The Bank That Mints When You Leave

Most token designs assume you hold the token. This one assumes you do not.

In The Standard Reserve, the token is not created when you earn it. It is created when you leave. The whitepaper is blunt about it: actual tokens are minted only when a banker withdraws. Until that moment your position is a ledger entry inside a branch you operate, accruing against a supply that does not yet exist.

That single inversion is the seed of everything else in the system, and it is worth walking through slowly, because almost every other choice follows from it.

The shape of the thing

The protocol describes itself in four nouns. One asset, one market, one signal, one authority. The asset is $STANDARD. The market is a single Uniswap v4 pool against ETH. The signal is net ETH flow through that pool. The authority is the central bank, which is not a person or a DAO but four thousand lines of immutable code.

You enter as a banker. A banker holds a charter, a charter operates branches, branches earn the currency the bank issues. There are 1,000 Founding Charters and each can run up to ten branches.

A charter is not a share of the protocol. This is the distinction almost every explainer gets wrong. Issuance is split across all branches in the system, pro rata, so what your charter earns depends on how many branches it runs relative to every branch everyone else runs. It is a seat that pays in proportion to how hard it is worked, and it dilutes as other seats are worked harder. A charter is not one thousandth of the supply, and any number built on that assumption is a category error rather than a valuation.

The charters went out on 14 September. The whitelist paid a fixed 0.15 ETH, one claim per wallet. Every unminted charter then fell into a public Dutch auction that decayed for thirty minutes down to the whitelist price and stayed there, open to everyone, one charter per transaction, three per wallet at most. All 1,000 went. The public side cleared in about ten seconds. 601 whitelist, 399 public, 583.6 ETH in total.

Ongoing protocol revenue splits 70 percent to the vaults, 15 percent to permanent liquidity, 15 percent to the team.

The man who wrote the four thousand lines

The protocol was built by a developer who posts as 0xBeans, and the biography matters here more than usual, because a system with no governance is a system where the deployment is the last decision anyone makes on your behalf.

His public record is a decade of smart contract work rather than a pitch deck. Public reporting places him at Coinbase in its early days, as a co-founder of Frame, an NFT-focused rollup that was later acquired, and on Abstract, the consumer chain out of the studio behind Pudgy Penguins. Coverage of The Standard Reserve says the Uniswap Foundation covered the cost of its audits, which is an unusual line item for a project with no investors to bill.

His GitHub is the more telling document. DRIP20, an ERC20 with gasless streaming. Mirakai, an on-chain NFT game. IAmTheOptimizor, a public competition to see who could write the cheapest gas. GenesisAndConclusion, a blitz project to mint one NFT in the last proof-of-work block and another in the first proof-of-stake block, which has no commercial purpose whatsoever and is the single most characteristic thing on the list.

The mechanism underneath The Standard Reserve is not new either. It came out of a hackathon he entered in 2023 and then shelved. An earlier hackathon project of his, from 2022, became the mechanic behind FWA. He has a habit of building a complete economic machine in a weekend, putting it in a drawer, and taking it out years later when it is worth doing properly.

None of that guarantees anything. It does explain the texture of the design, which reads like someone optimising a system for its own internal consistency rather than for how it will be described on a podcast.

Policy without a committee

Base issuance is 700,000 $STANDARD a day, scaled by a policy multiplier. The multiplier opens at 1.0 and lives between 0.2 and 1.25. It reads exactly one input: whether net ETH flow through the pool over a three day epoch was positive or negative.

The asymmetry is the entire design. A negative epoch cuts the multiplier by 0.15, immediately. A positive epoch adds 0.10, and only starting from the second consecutive positive epoch. Contraction is fast and unconditional. Expansion is slow and has to be earned twice.

There is no vote and no discretionary committee. The owner can lower base issuance but can never raise it. That is the whole of monetary policy: a number that falls quickly, climbs slowly, and answers to flow rather than to opinion.

This is the part everyone compares to Olympus, and on the surface the comparison is fair. But Olympus paid you in tokens you already held, so every holder woke up each morning with more of the thing they were being asked not to sell. The Standard Reserve never hands you that. You accumulate a claim. The supply arrives only at the moment you give the claim up.

Growth costs supply

Branches past your first are bought with expansion licences, and the licence is the other half of the engine.

Licences are capped per day across the whole system, three per charter per day, sold by falling-price auction, paid for in $STANDARD, and burned on receipt. The floor sits at roughly two days of a single branch’s yield. Each day opens at twice the previous day’s closing sale and the distance to the floor halves every four hours, so a day nobody wants licences is a day licences get cheap.

Read that loop again, because it is the whole machine. Issuance creates supply. Growth destroys it. A banker who wants to get bigger has to burn the thing they are paid in, at a price set by how badly everyone else wants to get bigger on the same day. The protocol puts it in one sentence: the highest expected-value action available to an incumbent is also the protocol’s largest supply sink.

There is a second-order detail worth spelling out. A burn does not just remove tokens in circulation. It lowers the maximum supply that can ever exist, and it does so even for tokens that were never minted, because a licence is paid out of a ledger entry rather than out of a balance. The cap is not a fixed billion. It is a ceiling that only ratchets down.

Day one gave a clean reading of the loop. The licence auction opened at 12,000, cleared at 11,888.34, sold out in three minutes, and burned 1,192,771 $STANDARD. That is over one percent of genesis supply destroyed in a single session, paid for entirely with demand for the right to expand.

A separate charter auction exists for new seats, opening at three times the last sale rather than two. It starts at zero per day and stays there until policy turns it on.

The price of the door

Withdrawing is not a transaction. It is a resignation.

