Roadmap

Now — Q3 2025: Testnet foundation

Everything listed below is live today on Sui testnet.

• AI-powered yield vault with Gemini as the decision engine
• 10 mock DeFi protocol adapters with randomized APY walks
• Automatic rebalancing via on-chain PTBs
• Permanent Walrus audit trail for every decision
• Real-time AI Log with full reasoning, input rates, and allocation breakdown
• Yield history chart and personal returns tracking
• Atomic withdrawal with auto-pull from adapters
• Open source Move contracts and agent

Phase 1 — Q4 2025: Mainnet with real protocols

The first mainnet release replaces mock adapters with live Sui lending protocol integrations.

• Suilend adapter — deposit into Suilend USDC lending pools, read live supply APYs
• NAVI adapter — integrate NAVI Protocol USDC market, including boosted reward APYs
• Scallop adapter — Scallop sUSDC market with isolated pool support
• Real USDC (Wormhole bridged and native Circle USDC on Sui)
• Smart gas management so users never need to top up SUI manually
• Full security audit of all Move contracts before launch
• Rate limiting and circuit breakers to protect against oracle manipulation

This is the phase that makes Aquifer real. The AI agent will see live on-chain rates from protocols handling hundreds of millions in TVL and move actual capital between them every few minutes.

Phase 2 — Q1 2026: Multi-asset vaults

Expand beyond a single USDC vault.

• USDT vault
• SUI liquid staking vault — aggregates stSUI, vSUI, haSUI across Aftermath, Haedal, and Volo
• Wrapped ETH vault (wETH across Sui protocols)
• Each asset runs its own AI agent tuned to that market's dynamics

Phase 3 — Q2 2026: Strategy modes

Not every depositor has the same risk preference. Aquifer should offer choices.

• Conservative: spread capital across at least five protocols, never more than 25% in any single one, prioritise stability over maximum yield
• Balanced: the current default, AI decides freely within protocol minimums
• Aggressive: maximise APY, concentrate into the top two protocols, accept higher concentration risk
• Depositors choose their mode at deposit time. Separate vaults per strategy.

Phase 4 — Q3 2026: Yield alerts and notifications

Keep depositors informed without them having to check the app.

• Email alerts when APY drops below a threshold the user sets
• Push notifications for large rebalances
• Weekly yield summary sent to connected email
• On-chain subscription mechanism so notifications work even if users change devices

Phase 5 — Q4 2026: Governance

Move control over the protocol to the community.

• Governance token distributed to early depositors based on vault usage
• Token holders vote on which protocols get added as adapters
• Votes on fee parameters (protocol fee on yield, if any)
• Emergency pause votes if a protocol is suspected compromised
• All votes executed on-chain via Move governance contracts

Phase 6 — 2027: Developer ecosystem

Let other builders use Aquifer as infrastructure.

• SDK: a TypeScript package to deposit to, withdraw from, and read Aquifer vaults in any dApp
• Adapter standard: a published Move interface so any Sui protocol can write a compatible adapter and apply for inclusion
• Vault factory: deploy custom vaults with different assets, agent addresses, and allocation constraints, no code required
• Webhooks: subscribe to rebalance events and receive them as HTTP POST requests to any endpoint

Phase 7 — 2027: Multi-chain expansion

Take the Aquifer model beyond Sui.

• Ethereum mainnet vault using ERC-4626 with a Gemini agent watching Aave, Compound, and Morpho
• Solana vault across Kamino, Drift, and Marginfi
• Cross-chain yield comparison: the AI can see rates across all chains and signal where capital should move if bridging costs justify it
• Unified dashboard showing positions across all chains in one view

Long-term: On-chain AI verification

The most ambitious item on the roadmap: proving cryptographically that an AI model made the decision, not a human.

• Use a trusted execution environment (TEE) to run the Gemini inference
• Generate an attestation proof that a specific model with a specific prompt produced a specific output
• Store the proof on Walrus alongside the existing blob
• Anyone can verify not just what the AI decided but that the AI actually decided it

This would make Aquifer the first verifiably AI-governed DeFi protocol, with no human in the loop at any step.