0G Labs Global Hackathon

Verifiable yield optimization for the AI-native Web3.

Mainnet-first optimizer, 0G proof trail, frictionless judge review, and strategy-agent NFT minting.

0G Mainnet default 0G Storage + ProofRegistry 0G Compute path YieldStrategyINFT

Live app: yieldboost-ai.vercel.app/judge

The DeFi AI trust gap

AI yield advice is easy to generate. It is hard to verify.

Opaque AI output Most dashboards show a recommendation, but not a durable proof trail for what was decided.

Review friction Hackathon judges should not need a wallet, faucet, or manual rerun just to audit the result.

Strategy ownership A completed strategy should become a portable artifact, not a temporary UI state.

YieldBoost AI turns an optimization into a visible chain of custody: compute, storage, registry anchor, and optional Agent NFT.

Product loop

One click produces a strategy, a proof, and a review surface.

Connect wallet User enters the normal dashboard flow with an active wallet context.

Optimize route Portfolio snapshot is compressed and routed through cache, embeddings, and compute.

Anchor proof Result is stored through 0G Storage and anchored through ProofRegistry.

Review or mint Judge Mode audits the proof. User can mint a Strategy Agent NFT.

Projected APY 23.84% latest proof snapshot

Confidence 96% route scoring

Proof coverage 100% stored run trail

Judge wallet 0x8a3c... read-only demo wallet

0G-native architecture

The system is built around a verifiable data path.

User input Wallet + portfolio snapshot

Prompt compression Compact intent and constraints

Semantic cache Reuse exact wallet/network state

Embedding reuse Alibaba text-embedding-v4

0G Compute Inference path + verified runtime metadata

0G Storage Persist proof payload

ProofRegistry On-chain anchor for review

The key design choice: optimization is not just displayed. It becomes a durable, wallet-scoped proof artifact that the judge can inspect without touching user funds.

Judge Mode

Audit-first UX: no wallet, no faucet, no rerun.

Hard-locked demo wallet 0x8a3c7524Aaed081825aC88eC7f4cCECFc583ee7D

Mainnet-first, testnet switchable Reviewers can compare networks without leaving the audit page.

Direct explorer path Latest storage tx, ProofRegistry anchor, contract addresses, and proof history are surfaced in one place.

/judge

Mainnet review

Mainnet review starts here.

Latest proof-backed snapshot, APY lift, and verification payload.

APY lift 12.38 -> 23.84%

Proof history 14 runs

Mainnet evidence

Live contracts and latest proof trail are already on 0G mainnet.

ProofRegistry 0x8e63e117E71A80Cfc10fDF375F079e2e29cd7D7D Anchors proof metadata and emits the reviewable proof event.

YieldStrategyINFT 0xb264D861264B0e4f8fb98A61B7694BA8a3B6BBe3 Mainnet Agent NFT contract for strategy ownership.

Attestation Oracle 0x216E7880D64D94335B583c539802d3e61958d4A2 Allows verified attestation hashes to be registered on-chain.

Latest anchor tx 0xa76f59de764dfb5dcd2fae3e8dff53cb0e213bab89162e7b4de16962309caa9b Mainnet ProofRegistry transaction linked from the product.

Runtime efficiency

Inference is optimized before the system spends compute.

Semantic Cache Scoped by wallet, network, normalized prompt, and portfolio digest before new inference runs.

Embedding Reuse Alibaba text-embedding-v4 can reuse similar optimization requests for the same wallet and asset signature.

Prompt Compression The optimizer rewrites noisy input into compact intent, constraints, and portfolio facts.

Raw intentUser request + portfolio

Compacted promptLess noise, same decision context

Reuse checkExact cache or vector similarity

Compute only if neededPreserve 0G proof flow

Strategy Agent NFT

Optimization results become ownable agent artifacts.

After a proof-backed optimization, the user can mint the strategy through YieldStrategyINFT. The NFT is not the proof itself; it is the portable strategy artifact linked to the verified runtime output.

Minted to the connected wallet Agent ownership follows the user wallet, not the server signer.

Encrypted metadata Strategy metadata is protected with AES-256-GCM at the app runtime layer.

/agents

Agent NFT

SaucerSwap LP Strategy Agent

Confidence 96% | APY 23.84% | Proof-linked mint

Roadmap

From verifiable optimizer to Proof-of-Optimization economy.

Phase 1: Mainnet proof product Live proof trail, Judge Mode, Agent NFT minting, and dual-network review.

Phase 2: $YA0G reward logic Reward high-quality optimization events backed by ProofRegistry and measurable outcome metadata.

Phase 3: Mining by optimization quality Proof-of-Optimization weights verified runs by confidence, impact, and durable proof coverage.

RunUser optimizes portfolio

Verify0G Storage + ProofRegistry

OwnMint strategy agent

RewardFuture $YA0G mining

YieldBoost AI is strongest as a mainnet-live verifiable finance product: AI recommendations with a proof trail judges can inspect in seconds.