โ€น back to the game

Trust assumptions

What's proven by the code, and what you're still trusting. We'd rather you read this here than discover it on-chain.

PROVENEnforced by the program, verified by exploit tests
๐Ÿ”“ Your money always has an exit
Every market state ends in a payout or a refund, and the exits are permissionless: anyone can resolve after the sampling window, anyone can force-void after 24h of keeper silence, voids refund every stake, losers reclaim their account rent. A dead operator cannot strand your funds.
๐Ÿงช Every drain we found is fixed and re-attacked on every change
Three adversarial audit passes found real critical bugs โ€” oracle-steering drains, a forged-price drain, a round-recycling over-claim. Each fix ships with an exploit test that attempts the drain and must fail. The suite re-runs on every program change.
๐Ÿงฎ Payouts can't overdraw the pool
Parimutuel with floor-division math: the sum of all claims is provably โ‰ค the distributable pool, and the vault can never drop below rent while any position is unclaimed. The protocol holds zero inventory โ€” it takes a rake, it never bets against you.
โ›“ Settlement reads the chain, not us
Prices come from on-chain pool reserves โ€” the median of multiple time-spaced reads, each requiring a sustained confirmation, collected over a mandatory window. No off-chain price feed, no operator-submitted number, no single instant to snipe.
๐Ÿ™ˆ Betting is blind by design
Cohort waves lock before their member tokens are even known, and the short bet window closes before the typical dev-exit lands. What makes the game fair to you also makes it unrideable for insiders.
YOU ARE TRUSTINGThe honest residuals
๐Ÿ”‘ Us, via the upgrade + admin keys
The program is upgradeable and an admin key sets parameters and withdraws the buyback vault. That's the standard trust assumption of every non-immutable Solana program: whoever holds those keys โ€” us, or anyone who steals them โ€” could ship a malicious upgrade. If you don't trust the operator, don't bet more than you'd tip.
๐Ÿง‘โ€โš–๏ธ The keeper picks which graduations enter a wave
Member outcomes can't be forged (grad prices are read on-chain, minimum wave size enforced), but the keeper still curates which tokens fill a cohort. Bounded to griefing/steering โ€” not a drain โ€” and documented as the known residual.
๐ŸŒŠ Extreme manipulation on ultra-thin pools isn't free โ€” but isn't impossible
The median + sustain-confirm oracle made price-steering roughly 200ร— more expensive than a naive single read, and a liquidity floor gates entry. An attacker willing to hold a fake price against arbitrage across the entire sampling window still has a theoretical path on the thinnest pools. We raised the cost; we didn't repeal economics.
๐Ÿ“ The audits are ours, not a third party's
Multi-pass adversarial review with executable proof-of-concept exploits โ€” rigorous, but self-performed. No external firm has audited this code. We publish the full findings, including what stays open, instead of a green badge.
HOUSE DISCLOSURE
The house runs a market-maker that places a small bet on every tranche of every wave so no round ever voids โ€” there is always a counterparty, and the winning tranche always pays. These wallets bet dumb, fixed coverage decided before the wave opens; they carry no inside information and follow no signal. Their addresses will be published here at mainnet launch so you can watch them do exactly that.
Don't trust the copy โ€” verify the chain. Every settlement is an on-chain read anyone can replay, and every index call is locked on-chain before it resolves.

The program ID, full audit findings (including open residuals), and the house market-maker addresses will be published on this page at mainnet launch. Nothing here is financial advice; bet what you can afford to lose on hour-old memecoins, which is the entire subject matter of this site.