A central bank with no one inside.
SPECIE issues a fixed-cap currency, watches one market, and moves one rate. Its constitution is a set of constants deployed to Base, no key can edit them, and no meeting could.
Overview
Four primitives define the machine. The unit: $SPEC, ERC-20, ceiling of one billion, permanent. The venue: a single ETH ⇄ $SPEC pool on Uniswap v4, carrying a custom hook. The input: net ETH crossing that venue. The operator: contract code that cannot be amended. Participants attach through charters; a charter runs branches; branches receive what the operator issues.
Inbound capital raises the issue and converts fee income into reserves. Outbound capital lowers the issue, points fee income at buybacks, and prices the door. Either way the system is fed, one state shrinks supply, the other grows the book.
The system map
Six objects, seven flows, the whole economy fits in one drawing.
| OBJECT | WHAT IT IS | IN SHORT |
|---|---|---|
| $SPEC | ERC-20 · 1B ceiling · 18 decimals | Created only on withdrawal; destroyed by license sales, buybacks, and half of every exit toll. |
| The pool | The only ETH ⇄ $SPEC venue, a hooked Uniswap v4 pool | All capital passes here; the hook meters it. |
| The operator | The central bank, unamendable contract code | Meters flow, sets the rate, splits the fees. |
| A charter | A soulbound seat NFT, running 1-10 branches | 1,000 issued free at genesis; later ones sold daily for ETH. |
| A branch | One claim on each epoch’s issue | Costs burned $SPEC to open; closing one converts its accrual to tokens. |
| The vaults | Two destinations for fee income | The reserve vault accumulates; the defense vault burns. |
A charter behaves like a firm holding up to ten seats on an exchange: seats are bought out of retained earnings, and the only dividend is handing one back. Returning the last seat closes the firm.
- Traders ⇄ the pool. Open to anyone; every swap pays the operator a toll in ETH.
- The pool → the operator. The hook nets inflow against outflow; nothing else is measured.
- The operator → branches. Each epoch’s issue, divided equally across every open branch.
- Bankers → the operator. Licenses for new branches, paid in $SPEC and destroyed.
- Bankers → the pool. Retired branches mint out their balances, which meet the market here.
- The auction → the operator. Seat sales in ETH, pooled with the tolls.
- The operator → the vaults. The epoch’s fee income, split by regime.
The unit
- Ceiling. 1,000,000,000 units, 18 decimals. No function exists that raises it.
- Genesis. 100M deposited one-sided into the v4 pool as protocol liquidity, full-range, owned by the code, with no withdrawal instruction anywhere in it.
- Allowance. The other 900M is everything the operator may ever issue. Exhausted means finished; from then on the economy runs on recycled fees.
Accrual is ledger-side; tokens materialize only at withdrawal. Three things destroy them, license sales (100%), defense buybacks (100%), exit tolls (50%), so supply satisfies one identity at every block:
A burned unit is gone for good, so the largest possible supply moves in one direction only:
Supply is the system’s own accounting, published as one number.
Whatever circulates equals what has left the bank, net of everything the bank has since bought back or destroyed. Anyone can read it, at any block, and know which way the economy is running.
The signal
Per epoch, the hook keeps two tallies: gross ETH paid in by buyers and gross ETH paid out to sellers. The difference is that epoch’s net flow; the policy signal sums the last two completed epochs:
- One point of measurement. Capital has a single door in and out of this economy, so the signal cannot be computed, or faked, anywhere else.
- Priced in capital. The unit is ETH, not trade counts or token volume. Bending the signal means pushing real money through a tolled door.
- Two clocks. The rate reads two whole epochs of history, so one manipulated hour moves nothing. Fee routing reads only the live epoch’s sign, so defense reacts within hours.
The rate
The press runs at 700,000 $SPEC per day before adjustment, scaled by the policy multiplier m. An epoch of d days issues
streamed second by second across every open branch, a branch opened at noon earns from noon. With N branches live, one branch’s day is worth 700,000 × m / N. The multiplier follows one rule, checked each epoch:
The asymmetry is deliberate. One bad epoch cuts the rate on the spot; a raise has to be preceded by two good ones. From launch, steady inflows need about nine days to restore the full issue and eighteen to reach the ceiling, while steady outflows ride the rate from ceiling to floor inside three weeks, an 84% cut to dilution exactly when the exit is busiest.
Cuts are reflexes. Raises are verdicts.
