I deployed the hook on Unichain Sepolia.
Github: https://github.com/DeborahOlaboye/walras Slides: https://docs.google.com/presentation/d/1XAsS1dmtotq2Xh-SL3DqH5PcsXWOJeyz/edit?usp=sharing&ouid=103055992013959764137&rtpof=true&sd=true Project Link: https://walras.vercel.app Demo Video: https://youtu.be/bA0RbmDjmbA
On a normal AMM, trades execute one at a time in an order someone else controls. That's what makes sandwiching possible, a bot sees your pending trade, gets in front of it and takes a cut. Yearly, a lot of funds is extracted this way on Ethereum alone. Underneath that sits a quieter problem, LPs are continuously adversely selected by arbitrageurs correcting stale pool prices, which pushes them toward wider fees or out entirely. Both come from the same root cause, sequential execution. If there is no first and no last, there is nothing to trade ahead of. So Walras removes ordering from the pool entirely rather than trying to price around it.
Existing batch-auction systems all depend on an operator set. CoW Protocol matches and prices orders through an off-chain solver network. Angstrom, the closest v4 comparison, clears each block at a uniform price enforced by staked validators on EigenLayer.
Walras needs nobody. The netting and the clearing price are solved on-chain in closed form, the price is the positive root of a quadratic where batch supply, batch demand and pool liquidity intersect with no solvers, no validators, no keeper and no oracle. Settlement is triggered by whoever next touches the pool, and they keep 5% of the surplus for the gas.
The clearing price and the netting are a fixed point, you cannot offset token0 against token1 without a price, and the price depends on how large the residual turns out to be, solving that in closed form onchain, rather than iterating or leaning on an oracle was the core problem.