Nevermind

DeFi yields are scattered across hundreds of protocols and chains. Finding the best rates requires constant monitoring, manual rebalancing, gas costs, and real risk assessment. Most users end up choosing one protocol and hoping for the best.

Nevermind changes that. You deposit USDC, choose your risk tolerance (Conservative, Balanced, or Aggressive), and an AI agent handles the rest — fetching live yield data from DeFi Llama, analyzing rates across Aave V3 and Compound V3 on Ethereum, Arbitrum, Base, and Optimism, then recommending and executing optimal allocations on-chain.

The key difference: the AI does not run on our server. It runs inside a Chainlink CRE workflow, executed by a Decentralized Oracle Network. Every step — data fetching, on-chain reads, LLM analysis, and rebalancing — is verified through DON consensus. No single party controls the AI decisions or fund movements.

Users connect via thirdweb (email or social login, no MetaMask required), deposit into the NevermindVault smart contract on Tenderly Virtual TestNet, and monitor everything in real time through the agent activity log.

Nevermind consists of three repos working together:
Smart Contract — NevermindVault.sol (Solidity 0.8.20, Foundry). A custom USDC vault with per-user balance tracking, risk profiles, and a CRE-authorized executeCRERebalance() function. Deployed on Tenderly Virtual TestNet (Ethereum mainnet fork, chain 73571) using real USDC at its mainnet address. Includes 18 Foundry tests.

CRE Workflow — TypeScript using @chainlink/cre-sdk and viem. A 4-step cron pipeline:

1. HTTPClient fetches yield data from DeFi Llama with consensusIdenticalAggregation.
2. EVMClient reads vault totalDeposits at LAST_FINALIZED_BLOCK_NUMBER.
3. HTTPClient sends yield data and risk profile to an LLM via OpenRouter for allocation analysis.
4. EVMClient submits a DON-signed report to executeCRERebalance().
5. Configuration is validated with Zod.

Frontend — Next.js 16, React 19, Tailwind CSS v4, thirdweb v5. The dashboard displays live yield data, AI recommendations with per-protocol reasoning and risk scores, portfolio positions, and an agent activity log. API routes (/api/analyze, /api/rebalance) connect the frontend to the CRE workflow and Tenderly VNet for on-chain execution via impersonated transactions.