Every moving part, written down.

The front page keeps the mechanics out of the way on purpose. This page is the opposite: the whole machine, in order, including the parts that are inconvenient for us. Where a number appears, it comes from the contract or from a measurement, and the measurement is named.

Revised whenever the contract changes · numbers verified on chain 4663

Section 1What an offering is

An offering is a subscription with a deadline. People send ETH into a contract during a window. If the founder fires the listing before the window closes, the contract spends that money buying the token on the launchpad — inside the listing block itself — and every subscriber owns a share of what was bought. If the listing never happens, everybody takes their money back.

Nothing here is custodial. Subscribed ETH sits in the contract from the moment it arrives until it is either spent on the listing or returned to the wallet that sent it. There is no third state, no treasury, no multisig, and no human step in between.

The life of an offering

1. CreationThe founder writes the terms into the contract: the window, the caps, the rounds and their weights, the vesting, the buyers' share of the bag, the floor on what the money must buy, and the launchpad configuration the money will be spent through. After this call, none of it can be edited. Not by the founder, not by us.
2. SubscriptionAnyone may send ETH between opensAt and closesAt, subject to the minimum, the per-wallet maximum and the hard cap. Each contribution is weighted by the round it lands in. Anyone may also withdraw in full, at any moment, until ten minutes before the deadline.
3. The listingThe founder calls launch. It only goes through if the goal was met and the window is still open. One transaction deducts the platform fee, buys the token, deploys the fee vault, checks the bag against the floor, and credits any change back to the subscribers.
4. Or the deadlineIf the founder never fires, or the goal is never met, the window simply closes. The exit reopens permanently and everybody withdraws the exact amount they put in. The founder gets nothing: no tokens, no fees, no consolation.
5. Afterwards, indefinitelySubscribers pull vested tokens as they unlock and pull dividends as the pool generates fees. There is no end date on either. The contract does not stop caring about a sale once it has listed.

What an offering is not

• It is not a promise of a fixed number of tokens. See §3 — this is the single most important design decision in the whole system.
• It is not a yield product. Dividends are a share of trading volume, and volume is usually zero. Nobody here will quote you a percentage.
• It is not a claim on the founder. The contract does not know how to sue anybody. What it enforces, it enforces mechanically; everything else is trust, and we will tell you which is which.

Section 2Rounds and weights

An offering has one, two or three rounds. A round is not a price. It is a weight — a multiplier applied to the ETH that lands in it. Your share of the offering is your total weight divided by the total weight of everybody.

A round is defined by two numbers: its weight, and the cumulative book size at which it closes. Round I with a cap of 1.0 ETH means "the first 1.0 ETH raised, by anybody, in any number of contributions". Rounds do not have their own clocks; they fill in order, and they fill by money, not by time.

What the contract enforces at creation

• At most three rounds.
• Every weight is between 1.0× and 3.0×. Nobody can mint a 50× founder round.
• Weights never improve: each round must be worth the same as or less than the one before it. Without this rule a founder could hang back and take the best deal at the end of the queue, which is the oldest trick in the presale book.
• Caps strictly increase, and the last round's cap must reach the hard cap — so there is no band of the book that no round covers and no way to reach the hard cap through a gap.

Contributions that cross a boundary

They are split at the boundary and weighted piecewise. If the book stands at 0.4 ETH, round I closes at 1.0, and you send 0.8 ETH, then 0.6 of your ETH is weighted at the round I multiplier and 0.2 at the round II multiplier. There is no rounding trick and no rejection: the arithmetic is weightFor(), it is a view function, and the app calls it before you sign so the number you are shown is the number you get.

What happens when somebody leaves

The exit is all-or-nothing: you withdraw your entire subscription, your weight goes to zero, and the book shrinks by exactly what you put in. Since rounds fill by book size, your departure reopens the cheap capacity you were occupying for whoever comes next. If Bruno exits in the example above, the book falls from 3.0 to 2.2 and the next 0.8 ETH is weighted from 2.2 upward — at round III, not at Bruno's old blend. Weights are a function of where the book stood when the money arrived, and nothing else.

