Whitepaper

Fees buy $ANSEM. Holders get it back.

Version 1.0 · 14 August 2026 · not yet launched

A description of what ANSEMFUND does, the measurements that forced each design decision, and the parts we have not solved. Every figure here was measured against mainnet; where something is assumed rather than measured, it says so.

1 · The problem

"Strategy tokens" pay trading fees back to holders as some other asset. The category has a consistent failure: the fees never reach anyone. Distribution thresholds are set so high that a coin dies before it qualifies, and the fee data supports that — we measured every Meteora Dynamic Bonding Curve pool that has ever existed, 1,559,237 of them, and ~98.6% never earn $250 in lifetime fees.

A second failure is subtler. A distribution based on who holds at one instant pays whoever happened to be there, so buying just before a snapshot and selling just after extracts the same reward as holding for a month. That is a transfer from patient holders to fast ones.

ANSEMFUND targets both: no distribution threshold on the pool, and rewards proportional to how long you actually held.

2 · The flywheel

The coin launches on pump.fun, which pays the coin's creator 0.30% of trading volume. That figure is measured, not read: pump's Global.creator_fee_basis_points says 5, which is vestigial. The authoritative value lives in the fee program's FeeConfig and says 30 bps — confirmed by decoding a real mainnet sale where the curve paid out 0.222932448 SOL and the creator vault received 0.000668798, exactly 0.30000%.

Fees accrue to a vault derived from the creator address. Anyone may sweep them. They are then converted to $ANSEM and distributed to ANSEMFUND holders by time-weighted balance.

3 · Time-weighted balance

A holder's share is their balance integrated over the epoch, divided by the epoch's length — not their balance at a chosen moment. Holding half the epoch earns half as much as holding all of it. Buying before a snapshot and selling after earns approximately nothing.

The computation is deterministic, which matters more than it sounds. Given (mint, startSlot, endSlot, exclusions) it enumerates token accounts, replays every balance change from transaction history, integrates balance × slots, and divides. No sampling, no wall clock, no randomness. Two strangers computing the same epoch get byte-identical results, which is what allows anyone — not just us — to publish a distribution and anyone else to check it.

Validated by replaying 240 real balance changes and confirming every reconstructed closing balance matched chain state exactly. That check is vacuous on a dead token with no activity, so it was run against one with real in-window trading.

4 · Cumulative roots

Each epoch publishes a merkle root committing every holder's lifetime entitlement, not that epoch's slice. A claim pays cumulative − already_claimed. Three properties fall out of that arithmetic rather than out of extra code:

• One claim settles every epoch. Skipping ten epochs costs nothing; the next claim pays all of them.
• Double-claiming is impossible without a flag to desynchronise. Re-claiming an unchanged cumulative owes zero.
• A malicious root cannot claw back. Lowering someone's cumulative pays them nothing; it cannot reverse a completed payment.

The economic consequence is large. The conventional design needs one rent-bearing account per claimant per epoch. Cumulative roots need one shared ledger, forever, which cut the marginal cost of a claim from $0.0924 to $0.0004 — a 246× reduction. On a real 528-holder curve that moves the share of holders for whom a $100 epoch is worth claiming from 13.4% to 56.6%.

The tree's structure follows Jito's distributor and is checked against vectors from its Rust across 92 cases. The leaf deliberately differs: it also commits the recipient's index, which is what makes a shared ledger safe. It is not Jito wire-compatible.

5 · Converting without an oracle

Turning treasury SOL into $ANSEM without trusting anyone is harder than it looks. A router CPI lets the caller supply the route, so the program must bound the price it accepts — and there is no oracle for a memecoin. An admin-set price reintroduces exactly the trust this removes.

Instead the treasury posts an offer and the market takes it. The price of SOL per $ANSEM opens unfavourable to the filler and improves with elapsed slots until someone finds it worth filling. Arbitrage competes it toward fair value. No oracle, no route to validate, no privileged actor.

Guards, all unit- and fork-tested: both legs settle in one instruction with the payload taken first, so a later failure cannot leave the treasury short; destinations are pinned by PDA constraints so a filler cannot redirect them; any single fill is capped at 25% of the spendable treasury; a cooldown separates fills; and the price range is required to open unfavourable — inverted, the first caller would take the treasury at the best price.

6 · Who can run what

We would rather name the exception than claim a trustlessness we have not earned. One hop needs us. It is the hop that keeps the money recoverable while the program is unaudited, and it is the first thing that should change once a formal audit is complete.

7 · What is not solved

• The published root is optimistic, not proven. It is defended economically — a competing post must double the standing bond and collects the superseded poster's — not cryptographically. A sufficiently funded attacker can outbid honest correctors. Doubling makes that geometrically expensive; it is a deterrent, not a proof. The genuinely trustless alternative, on-chain balance×time via a transfer hook, was rejected because transfer-hook tokens are poorly supported by the routers and charts a coin needs to be visible at all.
• 1,024 recipients per epoch. The ledger is a fixed array sized to stay under what a single instruction can create, so there is no grow path to get wrong. Above that, the only lever is a higher dust floor — fewer, larger payouts. Raising the cap requires a redeploy.
• Nobody is paid to crank. Whoever sweeps fees pays gas and receives nothing. The guarantee is that nobody can stop them, not that they are rewarded.
• Treasury solvency is not enforced on-chain. If a root promised more than the treasury holds, later claims would fail on transfer. Off-chain policy keeps roots within the balance.
• No formal audit. An AI audit has been run and its findings resolved. That is a floor, not a ceiling.

8 · The numbers

The launch-economics figures describe other tokens' history. They are published because they are the honest base rate any launch faces — including this one. They are not a forecast.