TL;DR
- The last playable round is on 12 August 2026. No new rounds run after that.
- Any referral and streak rewards you have earned will be sent to your Trepa wallet on 13 August 2026.
- The app stays online until 30 September 2026 so you can withdraw. After that it will not be reachable.
- If you have a balance, withdraw it before then. Open the app, go to your wallet, send it to an address you control.
- If you miss the deadline, you can recover access via home.privy.io. This only works if you signed up to Trepa with an email address.
We were not able to find product-market fit. The crypto market was difficult, but that does not really change the assessment. Below is what we got wrong, in case any of it is useful to someone building something adjacent.
What we bet on
Trepa paid you for how close you got, not for which side you picked. You named a number, and your share of the pool scaled with your distance from the outcome.
We were betting that people want to predict a number rather than pick a side and get paid for accuracy. Every incumbent asks for a discrete answer: yes or no, up or down, over or under. We were also betting that a third mechanic could take hold at all, alongside traditional trading and binary prediction markets.
We chose pari-mutuel knowing the cost. No fixed odds, so you cannot know your payout when you commit, and no way out of a position once entered. We accepted both because scoring with pro-rata distribution was the only structure we found that rewards numerical accuracy fairly and stays legible to an ordinary user.
Demand for one thing is not demand for another
We implicitly reasoned by analogy. People bet on uncertain outcomes at enormous scale, therefore people will want to bet on precision. That does not follow.
In a binary market your payout depends on being right. In Trepa it depended on being more accurate than everyone else in the pool, and on enough of them being there in the same window. Different product, different psychology.
Cognitive load will explain at least some of it. Choosing a number to decimal precision asks more of a person than choosing yes or no, and it probably removes some of the fun. We saw a version of this earlier, when a time decay weighting we added made payouts harder to understand without moving retention. Our read now is that in consumer betting, simplicity may be the whole product.
The trading mental model is also dominant and we could not re-educate around it. Users arrive expecting to trade an asset the way they do on Polymarket or a perp venue. Asking them to think in accuracy-scores and a contest format rather than in position did not seem to transfer.
Our growth plan diluted the thing we needed
A pari-mutuel pool with proportional scoring needs a crowd predicting in the same window. Thin pools mean small payouts, unstable variance and a worse experience for everyone in them.
We ran one live hour a day, 1 to 2pm UTC, with sixty-second rounds. That concentrated users, but it required them to build their day around us. Every day. Retention degraded under that ask, and degraded more the longer someone had been with us.
The scaling plan was more assets and more time windows. We recognized the contradiction too late: every unit of expansion dilutes the concurrency the mechanism depends on. We treated it as a tension to manage later. In the end we could not see a version of scaling Trepa that did not first make Trepa worse.
Reaching people was harder than we assumed
Referrals did not work and paid KOLs returned almost nothing. B2B2C through builder codes and an embeddable widget looked like our best channel on paper, but it ran into a wall: every partner conversation required us to justify the 1 to 2pm UTC window, and the answer was that the window exists because we need concurrency. Explaining your structural weakness as part of the pitch is not a strong position.
We were also wrong about the market. Geographic restrictions removed a large share of the addressable audience before we started. Onramping is a brutal hurdle for any crypto consumer app. And reported crypto DAU and MAU figures, including from sources we trusted, are softer than they look. Our working conclusion is that the genuinely active audience for a product like this is in the hundreds of thousands at the moment, not the millions.
Thank you
To everyone who showed up at 1pm UTC every day, and predicted through however many rounds. Some users did 2,000+ rounds. You were the reason any of this was worth building.
To @anamansari062, @reubence, @kioLmao, @0xapacx who built this. All of them are looking for their next mission, so if you are hiring, get in touch with them or with us.
To @CosmicDude3000, @0xllamii, @uretsky17510, @a_yo_figma whose work is still in what we shipped.
To our investors, who backed an unproven mechanism, UI/UX, and two founders. You took the risk knowingly and you deserved a better outcome.
To the @SolanaFndn, @solanamobile, and to the @superteam eco, with special shout outs to @SuperteamKorea, @SuperteamSG, @SuperteamDE, @SuperteamIN, and @SuperteamBlack. The support we had was not the thing that was missing.
To our infrastructure and security partners (@AdevarLabs, @PhageSec), who kept our product running and audited without drama.
To the friends who gave us feedback we did not want to hear, and kept giving it.
We still think there is something in the argument that people will pay to be measured on how close they get. We were not the team that found the way to build a business around it.
Jong and Leon
