// the take
A reference implementation of a Uniswap V2/V3 sandwich bot: a Rust strategy engine (built on the Artemis framework) for finding and simulating sandwich opportunities, paired with a gas-optimized Huff contract for execution. It's aimed at people studying or building MEV searchers, not at someone who wants to point it at mainnet and extract profit unmodified.
Clean separation of concerns — Huff contract handles the gas-critical execution path while Rust with revm does fast local simulation of candidate bundles before anything touches the chain. It covers both V2 and V3 pools and multi-meat (bundling several victim transactions into one sandwich), which most public examples only do for one pool type. The salmonella inspector is a genuinely useful, specific check: it flags ERC20 tokens whose transfer function uses opcodes that behave differently on mainnet than in simulation, a real failure mode for this kind of bot. The token-dust retention trick for cheaper gas on repeat trades shows this came from someone who actually ran one of these, not just read about them.
Dead: last commit is August 2023, and the README itself tells you to go use the author's rewrite on Artemis instead — this version is a frozen snapshot, not something getting fixes. It says nothing about the two things that actually determine profitability: mempool/order-flow access and bundle inclusion through a builder relationship; without those the bot has nothing to sandwich. There's no handling discussed for reorgs, failed-bundle costs, or gas-auction losses, which dominate real P&L for this strategy and aren't optional edge cases. It's explicitly flagged by the author as "highly experimental" and MIT "as-is," with no security audit of the Huff contract mentioned anywhere.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →