteen years ago, would it have been impossible? Is it because the current CFTC is more open, or because certain tech like stablecoins had to mature first?

Luana:
Partly, yes. Crypto’s existence, like early Augur, made the CFTC realize they need a legitimate, regulated alternative. Before, they could just say no. That played a role, maybe 5-10%, but not more.

Tarek:
More broadly, interest in prediction markets has existed since the 1950s. It’s a better signal source than many others. But the pressing need is recent. Society is more divided, information streams are fragmented, and incentives favor sensationalism. This creates a demand for prediction markets. I don’t think fifteen years ago we’d see today’s situation because the problems weren’t as acute.

Luana:
Most users aren’t trading—they’re consuming information. They look at the odds, compare with polls, and realize the market does a better job of aggregating information. It’s a compelling proof of prediction markets’ value and necessity.

The Distinction Between Prediction Markets and Entertainment: Trading vs. Betting

John Collison:
Another market design question. In sports betting, there’s a well-known phenomenon: betting companies suppress so-called sharp bettors—those who are too smart or too good at betting. Many don’t realize that for betting companies, the ideal bettor is someone who’s just supporting their team; the worst are those who find mispricings in niche markets. Because they might offer odds on thousands of markets, and if they get a few wrong, sharp bettors will exploit those errors. They can detect signals from behavior—if you just bet on your team, that’s fine; if you’re too professional, they might ban you.

It’s interesting—you think you’re just betting on their odds, but if you’re too good, they don’t want you. Like Las Vegas casinos asking you to leave. Does Kalshi face similar issues with sharp bettors? I thought you’d welcome them, but do market makers worry about facing too-smart opponents?

Tarek:
The problem is, sharp bettors are part of the market. To clarify—

Luana:
We don’t restrict winners. We don’t do that. We want the smartest people to come.

Tarek:
We need these sharps. Without them, the market wouldn’t be accurate. That’s the biggest difference from entertainment companies.

John Collison:
So your incentive system is completely different. You don’t profit from one side losing; you earn from transaction fees.

Luana:
Exactly. Our goal is for users to find the market fair, with good prices and stable liquidity, so they want to trade here. To achieve that, we also design incentives for different roles. That’s why we have various liquidity programs. If you provide liquidity and take on higher risk of being “sniped,” your fees should be lower; if you’re actively taking orders or trying to “snipe,” your fees should be higher to compensate for that.

John Collison:
So you use fees to promote pro-social behavior.

Luana:
Yes, I think that’s how traditional financial markets work.

Tarek:
Financial markets are built on that logic.

Luana:
Just not always so explicitly.

Tarek:
Fundamentally, it’s about giving those who create real value for the market a slight advantage, and those who just extract value a slight disadvantage.

John Collison:
What behaviors are pro-social, and which are anti-social?

Tarek:
Insider trading is obviously anti-social.

Luana:
That’s the most clear-cut.

Tarek:
And illegal. Regarding “sniping,” it’s also part of the market. When someone gains new information suddenly and trades on it, that happens every day in traditional markets. But if you want liquidity to persist and market makers to keep investing, you need to give them some incentives.

I think that’s one reason prediction markets are gaining acceptance—they align incentives with truth. More trading volume and liquidity lead to more accurate predictions. It takes time for people to trust this, but once established, they won’t want to go back to worse products.

John Collison:
Speaking of trading volume, can you give us a sense of Kalshi’s growth? It’s been very rapid.

Tarek:
In February, our trading volume was $10.4 billion.

John Collison:
That’s $10.4 billion in contract trading volume.

Tarek:
Yes. Compared to six months ago, it’s roughly an 11-fold increase, maybe more.

John Collison:
That’s so fast you don’t even bother looking back a year, it’s like ancient history.

Luana:
A year ago, it was completely different. We only had one sports market then.

Tarek:
Yes, that was in February. Overall, growth has been very rapid.

Matt Huang:
Apart from AI, probably the fastest-growing company.

Tarek:
I think so. Maybe even comparable to top AI firms. I don’t know the latest data from Cursor or Anthropic, but—

John Collison:
Even in AI, an 11x increase in six months is extraordinary.

Tarek:
Very fast. I think the reason is, we’re a real market with inherent network effects. As more categories and liquidity grow, user retention improves, participation and trading volume increase over time. This naturally drives more usage, which in turn attracts others because of higher liquidity and better products. Users are more willing to share with others. These forces combine to produce this rapid growth.

Matt Huang:
A large part of your early growth came from other broker platforms. Now the structure has changed. How do you view the broker channel? What’s its current share?

John Collison:
What do you mean by broker? Like Robinhood?

Tarek:
Yes. That’s an interesting question.

Luana:
I can explain the broker part first, then he can add specifics. Basically, since we’re an exchange and clearinghouse, our role is more like NYSE or, more precisely, CME. Brokers can connect to us. You can trade Kalshi contracts on Robinhood, and in the future, on Coinbase or other platforms too.