Every epoch lands in one of two regimes, expansion when the signal is positive, contraction when it is not:
| EXPANSION | CONTRACTION | |
|---|---|---|
| Issuance | steps up while inflows persist | cut at once |
| Fees | reserve vault, hard assets | defense vault, buy back and burn |
| Licenses | dearer, the floor tracks the rate | cheaper |
| Exits | 2% floor | crowd-priced, up to 60% |
| Rational move | build, every branch opened burns float | hold, leavers pay whoever stays |
Charters
A charter is the seat itself: an initially soulbound NFT whose holder is a banker, licensed to run branches and collect the issue.
1,000 Founding Charters at 0.01 ETH each, an allowlist portion and a public portion, one per wallet. The mint fee goes to the team; the genesis liquidity is owned by the protocol and cannot be withdrawn.
Supply resumes only through the daily Dutch auction, in ETH. Each day opens at 3× the previous close and decays toward an admin-set reserve floor; purchase is first-come at the running price, and the seat mints inside the buying transaction. The daily count is policy-set and begins at zero, seats are never unlimited. Proceeds join the fee engine like any other ETH.
A charter lives exactly as long as its last branch. Retire that branch and the NFT burns with it; the only road back runs through tomorrow’s auction. Seats do not revolve.
Branches and licenses
Every charter is born with one branch and can grow to ten. A branch is one claim on each epoch’s issue, and every branch past the first requires an expansion license from the daily auction:
| PARAMETER | VALUE |
|---|---|
| Licenses per day | 100 |
| Per-charter limit | 3 per day |
| Payment | $SPEC, destroyed in full |
| Open price | 2× the previous day’s closing sale |
| Floor | ≈ two days of one branch’s yield |
| Decay | exponential, across 24 hours |
Formally, with N branches live and Pstart = 2 × Plast:
The curve opens high on purpose and falls until someone blinks; where they blink is the market’s own estimate, and the floor rules out nothing except free seats. Since the issue streams by the second, a branch’s lifetime take is exactly proportional to its lifetime.
This is the supply sink at the center of the design: the most profitable act available to an insider, building, is also the protocol’s largest burn.
The daily auctions
One mechanism sells everything: a falling-price daily auction. Open high, decay for 24 hours, transact instantly at whatever the clock says, no bids, no escrow, no refunds, nothing to snipe. The two markets differ only in unit and destination: licenses cost $SPEC that is destroyed; charters cost ETH that funds the bank.
Double yesterday’s closing sale, or double the floor, if nothing sold.
Down the exponential curve of §7, toward ≈ two days of one branch’s yield.
First come, first served. Three per charter per day. Payment burns on receipt.
At 100 sold or 24 hours. Unsold licenses vanish; the last price paid seeds tomorrow’s open.
Triple yesterday’s closing sale, or triple the floor, which is the admin-set reserve.
The same exponential slide toward the reserve floor over 24 hours.
First come, first served. Seat and first branch mint in the buying transaction; the ETH routes to the fee engine.
At the day’s count or 24 hours. Unsold seats are never minted; the last sale seeds tomorrow’s open.
Price discovery is outsourced to impatience. Pay the morning premium for certainty, or gamble that supply lasts into the cheap afternoon, the price is wherever those two instincts balance.
Repricing is asymmetric by construction. Dying demand reaches the floor within a day, making contractions the cheapest time to build. A demand spike clears days of supply instantly while the open walks up to meet it, on the order of 100× in a week if it must, since each open can only multiply the last close.
Charters open steeper than licenses, 3× against 2×, because a scarce seat should chase demand faster than a daily commodity.
Getting paid
Yield accrues to the charter’s internal balance continuously. Taking it out means retiring branches, a payout and an amputation in the same transaction: the branch’s share of the balance mints to your wallet, and the seat that earned it is gone.
- Strictly pro rata. One branch of ten frees one tenth of the balance; all ten free everything, and burn the charter behind you.
- Net of the toll. The mint lands minus the resolution fee, locked at the rate in force when you commit.
- No extraction without contraction. Every withdrawal permanently shrinks your claim on all future issues.
The toll is congestion pricing on the exit itself. With W the tokens withdrawn system-wide over the trailing 7 days and D everything still on deposit:
A quadratic climb from a 2% floor to a 60% ceiling, saturating when a tenth of the bank heads for the door inside one week:
A run on this bank pays whoever refuses to join it.
A classic run rewards speed, first out is whole, last out is ruined, so running is always correct. Here the order is inverted: heavy exit volume raises the toll on the exiters themselves, half of every toll is destroyed, and the other half accrues to positions still open. The crowd behind you pays you for having stayed.
Withdrawals are never paused, gated, or queued, at any toll level. Price is the only instrument.
Inactivity
An idle charter is a leak: it draws issue away from bankers doing the work. The crowd is paid to plug it.