The corollary matters if you are thinking of exiting and re-entering: you cannot get your old round back. Re-subscribing is weighted at the book's current position, which is normally worse.

Section 3What your allocation is actually a share of

You own a share of the tokens the offering bought. Not a promised quantity, not a price, not a rate. This is worth being blunt about, because it is the difference between a presale that can fail its last customer and one that cannot.

Where the raise actually goes

Three deductions stand between the book and the bag, and all three are visible before you subscribe. Using the 3.0 ETH book from §2:

At 2.9695 ETH the curve delivers about 68.4% of the one-billion supply, roughly 684,300,000 tokens. That figure is not a guess: the curve was measured on chain and follows ETH = 1.37 · p / (1 − p) for a fraction p of supply, accurate to within 0.01 percentage points at every size we tested up to 15 ETH. It is also why a bigger book does not buy proportionally more tokens — the last ETH into a large raise buys far less than the first.

The bag, split

The bag is divided once, at the listing, into the buyers' share and the founder's payment. The buyers' share is a term of the offering (buyersBps), fixed at creation, and the contract refuses to create an offering that gives buyers less than half. Say this one set 80%:

The remaining ~31.6% of supply is not held by anybody — it is still in the pool, which is what makes the pool a market. The founder's payment is held by the contract and pulled, never pushed; see §4 for why that is not a courtesy but a requirement.

Individual allocations, from the same example

The floor

At creation the founder commits minBagTokens: the minimum number of tokens the purchase must actually yield, or the listing reverts and the money stays refundable. It is published with the terms, so before you subscribe you can read the worst bag the offering is allowed to accept. If the launchpad changes its curve, its fee, or its configuration between creation and listing, this is the tripwire that catches it.

Section 4The listing block

The founder fires the listing with one call. What that call does, in order, inside a single transaction:

1. ChecksCaller is the founder; the sale is neither listed nor cancelled; something was raised; the window has not closed; the goal, if any, was met.
2. Books the platform fee1% of the raise is set aside inside the contract as owed to the platform. It never leaves as part of the purchase, and it is booked exactly once — a subtlety that cost us an audit round, see §10.
3. Deploys the fee vaultA fresh FeeVault, belonging to this sale and no other, which will be the launchpad's fee recipient for this token forever.
4. BuysThe rest of the raise goes into the launchpad through the shared bundler, which creates the token and executes the purchase in the same block, then points the pool's fee stream at the vault.
5. VerifiesThe tokens received are counted and checked against the committed floor. Under the floor, the entire transaction reverts and the offering is exactly as it was a second earlier: refundable.
6. Splits and settlesThe bag is divided into the buyers' share and the founder's payment. Any ETH the launchpad hands back as change is credited to the subscribers as a dividend — not to the founder, not to us.

Why the same block matters

The launchpad reserves the listing block: in the block the token is created, the only buy that can execute is the creator's, which here is the offering's. A sniper's first opportunity is the block after, by which time the offering already holds its tokens and the subscribers already own their shares of them. This is the one structural advantage the format has, and it is borrowed from the launchpad, not invented here.

The per-wallet window, and what it means for you

For two blocks after the listing the token itself refuses to let any wallet end up holding more than 5% of supply — except transfers coming from the token's deployer, which is the bundler, so the offering can take delivery of its entire 68% bag without tripping it.

The consequence is on the way out. In the example above, Alice's 8.88% and the founder's 13.69% both exceed 5%. If they tried to claim during those first blocks, the token would reject the transfer. So:

• The founder's payment is held by the contract, not sent at listing, and becomes claimable only after the window has passed with a margin. Pushing it during the listing would make any large founder share revert the entire listing — a defect we shipped once and an auditor caught.
• If your own allocation is over 5% of supply, your claim will revert during the window too. Wait a few blocks. The app does this for you; if you are calling the contract by hand, this is the reason your transaction failed.

Every reason a listing reverts

In all seven cases the money stays exactly where it was: in the contract, refundable, in the subscribers' names.

Section 5Vesting

An offering may release the buyers' tokens gradually. Three numbers, all fixed at creation and all visible before you subscribe:

Unlocking is tge + steps × step, capped at 100%, where steps is whole blocks elapsed divided by the step length. It is a stair, not a ramp: nothing accrues between steps. Claims are cumulative — the contract remembers what you have taken and gives you the difference, so claiming early costs you nothing except gas.

Two things vesting does not touch. It does not touch the founder's payment, which is governed by the per-wallet window instead and is otherwise released in full. And it does not touch dividends: you earn on your whole subscription from the first block, unlocked or not. Vesting slows down what you can sell, not what you are paid.

Section 6The dividend, and how it is funded

Every trade in the pool pays a fee. That fee is the only source of dividends in this system. There is no emission, no treasury top-up and no rebate — if nobody trades, nothing is paid, and no amount of clever accounting changes that.

The path a fee takes

1. The pool accrues itThe pool sits in the 1% fee tier. Fees accumulate in both assets — ETH on one side, the token on the other — depending on which direction each trade went.
2. The launchpad takes its cutThe launchpad's locker keeps 30% before anything reaches us. This is not our fee and we cannot waive it. See §7.
3. The vault receives the restThe remainder is paid to this offering's FeeVault and nothing else is ever paid there. One vault per sale is what makes the accounting exact — there is no shared pot to misattribute, and nobody triggering a collection out of turn can strand or misdirect it.
4. Anyone harvestsharvest is permissionless. It pulls the collection, drains the vault, unwraps the ETH side and splits everything. It can only move money towards the people already owed it, which is why it does not need a guard. If there is nothing to collect it does nothing and says so.
5. You pull your shareCredited through an accumulator, so an offering with five thousand subscribers costs the same to distribute as one with five. Nothing is pushed at anybody.

The split

Those percentages are constants in the contract. There is no setter, no governance and no owner who could add one. The token side goes entirely to buyers because it is the same asset they already hold a claim on — it arrives as extra tokens attached to your next claim, not as a separate thing to remember.

Three properties worth knowing before you plan around this

Your share is frozen at the listing

Weights cannot change after the listing, because the exit closes at the listing. Whatever fraction of the book you held in that block is the fraction of every future dividend you will receive, forever.

The dividend follows the wallet, not the tokens

This is the most misunderstood part of the design, so plainly: selling your tokens does not end your dividends, and buying tokens on the open market does not start any. The right belongs to the subscribing address. It is not a token, it is not transferable, and there is no function anywhere that moves it. If you want the dividend, subscribe; if you want the dividend and someone offers to sell you "their position", they cannot deliver it.

Dust is carried, not swallowed

Integer division leaves a remainder on every harvest. That remainder stays booked as buyers' money and is added to the next harvest's pot rather than being split again or written off. Over thousands of harvests it comes out to a handful of wei in somebody's favour, and we would rather it be yours.

One loose end the contract also closes: if the launchpad ever routes tokens to the bundler after the listing, sweepBundler brings them in and credits them to buyers on the token side. It is permissionless, and it is the only path those tokens have.

Section 7Fees, in full

Every fee in the system, who pays it, and when. There are no others. If you find one this table does not list, the table is wrong and we want to know.

Two structural limits sit behind the table. Subscribers are always entitled to at least half the bag — the contract refuses to create an offering that gives them less. And the platform address is fixed when the contract is deployed; it cannot be repointed, and it has no authority beyond being an address that is owed money.

Section 8The emergency exit

Before the listing you can withdraw everything you subscribed, in one call, at any time. It is not a request, nobody approves it, and it does not depend on the founder being reachable, alive or cooperative. The money was never anywhere else.

Why the ten minutes exist

Without them, one wallet could occupy the entire hard cap for free, keep everybody else out of the book for the whole window, and walk away one block before the founder fires — killing the offering at zero cost and zero risk. The lock is the smallest window that makes that attack cost something: the attacker has to be genuinely willing to be listed. Ten minutes is short enough that nobody is trapped in a meaningful way and long enough that the founder can act on a book that will still be there when the transaction lands.

The lock is also the one moment in the whole flow where you are not fully in control of your money, so we would rather point at it than bury it. The offsetting rule: if the deadline passes without a listing, the exit reopens and never closes again. There is no window in which the money is stuck with no way out.