From the start, we’ve been clear that we’re primarily an exchange and clearinghouse, not something else. Establishing connections with Goldman Sachs, Robinhood, and others is crucial for understanding the ecosystem.

Last year, our first broker partners were Robinhood and Webull. During the rapid growth phase, broker channels accounted for a large share, which was good because brokers bring demand, and market makers are willing to step in to match retail flow. This also gave us time to develop direct-to-user products to today’s level.

Our current understanding is that the core remains exchange + clearinghouse. Users can access directly via our app, website, or API, or through any broker. We’re also investing more in institutional and international brokers. Soon, you’ll be able to trade Kalshi directly from Brazil. You tell me the numbers.

Tarek:
She’s reluctant to share exact figures, but our direct-to-consumer business—kalshi.com, the Kalshi app—is already growing faster than intermediary channels. I think it’s mainly because our brand has become well-known. Now, many people, when they disagree on something, their first instinct is to open Kalshi and check the odds or place a bet. The brand is becoming synonymous with this behavior. There’s a lot of organic growth now, and I believe this trend will continue in the coming months.

An Unconventional Market: Ordinary People More Important Than Institutions

John Collison:
You just discussed retail user growth—some come via broker platforms, others directly on Kalshi. But as an exchange, you also need to do market making. The NYSE doesn’t worry much about market makers because economic incentives are strong once scale is big. But I’m curious—how did you build your market-making system early on? Did you do it yourself? Partner with external market makers? How do you incentivize them? How did you start from zero?

Luana:
Kalshi’s markets can be divided into two types, with very different behaviors and incentives.

One type is long-tail markets, like “Will One Direction reunite?” These are hard to price and usually have low demand, so we need to attract market makers through various incentives, including recruitment bonuses. One of our ongoing concerns is how to establish stable, sustainable liquidity for these long-tail markets. Today, we might have 10,000 markets; in the future, 50,000 or 100,000—how to ensure liquidity then?

The other type is more classic markets, like crypto or sports. For these, market making is much easier because demand is clearer and pricing logic more mature. Incentives here are not direct payments but fee rebates, along with strict obligations—maintaining certain spreads or depth within specified times. In these markets, we’re more about incentivizing order book stability than just paying for quotes.

John Collison:
What do you mean by “order book stability”? Can you clarify?

Luana:
For example, during a live event or in hourly-settled crypto markets—

John Collison:
If no new information arrives, you don’t want prices to jump wildly, right?

Luana:
Exactly. Even if there’s new info—say, a touchdown is imminent—you don’t want the order book to suddenly lose all liquidity. You can allow some spread widening, but you still want traders to be able to trade. As we move toward a broker model, brokers will bring expectations from traditional markets. They’ll say, “We want spreads and depth maintained at certain levels at all times.” So we need to negotiate with market makers on how to design incentives. Because if we let the market decide freely, spreads might naturally widen during high volatility; but to serve all users—including broker channels—we need to design incentives carefully.

Matt Huang:
In times when spreads widen significantly, do market makers lose money? Do they use profits from calmer periods to offset losses during volatility?

Luana:
Now, because overall demand is strong, even with narrower spreads, they can still profit from the spread. But that’s precisely why we have incentive programs—to align the benefits. Maybe they lose a bit at times, but overall, the returns are worth it.

Matt Huang:
So your goal is to keep spreads tight across major markets at all times.

John Collison:
Meaning, maintaining narrow spreads continuously, which requires careful design.

Tarek:
Exactly. It’s difficult, but more interestingly, prediction markets are unique because liquidity doesn’t mainly come from traditional market makers but from ordinary people.

This loops back to the initial logic. We’ve solved the regulatory issue, but liquidity remains a challenge. Traditional exchanges like NYSE or CME spend years designing products and recruiting dozens of market makers in advance. That’s the traditional way. But prediction markets are different—they generate liquidity on a daily, hourly basis, constantly creating new markets. How to do that? The pace is highly dynamic, with new events constantly emerging.

John Collison:
Many find this counterintuitive—you need to incentivize market makers to provide liquidity. Because in stock markets, high-frequency trading firms invest heavily in low-latency links and compete fiercely. Is this because prediction markets are still early-stage, or is there a more fundamental difference?

Tarek:
It’s about the mode of liquidity provision—more immediate, more dynamic than traditional markets. Wall Street market makers don’t operate on such a rapid, real-time basis. You can’t expect them to set up a new desk in an hour to price political or cultural events.

The most interesting aspect is that prediction markets have a counterintuitive feature: often, the best predictors are not experts or authorities but ordinary people.

Matt Huang:
Internet anonymous forecasters.

Tarek:
Exactly. They are highly dispersed. It’s hard to say that a specific demographic is the best at pricing. Our success has come from cultivating a community of “super forecasters” who can quickly and efficiently price these events. Initially, it was hard to turn their interest into part-time work, then full-time. But as the market grew, this finally happened.

Luana:
Here’s a data point we can share: in the platform, the largest institutional market makers account for less than 5% of all matched orders.

