On April 2nd, 2026, I spent two hours in the Kalshi Discord. I wasn't there to pitch anything. I was there to read. I wanted to see, in plain speech, what actual users of the biggest CFTC-regulated prediction market were complaining about.
Fifteen minutes in, a user named "ice" asked: "how do i view my pnl graph this site is so hard to navigate." Someone else replied, casually: "most people use other tools."
That was the moment the thesis stopped being theoretical for me. Kalshi is a CFTC-regulated Designated Contract Market with a capable product team and real volume. And its own community, in one line, said the quiet part out loud: the people who are actually making money on this venue are doing their real work somewhere else.
That "somewhere else" does not exist yet. Not in any finished form. That is the gap. And the dominant thing being built into that gap right now is the wrong thing.
Everyone is building for the wrong user
Look at what's getting funded in the prediction-market stack in 2026. The pattern is obvious after a few decks:
- Market-making SDKs for proprietary trading firms.
- Order-book analytics dashboards aimed at funds with a six-figure seat fee.
- Cross-venue execution routers for professional liquidity providers.
- Bot frameworks for "algorithmic strategies on prediction markets."
Every one of those products is useful. I am not knocking any specific team. What I am saying is: they are building Bloomberg for a market whose first ten million users are going to look nothing like a Bloomberg user.
The people who made Polymarket do $3.6 billion of volume on the 2024 US election were not prop shops. They were twenty-three-year-olds with a take on how Michigan breaks. They were retirees who read three political newsletters a day. They were the poker-adjacent subreddit trying to calibrate. They were, in aggregate, the reason the Polymarket price called the race hours before the networks did.
If you had only served Bloomberg users in November 2024, you would have missed the event entirely.
Retail is the category driver
This is true in a way that is actually structural, not slogan-y.
Prediction markets work well when informed participants show up in size. Who has an edge on "what the Fed says next Tuesday" is not a hedge fund — it's people who read transcripts. Who had an edge on "will Altman come back as CEO" in November 2023 was not a quant desk — it was people with source networks that mapped to a few blocks in San Francisco. Who has an edge on "will Senator X vote for appropriations Y" is not an institutional allocator — it's a lobbyist or a Hill staffer.
The right-skewed distribution of edge in prediction markets sits, overwhelmingly, with retail. This category is structurally different from equities or fixed income, where institutions have systematic advantages in speed, data, and balance sheet. Prediction markets are built to be won by informed amateurs, and the amateurs know it.
Which is why the lack of a retail-first interface is such a specific and honestly funny hole in the stack. The people who will make the volume, who will find the mispricings, who will make this market mature — they are being served right now by, roughly, a spreadsheet and a group chat.
What retail actually needs
I want to be concrete, because "build a better interface for retail" is the kind of line that means almost nothing on its own. There are five specific things that are missing, and every single one of them is its own product category:
Cross-platform discovery. There are eleven venues that matter in this space. Kalshi, Polymarket, Gemini Predictions, Manifold, PredictIt, Smarkets, Myriad, SX Bet, Metaculus, Good Judgment Open, Insight Prediction. A real retail user today has to open each site independently to see what's trading. They miss ninety percent of the interesting markets because they're on venues they've never heard of. The unified search has to exist. This is not hard software. It is just software nobody has bothered to build.
Portfolio tracking that is actually correct. The same Kalshi Discord I sat in had a recurring complaint that the site's P&L glitched and wiped eighty percent of users' reported balances overnight. The fix is independent calculation, sourced from the trade tape. A retail user shouldn't have to take the venue's word for their own performance. They certainly shouldn't have to take four different venues' words.
Community that maps to positions. The real research in this market happens in Discord servers, Twitter threads, Substack comments, Telegram groups. None of that is linked to your actual positions. When ice asked about the PNL graph, ice did not get to see that seventeen other users also held the same market, or that a macro thread from the week before had priced the exact call. The social layer is extremely undersupplied for a category that is this research-driven.
AI research that reads your book. Generic chat interfaces are not useful here. What a retail trader needs is an assistant that knows exactly which positions they hold, can read the overnight news against those specific positions, and surfaces the one thing that actually changed the probability. Not "here are ten reasons the Fed might cut rates." One specific, calibrated take. Personal, sourced, timestamped.
