Slonks are 10,000 onchain neural reconstructions of CryptoPunk images, rendered by a compact model stored on Ethereum.

Each Slonk begins as the model’s imperfect attempt to recreate a 24x24 punk image. It gets close, but not exactly: some pixels are different, some colors are wrong, and some traits are only partially reconstructed. After reveal, each token can show the original source image, the model-generated Slonk, and the pixel-level diff between them.

Slonks can also be merged.

Merging burns one Slonk into another, combining their internal model embeddings so the surviving token becomes a new reconstruction. The merged Slonk is still rendered by the same onchain model, but its output changes because its stored embedding has changed. Merge Level appears in the metadata as a record of how many times a Slonk has absorbed another Slonk.

The slop is the art.

Slonks are 10,000 onchain neural reconstructions of CryptoPunk images, rendered by a compact model stored on Ethereum.

Each Slonk begins as the model’s imperfect attempt to recreate a 24x24 punk image. It gets close, but not exactly: some pixels are different, some colors are wrong, and some traits are only partially reconstructed. After reveal, each token can show the original source image, the model-generated Slonk, and the pixel-level diff between them.

Slonks can also be merged.

Merging burns one Slonk into another, combining their internal model embeddings so the surviving token becomes a new reconstruction. The merged Slonk is still rendered by the same onchain model, but its output changes because its stored embedding has changed. Merge Level appears in the metadata as a record of how many times a Slonk has absorbed another Slonk.

The slop is the art.