The first AI assisted aggregator on Robinhood Chain · 4663
Read the skybefore you trade.
Galaxy Exchange indexes every liquidity pool on Robinhood Chain, splits one trade across the venues that price it best, and shows the arithmetic it used to decide. Ask it in words, and the copilot quotes, fills the ticket, and stages the swap for your wallet to sign.
Read from the running index at block 57322655, 106 behind the head.
01The field
Nobody publishes a listof every venue.
So Galaxy Exchange does not read one. It scans the chain for swap events by their topic hash with no address filter at all, which means it finds venues that were never configured anywhere, including forks with no name, no documentation and no front end of their own.
Twenty two came back. Three carry names, taken from Uniswap’s published deployment file for chain 4663 and each confirmed against the chain before being written down. The other nineteen are shown as addresses, because inventing a name for a contract is how a router starts describing trades it cannot actually account for.
- 552
- Uniswap v3 pools
- 166
- Uniswap v4 pools, one singleton
- 71
- Uniswap v2 pools
- 19
- venues carrying no name at all
Fees across those venues run from nothing to 99.08 percent. Eight pools charge more than four fifths of the trade. They are indexed and reported, and a pool charging above a quarter is never routed through.
02The route
One trade rarely belongsin one pool.
A pool prices the beginning of a trade well and the end of it badly, because every token taken out moves the price against the next one. Past a certain size the cheapest thing to do is stop buying from one pool and start buying from several at once.
Galaxy Exchange searches the paths between two tokens, then decides how to divide the trade between them. The division is never even. A leg gets the share it earns, and a route that would revisit a token is thrown away rather than quoted, because that shape is an arbitrage loop wearing a swap’s clothes.
Hook mediated pools are asked the chain directly, at four fixed shares. A share the chain was not asked about is not offered rather than guessed at.
Read the size of the advantage honestly. It is nothing at small sizes and about fifteen basis points at large ones. It is exactly zero whenever a single pool can absorb the whole trade, and it grows the moment one cannot.
An earlier run of the same ladder, hours apart, showed figures four times larger. Nothing in the router had changed. Liquidity had moved. A single comparison at a single moment measures nothing, which is why every figure this project quotes carries the block it came from.
03The proof
Checked against the chain,to the wei.
144 / 144
quotes matched Uniswap’s own QuoterV2 exactly. 120 of them crossed at least one initialised tick.
12 / 12
real Uniswap v4 swaps replayed out of their logs and reproduced to the last wei.
7 / 7
routes the contract executed against live pool state, delivering exactly what was quoted.
Matching to the wei is the only standard worth holding. A router that is merely close delivers less than it promised, and a swap that delivers less than its minimum reverts and charges the trader gas for nothing.
The test that proves it is also what found the bug that made it necessary. The first execution came up 5.5 parts per million short, because Uniswap v4 charges its protocol fee on top of the liquidity provider fee rather than out of it, and the engine had been reading only one of the two.
04The instrument
Open the terminal.
A quote screen with the route drawn out, the price impact stated whether or not it flatters the trade, and every number tabular so it stops moving under your eye while it updates. Slippage sets itself from the quote and shows its reasons. A copilot sits beside the ticket: ask what a trade returns, tell it what to set up, tell it to go ahead, and your wallet asks you to confirm. Every figure it says came from the engine, and it cannot sign.
Quotes are live. Wallets connect. Swaps are not.
The router contract is written and it compiles to 7.7 kB, under a third of the size limit. It has been executed against live pool state at the block it was quoted for, and it delivered what was promised in every case. It has not been deployed to any network, and nobody outside this project has reviewed it.
It takes custody of a trader’s tokens for the duration of a swap, which makes it the most dangerous code here by a wide margin. Passing tests is not an audit and this project will not pretend otherwise. Until it is deployed, connect a wallet, quote against your balance, and let the copilot set the ticket. The swap control turns itself on the day the contract goes live.