the documentation

begin with the way tokens actually die. they do not die of bad memes or slow weeks; they die when the liquidity that made them tradable is taken back out, and everything else is commentary. the market has evolved elaborate rituals to manage this fear: locker contracts with expiry dates, team promises with nothing behind them, vesting dashboards that measure nothing but hope. all of them share a structure, a human holding a right and pledging not to use it, and so all of them are bets on a person. this experiment removes the person. the future of this token's depth was not promised, it was computed, sealed into a program at genesis, and published to the slot. anyone with the deployment constants and a calculator can tell you the size of the floor on any date for the rest of its existence. the question the experiment asks is what a market does when the one thing every other market must trust is, here, simply known.

the reservoir is where it starts. at genesis a fixed quantity of liquidity was deposited into an account owned by the program itself, an address derived from the program's identity, deliberately constructed to have no corresponding private key, which means no signature can ever move it except the program's own. there is nothing to steal, nothing to phish, nothing to subpoena, and no late night on which a holder of the keys can change their mind, because there is no holder and there are no keys. the reservoir's balance is public, its address is published on this page's ledger, and its only exit is a single instruction whose behavior is arithmetic. everything this project claims reduces to one auditable sentence: the reservoir can empty only into the pool, only at the rate the law prescribes, and only in the direction of down.

the law itself is the oldest one in physics. the amount remaining in the reservoir at any moment equals the initial amount multiplied by two raised to the negative of elapsed time divided by a constant, the same form that governs the decay of unstable atoms, chosen not for poetry but for its properties. it is smooth, it is memoryless, and it is total: it specifies the state of the reservoir at every instant from genesis to the end of time, with no schedule to consult and no meeting to hold. the constant was fixed at deployment and is stored in the program's config account, where anyone can read it. on-chain the computation runs in fixed-point arithmetic with rounding rules chosen so that error cannot accumulate in anyone's favor; the dust lands in the pool. from those two numbers, the initial amount and the constant, the entire depth chart of this token's future falls out as a closed-form expression.

settlement is permissionless, and this is the part worth slowing down for. the program exposes a crank instruction that anyone on earth may call at any time. when invoked, it reads the clock, computes how much the law says should have left the reservoir since the last settlement, transfers exactly that amount into the locked position, records the new state, and pays the caller a small fixed bounty for their service. the crank does not decide anything; time already decided, and the crank merely collects. call it every block or once a month, the outcome converges to the same curve, because the law is evaluated lazily and owes what it owes. no operator runs this. the incentive runs it: as long as one stranger anywhere finds the bounty worth the transaction fee, the reservoir keeps its appointment with the pool. the experiment does not employ anyone; it employs arithmetic and greed, which have never once called in sick.

now the absences, which are the actual product. the program contains no instruction to refill the reservoir, so its contents can never be diluted into a larger promise. it contains no instruction to drain it elsewhere, so the pool is the only possible destination. it contains no pause, no acceleration, no retuning of the constant, no migration, no admin path of any kind, because those instructions were never written, and on solana an instruction that does not exist cannot be invoked by anyone, including the author. the upgrade authority is burned, which closes the last door: the bytecode as deployed is the bytecode forever. the audit for all of this is unusually democratic. you do not need to trust a firm's pdf; you read the program's instruction list, confirm what is present, confirm what is absent, and go about your day. what is not there is the guarantee.

the destination is a locked liquidity position, also owned by the program, also keyless. deposits arriving from the reservoir are added to the position and become depth: standing orders on both sides of the price that make the token tradable at size. the position earns the pool's ordinary trading fees, and those fees compound back into the position itself, so the floor grows by two flows at once, the scheduled one from the reservoir and the earned one from activity. like the reservoir, the position has no exit instruction. depth here is monotonic: it can arrive, it can compound, and it can never be recalled. the floor of this market moves in one direction for as long as solana produces blocks, which converts the most common catastrophe in this asset class from a risk into a mathematical impossibility.

the shape of the law rewards attention. exponential decay is front-loaded: half of the entire reservoir arrives in the first period, three quarters by the end of the second, seven eighths by the third. depth compounds fastest exactly when the token is youngest and most fragile, which is the opposite of every vesting schedule ever inflicted on a chart. but the tail is the more interesting half. the flow thins forever without stopping; there is no final deposit, no cliff, no date after which the mechanism is spent. a millionth of the reservoir is still scheduled to arrive in some distant year, and the crank will still pay whoever shows up to settle it. most tokens have a moment when the incentives end and everyone quietly leaves. this one's schedule has no last page.

the chart on the front page is the experiment's laboratory instrument. the dotted line is theory: the depth curve computed in closed form from the two deployed constants, drawable for any date, past or future, by anyone. the solid line is measurement: the actual cumulative settlements recorded on-chain. the claim under test is that the two lines never diverge, and the claim is falsifiable in the strictest sense, since a single missed or mismatched settlement would open a visible gap that no press release could close. the ledger row labeled deviation from the law is the same test expressed as one number, the difference between owed and settled, which should sit at zero forever. science is usually a metaphor in this industry. here it is a method: a prediction published in advance, an instrument anyone can read, and a result that does not care what anyone hoped.

the numbers on this site arrive by two roads. market data, the cap, the holders, the pool depth, the day's volume, and the trades along the bottom of the page, is read through the birdeye api every second, the same public data layer available to anyone. the reservoir's balance, the cumulative settlements, and the deviation figure are read directly from the program's own accounts on-chain. there is no privileged feed, no internal dashboard, no cleaner view reserved for the author. and where any value cannot yet be verified, the site prints a dash instead of a guess. the dash is a small discipline with a long reach: a page that will not fabricate a number when it is inconvenient is a page you can believe when the number finally appears.

which leaves the question the experiment was built to ask. fear of missing out is the engine of this entire asset class, and it has always operated in the dark, feeding on uncertainty about the one variable that decides survival: whether the floor will still be there tomorrow. here that variable has been removed from the realm of trust and moved into the realm of arithmetic. the floor's entire future is public knowledge, to the slot, forever. what remains uncertain is only demand, which is to say, only each other. maybe a market with a knowable floor compounds belief in a way this industry has never seen. maybe attention drains anyway and the reservoir spends decades pouring liquidity into a market nobody visits, the most solvent ghost town on the chain. either ending is acceptable, because either ending is data. do not trust the author, who is a stranger. trust the program, whose instructions you can read, and the chain, which does not care what anyone hoped.