To take value out you retire a branch, permanently. Retire your last branch and the charter itself burns. The whitepaper calls this no revolving doors, and it means it literally: there is no version of this where you take profit and keep the vehicle that produced it. Once transfers are switched on, selling the seat whole becomes a second exit, but it is still an exit.

The cost of leaving is not a fixed number either. The resolution fee is quadratic in total withdrawals across the entire system over a rolling seven days. Half of every fee is burned. The other half is paid to the bankers who stayed.

That curve is the most interesting object in the whole design, because of what it does to a bank run.

In an ordinary run, the first person out is made whole and the last person out absorbs the loss. Running first is always correct, which is why runs happen. Here the fee rises with how many people are leaving at the same time, and half of what the leavers pay lands with the people who did not leave. Exiting alone is cheap. Exiting with the crowd is ruinous, and it funds the crowd that stayed. The payoff structure of a run is turned upside down.

Idleness has a price too. Thirty days without activity exposes a charter to a 70 percent revocation fee, with 2 percent paid as a bounty to whoever reports it. The number is deliberately worse than the 60 percent worst case for leaving honestly, so going dark is never the cheap way out. Staying active is free, and a zero-cost check-in resets the clock. This system does not want passive holders. It is explicitly hostile to them.

The three defences also fire together, which is the part a static reading misses. Capital leaving does not trigger one mechanism. It cuts issuance within an epoch, flips fee routing into buybacks, and raises the exit fee on the people causing it, all at once.

The pool only fills

The liquidity design deserves stating carefully.

The canonical market is a single Uniswap v4 pool, ETH against $STANDARD, one percent fee. The 100 million genesis tokens are the only pre-mint, and they are not paired with anything. They sit single-sided above the launch price with no ceiling, owned by the protocol, and can never be withdrawn. Buyers’ ETH becomes the liquidity beneath the price as it discovers. The launch price itself is a free parameter the owner sets, and it opened at 0.00001013592 ETH.

Outside liquidity is closed while the launch tax schedule decays and opens once it reaches its floor. After that, third-party positions are allowed, but withdrawing a position’s principal is taxed by the hook as the swap it completes, while accrued fees are never taxed. A range order is not a cheaper way around the trading fee. I checked the hook on chain that morning and found outside liquidity reverting, which is what the schedule says should happen during the launch window.

The vaults run the other side of the ledger. In a positive epoch, fee revenue accumulates as hard reserves, tokenized gold and comparable assets, plus permanent liquidity. In a negative epoch, the same revenue finances buybacks and burns, rate limited so it cannot be front run in a single block. Fees arrive in ETH whether people are buying or selling, which means the protocol converts volatility itself into balance sheet regardless of direction.

The document is not the policy

The project’s own framing is that there is no committee, only code. That has a consequence the marketing does not spell out: if the code is the policy, then the paper is not the policy.

You could watch this happen in real time on launch day. The version of the whitepaper published that morning described the anti-snipe tax as opening at 20 percent on buys and 40 percent on sells, decaying linearly. The contract I read on chain a few hours later was set to 90 percent on both sides, decaying on a four minute half life to floors of 2 and 3 percent. By that evening the paper had been rewritten to say 90 and 90, halving every four minutes, matching the chain. The same rebuild also added a section on outside liquidity that had not been there in the morning, and dropped a passage describing how mint proceeds were split.

No changelog, no version bump. I am not going to tell you why it moved, because I do not know and I have not reproduced the on-chain reading independently. The general lesson is the one worth keeping. A document describes intent at the moment it was written. Deployed bytecode is the only text that executes. When a project tells you the code is the policy, it is also telling you where to look, and on this one the paper spent launch day catching up to the contract rather than the other way round.

To be fair to them, the paper is unusually explicit about which dials still turn. Epoch length, tax rates within a ten percent ceiling, licences and charters per day, fee splits with the team share capped, auction floors and decay half-lives. The hard cap, the base-rate ceiling, the multiplier rule, the fee curve and the launch schedule are fixed. Two switches, charter transferability and permissionless execution of buybacks, only ever move in the direction of less control.

What could break

Three things are worth holding in view, and none of them are secrets. All three sit in the published parameters.

The largest burn mechanism has an expiry date. Every charter caps at ten branches, so saturating all 1,000 charters takes 9,000 expansion licences. At 100 a day that is ninety days of buying, minimum, and after that the licence sink stops unless the charter auction is switched on. The supply sink that makes the loop work is not permanent by construction. It is a phase.

The absence of a committee cuts both ways. Nobody can vote to inflate you, and nobody can vote to stop anything either. An owner still sets the launch price, can lower issuance, can turn the charter auction on, and can move epoch length inside the published bounds. A system with no governance has no veto in it.

And it does not run on Ethereum mainnet. It launched on Robinhood Chain, which most write-ups of the protocol leave out entirely. Every property described above is a property of contracts on that chain, with whatever that implies about the base layer underneath them.

What I actually find interesting

Not the yield. There is no yield to discuss yet, and anyone showing you one this week is showing you an assumption wearing a number.

What holds my attention is that the incentives here point at behaviour instead of ownership. You cannot buy your way to the top, because charters are soulbound until a switch that is not on. You cannot sit still, because dormancy is punished harder than leaving. You cannot grow without burning. And you cannot leave quietly, because the price of your exit is set by how many people are leaving with you, and it is paid to the ones who are not.

Whether that survives a real drawdown is a different question, and ninety days of issuance will answer it better than any model I could build. But as a piece of mechanism design it is the most coherent attempt I have read at the problem Olympus never solved, which was never emissions. It was the exit.

---

Disclosure: I hold a Genesis Charter, and I made a series of short films for the project. This is a description of how the system works. It is not investment advice and not a recommendation to participate. On-chain figures were measured on 15 September 2026 and will have moved since. Whitepaper references are to the version live that evening, which is not the version that was live that morning.