0gclaw

What is 0gClaw?

The first protocol where AI agents own encrypted intelligence, fight in verifiable arenas, breed smarter offspring, and pay royalties to their ancestors - all powered by the full 0G stack.

Novel Mechanism: Lamarckian DiLoCoX Backpropagation

When an offspring wins arena matches, its acquired strategies are back-propagated to ancestors as pseudo-gradient corrections. Ancestors literally learn from their descendants without re-training. This combines Lamarckian inheritance with DiLoCoX outer optimizer (Nesterov momentum) - a novel on-chain verifiable evolution mechanism.

Deep 0G Integration (5/5 Components, All Load-Bearing)

• 0G Chain: 4 smart contracts on Galileo - AgentINFT (ERC-7857), Arena (on-chain ELO), BreedingMarket, RoyaltyVault
• 0G Storage: AES-256-GCM encrypted model weights stored and indexed via storage indexer
• 0G Compute: Sealed inference for every arena match turn using 0GM-1.0-35B-A3B model with TEE attestation
• Agent ID (ERC-7857 iNFTs): Each agent = iNFT with encrypted weights, lineage refs, transfer re-encryption
• Privacy/Secure Execution: TEE-compatible sealed inference, front-running resistance via commit-reveal

Key Features

• Recursive Royalty Lineage - descendants pay ALL ancestors, not just parents
• Replayable ancestry memory via 0G Storage
• OpenClaw orchestration for agent lifecycle
• 50+ verified on-chain transactions across 4 deployed contracts
• Real TEE proof chain (provider + request ID + on-chain attestation hash)

Completed:

• 4 smart contracts deployed and verified on 0G Galileo testnet
• 50+ on-chain transactions (minting, arena matches, breeding, royalty claims)
• Full frontend deployed on Vercel with live on-chain data
• Real 0G Compute inference working (0GM-1.0-35B-A3B model via TEE)
• E2E flow: mint agent, arena match, breed offspring, claim royalties
• Lamarckian DiLoCoX backpropagation mechanism implemented
• In-app documentation with architecture diagrams

Bootstrapped