Whitepaper TLDR: Slonks is essentially an on-chain AI/neural network experiment that reconstructs CryptoPunks as NFTs: it places a small image model on the Ethereum blockchain, with each Slonk corresponding to an original CryptoPunk. The model generates an 'imperfect replica,' and the number of pixels that differ between the generated image and the original Punk is called slop. According to the official website, the model's packed weights (214KB) are stored on-chain, with 24×24 pixels using CryptoPunks' 222-color palette. The slop is calculated in real-time every time it's read, rather than being pre-stored. In the whitepaper, SLOP turns this 'slop value' into an ERC-20 token: you can send a Slonk into the void, and SLOP is minted based on the current slop value. Burning SLOP allows you to randomly resurrect a Slonk from the void and give it a new embedding.