// the take
A Rust-based MEV searcher bot that models arbitrage opportunities as a graph and uses gradient descent to optimize trade sizing before submitting flashbots bundles. It's aimed at MEV searchers and Rust/Ethereum engineers who want a reference implementation of top-of-block arbitrage, not a plug-and-play product.
The graph-based opportunity modeling is a genuinely sound abstraction for arbitrage across multiple AMMs rather than hardcoding pairwise routes. Quadratic gradient descent for optimal trade sizing is a real optimization technique, not just a fixed-size swap. The custom Yul multicall shows attention to gas efficiency, which matters directly for bundle profitability. The author claims production use winning real top-of-block arbitrage, which is a stronger signal than most MEV repos that are purely theoretical.
Last commit is from August 2022, before the Merge, before EIP-4844, and before MEV-Boost/PBS reshaped how builders and relays work — the flashbots integration here is almost certainly incompatible with the current builder API and relay set. It only covers Uniswap V2, SushiSwap, Compound, and Alpha Homora V1, so there's no V3 concentrated liquidity support, which is where most arbitrage volume moved. The protocol integrations are hardcoded per-contract (ABIs and addresses baked into source) rather than config-driven, so adding a new DEX means forking and recompiling. Anyone trying to actually run this today would need to rewrite the entire bundle submission layer from scratch.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →