One flow in.
Fixed bands out.
A fixed share of every trade enters the prism as one undivided flow. It leaves as three bands, and the angle each band leaves at was set at deploy. There is no function that re-aims them, because the function was never written.
A fee with no fixed angle is just a treasury with better manners.
Most protocols take a cut and then decide where it goes. The decision sits behind a multisig, a governance vote or a variable nobody reads, and it can be re-pointed on any afternoon. The cut is real. The destination is a preference.
The destination is a setting
A treasury address in a storage slot is one transaction away from being a different treasury address. Nothing about the fee itself stops that.
You cannot price a variable
If a third of the fee might become buybacks, or might become salaries, then the fee has no meaning you can hold in your head while you trade.
Fix the angle, not the promise
Refraction does not consult anybody. Give each band a destination at deploy, then delete the ability to change it. What is left is arithmetic.
The glass decides the spectrum. Nobody else gets a vote.
A prism does not choose where each colour goes. The apex angle and the index of the glass decide it, and they decide it identically every time, for everybody, forever. PRISM is that idea applied to a swap fee.
The split happens inside the swap
The rate comes off in the swap path, not in a separate transaction a keeper has to remember to send. There is no queue, no batch and no window where the fee sits somewhere waiting to be routed.
Each band is paid to its destination in the same call. If the swap succeeds, all three bands landed. If any band would fail, the swap reverts and nothing moved at all.
| Wavelength | Index n | Deviation |
|---|
Snell at a 60 degree apex, Cauchy index. True spread across the visible band is about 1.6 degrees, so the fan in the render above is opened by a constant factor to be visible at this size. The figures in this table are not.
Three destinations. Set once, at deploy.
The split rate is a parameter. The way the split divides is not. Fifty, thirty, twenty, in that order, to those three places, on every trade.
Band one, into depth
Paired into liquidity and the receipt burned in the same call. The depth is real, it is on the book, and no address holds a claim on it.
Band two, into the burn
Buys on the open book at whatever the book says, then destroys what it bought. It pays the same price as anyone else trading in that block.
Band three, into the reserve
Held on chain in a contract with no withdraw function. Visible to everyone, spendable by nobody, including whoever deployed it.
Set the volume. Read the bands.
Nothing here is a projection. It is the split arithmetic run on a number you choose, which is the same arithmetic the contract runs on a number the market chooses.
The split rate is the only thing the glass changes. The fifty thirty twenty division is the same in all three.
Figures are the split arithmetic only. They assume the volume you set actually happens, which is the one thing no contract can promise.
Three glasses. Denser glass bends the band further.
A pool picks its glass once, at deploy. The index is written into the pool and the pool cannot be re-cut afterwards, so what a trader reads on day one is what a trader reads on day four hundred.
- The light setting. Lowest drag on the trader.
- Suits pools that expect high turnover and thin margins.
- Same three bands, same fifty thirty twenty.
- The default. Twice the depth per unit of volume.
- Suits a pool that wants its book to thicken visibly.
- Same three bands, same fifty thirty twenty.
- The heavy setting. Highest drag, fastest accumulation.
- Suits a pool that would rather have depth than volume.
- Same three bands, same fifty thirty twenty.
One flow, three fixed angles, step by step.
It plays on a loop. Click a dot to jump to a stage, or pause it and read.
What the glass is mounted on.
Three pieces, none of them ours: the rollup it settles on, the unit it settles in, and the venue the liquidity sits in.
Robinhood Chain
An Arbitrum Orbit rollup, chain id 4663, EVM equivalent throughout. Testnet is 46646 and runs the same bytecode.
USDG, natively
The chain settles and charges gas in USDG, eighteen decimals. Nothing here has to be wrapped on the way in or out.
Uniswap v4 pools
Liquidity sits in ordinary v4 pools on the same chain, so any router, wallet or explorer reads the market without a special integration.
PRISM settles on Robinhood Chain.
A USDG native L2, an Arbitrum Orbit rollup, chain id 4663, fully EVM. The splitter is an ordinary contract with no owner, and every band payment is a public transaction anyone can replay.
Burned receipts and destroyed supply
Both are arithmetic on chain, not a figure this site reports to you.
All three bands run inside the swap
Capped gas, public calldata, no keeper and no settlement window.
No admin path exists
Nothing to pause, upgrade, re-aim or drain. The functions are absent.
USDG
Orbit
4663
EVM
The values you need before you point a wallet at it.
Click any value to copy it. Every endpoint below was answered by the chain itself, not read off a docs page.
Cut the glass once.
These are not defaults with a setter behind them. Each one is a constant in the deployed bytecode, which is a different claim, and a checkable one.
| Parameter | Value | Changeable |
|---|---|---|
| Band one share | 50% | No |
| Band two share | 30% | No |
| Band three share | 20% | No |
| Band destinations | 3 addresses | No |
| Split rate | Per pool, at deploy | No |
| Reserve withdraw | Absent | No function |
| Owner | None | Renounced at deploy |
Cut, then widen.
The glass
Splitter contract, three bands, no owner. Verified source published with the first pool.
First pool
One pool at one index, so the arithmetic is observable on a real book before anything else exists.
Three glasses
Crown, flint and sapphire available at deploy time, same bytecode, different constant.
Open the bench
Anyone can cut a pool without asking, since there is nobody to ask.
The three things a fixed split does not fix.
Depth is not price
Band one thickens the book. A thick book still moves down if enough people sell into it. Permanent liquidity is a floor under the spread, not under the chart, and anyone telling you otherwise is selling something.
No volume, no bands
Every number on this site is a function of volume. If nothing trades, the split takes nothing, the burn destroys nothing and the reserve stays where it is. The mechanism cannot manufacture its own inputs.
Immutable includes the mistakes
The same absence of an admin path that stops a treasury being re-aimed also stops a bug being patched. That is the actual trade. Read the source before you decide it is the one you want.
The questions worth asking.
Can the three shares be changed later?
Who holds the reserve?
Why three bands and not two, or five?
Does the split rate change with volatility?
What happens if a band payment fails?
Is there a token?
Cut it once.
Read it forever.
One flow in, three bands out, at an angle nobody can change after the glass is cut.