The quote is the product
BTR is a venue. You show up to trade, or you leave liquidity, or you build against the contracts. What you meet, every time, is the same object: a quote that already knows the mark, sitting on inventory shared across every asset in the pool, leaning when the book leans.
That’s the product. The rest of an AMM is plumbing.
We run it as an AIMM, an Adaptive Inventory Market Maker. Shared liquidity, inventory that manages itself: that’s what makes a prop book good. The logic sits on chain, where you can read it: that’s what makes an AMM trustworthy. We kept both.
The leak is in the price
If the leak is in the quote, you fix the quote.
Ask someone who provides liquidity what they earned last month and you’ll get two numbers: the one the interface showed them, and the one sitting in their wallet. The usual explanation blames the fee tier. Wrong suspect.
The money doesn’t leak through the fee. It leaks through the price.
A constant-product curve names its price from its own reserves, which is another way of saying it names it from its own history. A concentrated position names it from a range somebody drew by hand, possibly weeks ago, possibly while looking at a market that no longer exists. Neither of them has the faintest idea what the asset is worth right now. So the moment the real price moves somewhere else, the pool keeps offering the old one, patiently, until somebody who does know turns up and takes the difference.
That person has done nothing wrong. They were shown a price and they took it. That’s what prices are for. The one who pays is the one who deposited.
There’s a name for this in the literature, loss versus rebalancing. On most venues it’s the single largest cost a liquidity provider carries. Ahead of gas. Frequently ahead of every fee they collect over the same period. We didn’t take that on faith: we measured it for two years, tick by tick, and fees lost to adverse selection in 14 of 14 market regimes. The industry’s answer has been to paper over it with emissions, which works beautifully right up until the emissions stop.
The quote sits on a live mark
Instead of pricing from its reserves, a BTR pool prices against a mark it keeps fresh. That mark can come from a high-frequency oracle, from a third-party pool, or from any other IExternalOracle implementation - an on-chain TWAP, a vault share price. Every pool we curate prices off signed keeper marks: a k-of-n quorum signs each push, and anyone can re-verify it off chain. When the mark moves it reprices every asset in the pool inside the same block. The window where you’re quoting yesterday to somebody who already knows today isn’t there.
One reserve backs every asset
The liquidity underneath isn’t pinned to a pair. One reserve backs every asset in the pool, so a dollar of deposits produces depth on every route through that pool instead of sitting idle on all but one of them. In theory and simulation, you’d need up to fifty times that TVL spread across a Uniswap V4 deployment listing the same tokens in multiple pools to match BTR’s quotes, and those pools would still pay LVR.
The factor depends on how many assets sit in each BTR pool, median swap size, and risk parameters, and we haven’t optimized for it. It also degrades under concurrent drains. That’s why the derivation is public. Check the maths. Come back if it doesn’t hold up.
What the spread charges for
The three things a BTR quote is built from, and what each one is paid for:
| the pool knows | it charges for | so the LP is paid for |
|---|---|---|
| how fast the asset moves | volatility | holding risk while the market runs |
| how old its mark is | staleness | the window an arbitrageur could exploit |
| how sure the feed is | confidence | quoting into genuine uncertainty |
None of those is a fee tier someone picked. They are terms in one spread, and the derivation is public.
The quotes lean with inventory
When the pool’s inventory tilts one way, the price improves for whoever pushes it back toward balance and gets worse for whoever leans on it harder.
Anyone who has run a book recognises it, because it’s the first thing you learn on a desk: you don’t quote both sides the same when you’re already long one of them. It has been almost entirely absent from AMM design. It’s the whole difference between absorbing adverse selection and charging for it, which is what lets depth hold without a drip of incentives underneath.
The same quote, both sides
A trader gets a quote that already sits on the mark. Spreads form around it, not around last week’s reserves, so you aren’t filling a price a faster book has already left behind. Depth isn’t trapped in a pair, so size finds the route you’re actually taking. Push the book further out of balance and the price moves against you. That’s how a desk is supposed to quote.
For someone providing liquidity, deposits are single-sided, so there’s no pair to assemble. Concentration happens on its own and stays wherever the asset is actually trading, tighter than the smallest tick Uniswap V4 will let you set. Nothing to rebalance. No chart to keep an eye on. The position is fungible and it doesn’t ask you for anything.
We didn’t fork any of it
None of this arrived out of nowhere, and we’re not going to pretend it did.
- Wombat worked out single-sided coverage and how to ringfence solvency asset by asset.
- Gyroscope showed what curated concentration looks like, and how to put idle inventory back to work.
- 1inch Aqua made the case for mutualising liquidity across markets rather than trapping it in pairs.
- DODO’s PMM replaced reserve-driven discovery with an oracle-anchored proactive mid.
- Swaap and the Solana prop AMMs demonstrated mark-centered quoting with volatility-aware spreads.
- Uniswap v4’s hooks showed how to make yield composable without pushing policy off-chain.
And underneath every one of those sits older work: Avellaneda and Stoikov on inventory reservation, Bergault and Guéant on oracle-referenced quoting, Milionis and his co-authors on precisely why a stale reserve quote ends up paying arbitrage rent.
We’re inspired by all of them and we’ve forked none of them. The curve, the inventory logic and the pricing policy are ours.
Live on Arc testnet
Put it together: an unusually capital-efficient AMM, running on the cheapest signed oracle on the EVM. We know how claims like that read. It’s why the capital-efficiency derivation is published, why the oracle’s gas numbers come with their own corrections attached, why the oracle is on chain, and why all of it is one click from this paragraph.
BTR goes toe to toe with the prop market makers taking share right now on their own foundations - high-frequency marks, tight inventory, shared liquidity - with the curve and the pricing logic on chain, where theirs are closed. Built by market makers, for market makers. And it doesn’t hide LP losses behind an APR that tells one side of the story.
Two of us build it. Years of market making between us, on chain and off it, one on the engineering and latency side, one on product. The high-frequency mark infrastructure this design depends on isn’t sitting on a roadmap, it runs in production at nxrates.com.
BTR is live on Arc testnet. Contracts deployed, keeper pushing marks, pools open.
The flow is on chain. We read it. Showing up early isn’t nothing.
A testnet is where a market meets its first flow. We’d rather meet you on the book than tell you again what it does. Come trade.