Thirty days without a transaction makes a charter reportable, by anyone.
The informant collects 2% of the dormant balance, capped at 100,000 $SPEC.
The absentee is charged 70%, set above the worst-case 60% exit toll, so going dark is never the cheap way out. Half burns; half pays the bankers still working.
Branches close, the charter burns, and the remaining 30% returns to the owner’s wallet.
Staying active costs nothing to prove: any interaction resets the clock, and a gas-only check-in exists for pure holders. Lost keys and tourists dilute nobody.
Treasury and defense
The hook levies 2% on buys and 3% on sells, in ETH, heavier during the first 24 hours after launch, to tax snipers. Every unit of protocol ETH, tolls and seat sales alike, is routed once per epoch:
| 70% | the vault matching the regime, the reserve vault stockpiles hard assets in expansions; the defense vault finances buybacks in contractions |
| 15% | protocol-owned liquidity, half swapped to $SPEC, paired, and added to the position forever |
| 15% | the team |
The reserve vault converts fee ETH into hard reserve assets, held by the bank itself.
The defense vault is the standing bid: it buys $SPEC on the open market and destroys every token it touches, in hourly rate-limited steps that cannot be baited into one blockable shot. Each tick, with vault balance V against pool reserves R, spends
bounding defense near 5% of pool depth per day at launch settings. Unspent budget rolls forward. The vault has no sell path, it is a one-way furnace.
Protocol-owned liquidity is monotonic. The genesis position plus each epoch’s share only ever grows: no withdrawal function exists to govern, subvert, or socially engineer, verify it onchain. Fees earned in $SPEC are burned outright.
What cannot change
There is no upgrade path. No proxy sits in front of the code, and no migration can move the state. What flexibility exists is fenced three ways:
The 1B ceiling · the 700k/day base rate · the multiplier rule · the toll curve · the 70% revocation charge · the 10% fee ceilings · the vault execution bounds.
Epoch length (1-7 days) · fee rates (≤10%) · licenses per day (≤2,000) · charters per day (≤100) · splits (team ≤20%) · auction floors and windows. Every change waits out a 48-hour timelock in public before it lands, and the delay itself is raise-only: it can lengthen, never shorten.
Charter transferability (off until on, then on forever) · permissionless vault execution (admin-cranked until opened to everyone, then open forever). Each flip surrenders control; none reclaims it.
A guardian address holds exactly one emergency power: halting auctions and vault purchases. It cannot stop a swap, cannot stop a withdrawal, cannot reach funds or parameters, and it can be renounced.
Transferable seats (future)
Charters launch soulbound. A one-way switch can later make them transferable, opening a second way out: sell the seat whole, branches, accrued balance, and all. A seat changing hands puts zero sell pressure on $SPEC; one banker replaces another, one for one, and the bank notices nothing.
Reinforcing loops
Four loops fall out of the mechanics above. None needs a token incentive, a lockup, or anyone’s goodwill.
Day-one settings
| PARAMETER | LAUNCH VALUE |
|---|---|
| Ceiling | 1,000,000,000 $SPEC |
| Genesis liquidity | 100M $SPEC, one-sided, code-owned, full range |
| Base issuance | 700,000/day × m |
| Multiplier m | 0.20×-1.25×, starts 0.80×, −0.15 at once; +0.10 from the 2nd consecutive positive epoch |
| Epoch | 3 days |
| Founding charters | 1,000, 0.01 ETH, one per wallet |
| Charter auctions | start at 0/day, policy-enabled, Dutch, in ETH, floor = reserve price |
| Branches per charter | 10 max |
| Expansion licenses | 100/day, Dutch, in $SPEC, destroyed in full, max 3/charter/day |
| License floor | ≈ two days of one branch’s yield |
| Trading fee | 2% buy / 3% sell, heavier in the 24h launch window |
| Fee split | 70 active vault / 15 POL / 15 team |
| Exit toll | 2%-60%, quadratic in 7-day system-wide exit pressure |
| Dormancy | 30 days, 70% revocation, 2% informant bounty capped at 100k |
| Buyback execution | hourly rate-limited ticks, ≈5% of pool depth/day at launch |
| Network | Base |
The full set at launch. Tunable values move only inside the hard bounds of §12.
Risk statement
SPECIE is a live experiment in self-referential monetary design, not an investment product and not a financial institution of any kind. It pays out of what flows in; when flow reverses, it gets smaller. The mechanisms in this paper are brakes, not guarantees, total loss is a possible outcome. Nobody can freeze your withdrawal, and nobody can hold your price up. Reserve assets are protocol property and are not redeemable. The contracts are final. Audit the chain, not the prose.