Inspiration Graph

2. Inspiration Graph: Technical Data Model

2.1 Graph Representation

The Inspiration Graph is formalized as a Directed Acyclic Graph (DAG):

$$\mathcal{G} = (\mathcal{N}, \mathcal{E})$$

• 𝓝 (Nodes) → Canonical, unique representations of minted creative works.

  • Canonicalization → Performed via Keccak-256 hashing of normalized metadata.

  • Identity = id = Keccak256(IPFS_CID || creatorAddress || timestamp)

• 𝓔 (Edges) → Directed inspiration links:

$$e_{j \to i} \in \mathcal{E} : I_j \to I_i$$

• Each edge must be cryptographically provable via commitment to an onchain zk-SNARK proof.

• Acyclic Constraint → Cycles invalid → enforced at verification level to prevent infinite recursive payouts.

---

2.2 Node Metadata Schema (EVM-Indexed)

Solidity-Friendly Metadata Schema (Struct):

struct InspirationNode {
    bytes32 id;                // keccak256(IPFS || creator || timestamp)
    address creator;           // EOA or contract
    bytes32 parent;            // Parent node hash (0x0 for originals)
    uint8 depth;               // Depth in DAG
    bytes32 proofHash;         // Commitment to zkProof onchain/offchain
    bool declaredInspiration;  // Explicit declaration
    uint16 royaltyShareBP;     // Basis points (e.g., 500 = 5%)
    uint256 timestamp;         // Unix epoch
}

• Gas Optimizations: uint8, uint16, and packing for efficient storage.

• Canonicalization Contract → Optional pre-deployment content integrity check.

---

2.3 Proof Integrity and Verification System

a) zkProof System Design

• proofHash → SHA256 commitment of zkSNARK output.

• zk Circuit → Takes as inputs:

→ Computes a proof of derivation where:

1. parent_hash → Keccak256 of parent content

2. child_CID → IPFS CID (content address)

3. salt → Random nonce for uniqueness in commitments

Zero-Knowledge Privacy → The content of derivation is hidden; only the proof verifies cryptographically that derivation took place from parent to child.

---

b) Advanced zkML Proofs (Future Planned)

• zkML Integration (WIP) → Use zkSNARKs to encode style similarity or content feature extraction.

  • Example circuits → ResNet embeddings, audio fingerprints, or text semantic embeddings.

  • Prove “inspiration proximity” in latent feature spaces without revealing actual embeddings.

Input → Embedding(child) - Embedding(parent) → zk circuit → ≤ threshold → Proof

---

c) Onchain Verification Pipeline:

1. submitProof(bytes calldata proof, bytes32 parent, bytes32 child)
2. zkVerifier.verify(proof, parent, child) → returns (bool)
3. emit InspirationProven(child, parent, msg.sender)

• Verifier Contract → Modular, supports circuit upgrades.

• Proof Hash Storage → Immutable commitment → protects against forgery or post-hoc manipulation.