For the past years we’ve celebrated smart contracts as the final word in trust-minimised finance, but they only verify state. The next leap is code that can also think and act on that state. We call these Agentic Protocols (APs). Most “crypto AI” experiments still finish with a chatbot pushing Swap while the user shoulders every risk. APs invert that flow: you state a clear intent: grow this balance, keep draw-down under two per cent—and an onchain agent decides how another one executes every block. Outcomes are transparent, non-custodial, and never need a sleeping bag.
This is finally practical because the substrate is ready. Solana now finalises blocks in under a second for fractions of a cent, so strategies that once bled fees can rebalance hundreds of times a day. The plumbing is equally deep: money markets like Kamino and marginfi expose program-callable APRs, perp DEXs such as Hyperliquid stream funding rates in real time, and Pyth oracles publish millisecond prices. Model Context Protocol (MCP) stitches it all together, letting an agent run its entire decision loop inside every transaction, so the whole brain lives onchain where anyone can audit it.
The early proof-points
The first wave of DeFAI projects shows that fully autonomous agents already attract serious capital, and that the field is still wide-open.
Giza’s ARMA stable-yield vault, live on Base since January 2025, has executed 100 000+ onchain moves, routed $58M in volume and now manages about $2.4M while users simply set risk prefs and let the bot rebalance.
BasisOS pushes beyond lending: its cross-chain basis-trading agent (spot long on L2s, perp short on Hyperliquid) launched in April 2025 and has drawn $4.7M AUM while staying funding-fee positive.
Glider, backed by a16z, Variant and Bankless Ventures, lets users drag-and-drop rules like “50/50 ETH-BTC + yield-farm the rest”; the beta is live and a $4 M seed round just closed.
The lesson is clear: performance-driven automation is no longer theory, it’s already managing millions and scaling fast.
The market gap
Yet billions of USDC still sit idle because most holders lack either the time or the confidence to loop, hedge, and babysit a dozen protocols. Quants, on the other hand, possess high-alpha strategies but only so much personal capital. What we lack is a permissionless venue where any strategist can launch an agent and any wallet can fund it with a single click—without ever handing over custody. That missing link is what we’re building at yld.fun.
The (3,3) flywheel
Our economic engine is a simple (3, 3) flywheel. Believers stake stablecoins and harvest optimised yield. Quants publish their automated strategies and collect performance fees. The protocol retains a tiny slice, uses it to buy and burn $YLD, and rewards holders. Higher net yield and a shrinking supply attract more capital and more strategists, which tightens the loop again. Everyone’s incentives compound.
YLD Liquidity: programmable capital
We released the first proof of concept last week. YLD Liquidity capped at 50,000 USDC and routed funds through a handful of Solana protocols. Even as chain-wide stable yields fell roughly 30%, the agent still produced a 6.53% annualised return, about 60% above the Solana stable-yield average reported by DeFiLlama. Every 1 of 50+ reallocations is visible in wallet:
5HnitPoEeKrxNk5Gyu7U9uAxnPi57DKB2uUkqdhgEE9
Where do we go next?
Phase 0 · Validation (now)
Launch small, capped vaults to A/B test autonomous vs. manual farming.
Publish transparent performance data to build trust.
Stress test risk guards, fee logic, and monitoring dashboards.
Phase 1 · Strategy Studio
Ship a drag-and-drop, no-code console.
Allow users to chain building blocks—lending loops, LP, restaking.
One-click backtests and onchain deployment.
Phase 2 · Permissionless Marketplace
Any strategist can deploy a vault; any wallet can fund it in one click.
Deposits remain non-custodial. Funds stay in user-controlled smart wallets.
Reputation, performance, and fee curves surface the best agents automatically.
Phase 3 · Cross-Intent, Cross-chain
Combine multiple goals (earn, hedge, restake) in a single autonomous loop.
Users set risk bands once; the agent juggles allocations to stay inside them.
Expand to other chains (EVM L2, Hyperliquid, SUI)
Smart-contract AMMs rewrote market-making in 2020; Agentic Protocols will rewrite money management in 2025. Swaps are already trustless, your yield strategy should be as well. If you’re a developer or quant, keep an eye on us for the early access to the Strategy Studio and stress-test the agents. If you’re a yield seeker, join the next capped vault or watch the dashboard at yld.fun update in real time. And if you’re simply curious, drop into our Telegram and vibe with the community.
Your capital already has an intention; it’s time to give it a real engine