erc721-ai

Explores on-chain anchors for fine-tuned model weights: who owns what, lineage, and transfer rules—without pretending full weights live on L1.

Design question

Fine-tuned models sit at the intersection of IP, data rights, and compute. Chains are strong for ownership and history; storage belongs off-chain with commitments and metadata on-chain.

What we ship

• Solidity reference implementation (MIT)
• Patterns linking off-chain weight storage to on-chain ownership
• Hooks for royalties and licensing

Research directions

Verifiable training claims, ZK/TEE attestations, and marketplace discovery for model artefacts—tracked in repo issues and the blog.

Quick start

git clone https://github.com/kcolbchain/erc721-ai.git
cd erc721-ai
npm install
npx hardhat compile
npx hardhat test

Adjust commands if the repo uses Foundry—check the README on GitHub for the current toolchain.

MIT · github.com/kcolbchain/erc721-ai

Intro: erc721-ai: on-chain provenance for model weights