Cancellation

The founder can cancel at any time before the listing. Cancelling is permanent — it cannot be undone, no further subscriptions are accepted, and it reopens the exit for everybody immediately, including during the final ten minutes. It is the founder's way of saying "this is not happening, take your money now" without waiting for the clock.

Withdrawal is all-or-nothing: there is no partial exit. And after the listing there is no exit at all, in any circumstance, for anybody — the ETH is not ETH any more, it is a pool and a bag of tokens with your name on a share of them.

Section 9What the contract can and cannot do

The useful question about any contract holding other people's money is not what it promises but what it is capable of. Both columns below are statements about code, and both are checkable in Schedule B.

One contract, many offerings

A single deployment escrows every live offering at once. That is efficient and it is also the sharpest edge in the design: an accounting error in one sale could reach into another sale's escrow. It has happened — see the second audit round in §10, where a platform fee counted twice would have paid one sale's subscribers out of every other sale's money. Every balance the contract reads is now measured as a delta against a baseline taken in the same call, never as a share of the total balance. If you audit one thing in this contract, audit that.

What it cannot do for you, either

• Make the token worth anything. It buys a memecoin; the market does the rest, usually downward.
• Force a founder to list, or to be interesting after they have.
• Stop the founder from selling their share the moment the restriction window drops.
• Get your money back after the listing.
• Recover a lost wallet. The subscription, the vesting and the dividend right all live at that address and nowhere else.
• Do anything at all about the launchpad's decisions — see §11.

Section 10Audit history, and what was fixed

The contract has been through two adversarial review rounds, each run by several independent reviewers working from different angles, with no instruction to be kind. The first round found five critical defects. Naming them is the point of this section — a list of what was found is worth more than a badge saying it was.

First round — five criticals

Second round — the fixes were audited too

The second round confirmed nine of twelve fixes and found two regressions introduced by the fixes themselves. The first: the platform fee was being counted twice — once as fee and again as change — which would have paid one offering's subscribers out of every other live offering's escrow. The second: a fee route that, as written, could never have been triggered at all, quietly paying nobody. Both are fixed.

We are recording this because it is the honest lesson and not a flattering one: patching money code without re-auditing the patch is the same as not auditing. The second round existed only because we assumed the first round's fixes were safe, and they were not.

Section 11What we don't control

The token, the pool, the curve, the fee schedule and the fee locker all belong to the launchpad. We are a customer of it. That dependency is load-bearing and you should price it.

What could change, and what it would do

• The launchpad raises its cut of the trading fee. Dividends shrink, proportionally and immediately. We cannot stop it and cannot opt out. The measured 30% is what it was when we measured it, not a guarantee.
• The launchpad changes the curve or the listing fee. Existing offerings buy a smaller bag than modelled. This is exactly what the committed floor is for: under the floor, the listing reverts instead of proceeding on worse terms.
• The launchpad changes the per-wallet window. A longer window delays large claims. Nothing is lost; claims simply revert until it lapses.
• The pool goes illiquid, or the token stops trading. Dividends go to zero and stay there. Nothing breaks; there is simply nothing to distribute. This is the ordinary outcome, not the exotic one.
• The launchpad's own contracts fail. A listing that reverts leaves the money refundable, which is the good case. A pool that misbehaves after the listing is not something this contract can undo.

What none of them can do is reach your subscription. The launchpad never holds it. Neither do we. The worst a change on their side can do before the listing is make the listing fail — and a failed listing is a full refund.

Section 12Plain questions

Schedule AParameters and limits

Fixed in the contract, identical for every offering

Chosen per offering, frozen at creation

The launchpad's numbers, measured on chain

Schedule BEvery function, and who may call it

The complete external surface. There is nothing else — no admin namespace, no proxy, no initializer, no emergency function. If a capability is not in this table, the contract does not have it.

Absent from that list, deliberately: any owner, any pause, any upgrade path, any setter for a fee or a split, any function that moves subscribed ETH to a third party, and any way to redirect a pool's fee stream away from its vault.