Category-specific depth. A prediction market on "NVDA beats consensus EPS this quarter" is not the same kind of market as "Trump wins 2028." The first one needs options skew, insider flow, whisper numbers, segment guidance. The second needs polling aggregates, ground-game reporting, state registration data, cable share. A generic interface that treats both as a line item with a price is going to lose to an interface that knows what each category actually needs and pulls it in.
None of those five things exists well anywhere right now. None of them were designed to. Exchanges built exchanges — and exchanges are not retail-facing products by nature. Nasdaq is not a retail product. Robinhood is.
The Robinhood moment is due
Equities in 2012 were already a mature asset class. You could trade them on a dozen brokerages. Public data was abundant. What did not exist, for a twenty-four-year-old with four hundred dollars and an opinion about Tesla, was a product that treated them like a user rather than a compliance problem. Robinhood happened. You know the rest.
Prediction markets in 2026 look the same way. Mature exchanges. Deep liquidity in the top venues. Enormous informational content priced into the contracts. What does not exist, for a twenty-year-old with five hundred dollars and a specific take on the next jobs report, is a product that reads like a modern consumer interface rather than a regulatory filing.
Somebody is going to build that product. There is too much volume sitting in the category, too many underserved users, too obvious a gap. The only real question is whether that product ships in the next eighteen months (my call) or drags on for eighteen years (nobody's base case).
Tykhy is our shot
The part where a post on a company blog is supposed to get coy. I'll skip it. Tykhy is our attempt, and I'll tell you exactly what we think we're building:
A single workspace for the retail prediction-market trader. Eleven venues unified into one portfolio. P&L calculated independently against the trade tape, not trusted to the exchange. Research layered on top of your actual positions with an AI that reads the news against the markets you hold. Community channels tied to categories, so the thesis and the price sit in the same place. Alt-data surfacing where it matters — options skew next to NVDA-earnings markets, insider flow next to M&A markets, polling aggregates next to politics markets, on-chain flows next to crypto markets.
Every trade routes through the user's own exchange API credentials. We do not custody money. The user stays the principal, the exchange stays the custodian, Tykhy is a non-MTL routing layer — the same posture Plaid took for screen-scraped bank data. That's not a marketing talking point. It's a structural claim: when the next Mt. Gox happens in this category, and at some point one will, it cannot happen to Tykhy because we never held the keys.
We are nineteen years old, writing this from a dorm room. We are not the team with a Bloomberg deal. We are specifically the team that knows the retail experience because it is ours. Every complaint in every Discord we read is one we've had ourselves in the last twelve months. Every missing feature on the list above is missing from our own workflow. We are building what we would have paid for in January 2025 and couldn't find.
Why this works
The skeptic in the room is reasonable: "fine, but plenty of retail-first ideas fail. Why this one?" Here's what's different:
- The category is pulling the product. You don't have to evangelize prediction markets in 2026 the way you had to evangelize stock trading in 2005. The volume is already there. Polymarket 2024 was a cultural moment, not a niche experiment. Kalshi is a household name in political Twitter. The wind is behind us.
- The gap is retail-specific. Every existing tool in the stack serves institutions. That leaves an entire category for the team that builds the consumer product first. This is not a contested niche — it is an unclaimed one.
- Retail drives the informational edge. Our users' research loop is the category's price-discovery loop. Good retail tooling makes the entire market better, which makes our users more right more often, which compounds. That's not true of every consumer product. It's true of this one.
- The moat is proximity, not infrastructure. Anyone can write a router to Kalshi. What's hard is knowing what a retail user actually needs when they wake up and open their phone. That is earned ear-to-the-ground, day by day, out of our own users. A product team shipped in from Goldman cannot recreate it on a six-month timeline.
We are nineteen. We have been trading prediction markets for two years. We have read more Discord complaints about missing features than any prop desk ever will. If you are building or trading in this space, come say hi: hello@tykhy.com.
The first retail product in a new asset class is usually the one that defines the asset class. Schwab defined online brokerage. Robinhood defined commission-free. Coinbase defined crypto onboarding. Tykhy is going to define this one. Come be early with us.
— Ilhan