Tarek:
That means they only provide a small portion of the liquidity.

John Collison:
Really?

Luana:
Yes. Less than 5% of all matched orders come from well-known large institutions. Over 95% are peer-to-peer or from small funds and teams.

Tarek:
That’s very rare in exchanges.

Matt Huang:
How many of these small, full-time market-making teams are there?

Luana:
About 2,000 people actively market-making on Kalshi.

John Collison:
Matt, you’re asking who exactly are Kalshi’s market makers—are they firms like Jane Street, Akuna, or just someone coding in a garage at 3 a.m. with Red Bull?

Tarek:
The garage coders are actually the most important.

Matt Huang:
And you said they account for 95%?

Tarek:
Yes. They are crucial because they price quickly, monitor the market constantly, and observe the situation in real time. They are the true frontline observers.

John Collison:
So, Kalshi is built on a community of people who keep a close eye on the situation.

Tarek:
Exactly. For example, in recent years, the best predictors of inflation on Kalshi weren’t big institutions or hedge funds but a guy from Kansas who had never traded before, just liked reading news, and had a feel for inflation trends. You find many such people on the platform. Thousands are fully dedicated, and tens of thousands have a broad knowledge of various topics, actively pricing and earning rewards for their insights.

Luana:
Let me tell you about my favorite user.

Tarek:
And I have a new favorite user I want to mention.

John Collison:
Great, each of you tell us about your favorite user.

Tarek:
I was just thinking about this morning. The person from the Wall Street Journal article about taxes—

Luana:
Oh, yes, he’s a very strong candidate. But my favorite is a huge Ariana Grande fan. He discovered Kalshi during election season but isn’t interested in elections at all. Later, he found our Billboard ranking markets—like those ranking charts.

John Collison:
He considers that a very important market.

Luana:
Yes, it’s very important to him. He’s earned over $150,000, paid off student loans, got a master’s degree, and bought a car. He had never traded before, never done anything like this, but has an intense, almost obsessive interest in music charts. For the first time, he could monetize this hobby. He’s also very friendly on Twitter.

Tarek:
I have many favorite users, but recently one stands out. Last week, The Wall Street Journal published an article about a tax accountant named Alan, who’s very active on Kalshi. When DOGE first appeared, many wondered how much it could cut costs. He dug into extensive tax laws and regulations, reached a very confident conclusion that the market’s expectations were unrealistic. He told his wife he was extremely confident about this trade. In a way, he’s like Michael Burry from “The Big Short,” but shorting DOGE. He heavily invested and ultimately won.

This shows prediction markets’ power: if you have specialized knowledge—often niche, like tax law—you can research and understand the world better, then profit from it. It’s fantastic.

John Collison:
Early AI applications included poker bots. Do you see any advanced AI market makers now? Because no one’s read all those tax laws, and now, besides Claude, maybe that’s less relevant.

Luana:
That’s true. Maybe we should ask AI.

Tarek:
We see more and more people using agents for trading, especially on APIs. It’s very obvious now.

John Collison:
Have you seen users successfully operating highly agentic market-making systems—where most processes are driven by AI?

Tarek:
Users generally won’t share their strategies in detail.

John Collison:
But you do talk to them?

Tarek:
Yes, John, we do. My understanding is that early Renaissance Technologies was already using some agent-based models. Of course, those were very early versions. Today, it’s an ongoing evolution, and AI is getting stronger. Many traders on our platform already have AI-based summary and judgment modules integrated.

John Collison:
I’m really curious about fully autonomous systems—no humans involved, ingesting information and quoting prices on their own. It feels like that’s coming very soon, if not already here. For example, Claude itself might be acting as a market maker.

Luana:
I’m not sure if fully unmanned systems exist yet. For instance, in international elections, many systems automatically translate documents, poll data, and perform analysis. But I don’t know if it’s fully automated.

Tarek:
We also don’t know if the models have reached that stage. Recently, we launched Kalshi Research, aiming to collaborate with research labs to establish a new benchmark—testing which models are better at predicting the future. This benchmark is unique because it tests understanding of the world, not just memory. I’m very excited about the results.

John Collison:
How will you evaluate them?

Tarek:
It’s not fully decided yet. The general idea is to run the same model on a set of markets for a month or two, then compare performance—accuracy, long-term PnL, etc.

Differentiating Prediction Markets from Entertainment: Trading vs. Betting

John Collison:
Another question about market design. In sports betting, there’s a well-known phenomenon: betting companies suppress “sharps”—those who are too smart or too good at betting. Many don’t realize that for betting companies, the ideal bettor is someone supporting their team; the worst are those exploiting mispricings in niche markets. Because they offer odds on thousands of markets, and if they get a few wrong, sharps will exploit those errors. They can detect signals from behavior—if you just bet on your team, that’s fine; if you’re too professional, they might ban you.

It’s interesting—you think you’re just betting on their odds, but if you’re too good, they don’t want you. Like Las Vegas casinos asking you to leave. Does Kalshi face similar