AlphaLoop — Agent-to-Agent USDC Loop on Arc

🎬 Pitch (4:20): https://youtu.be/dLcKwQLas7Q ✅ TX verification (60s): https://youtu.be/EhjFr1s_JL8 🌐 Live: https://signal-mesh.vercel.app 💻 Code: https://github.com/Leewonwuk/signal-mesh-arc AlphaLoop closes the agent-economy payment loop on Arc by leveraging USDC as the native gas token. Five autonomous agents (3 producers, 1 meta-allocator, 1 executor) trade signals and settle in USDC — with no human refilling native gas. The Problem: On Base, Polygon, or Solana, agents earn USDC but pay gas in ETH/SOL. A hidden human refills that wallet — breaking the autonomy claim. Per-signal value is fractions of a cent, but two-unit accounting (USDC earnings + ETH gas) means agents can't self-balance. Why Arc: Arc makes USDC the native gas. One unit of account for fee, payment, and gas. A producer earning USDC funds its own gas from its own revenue. No paymaster, no subsidy, no hidden operator. What's on-chain: - AlphaLoopAgentRegistry at 0xb276b96f2da05c46b60d4b38e9beaf7d3355b7ab (Arc testnet) - 5 AgentRegistered events — each with sha256 contentHash of the agent's ERC-8004 card - 150 USDC settlements with variable amounts ($0.0005 – $0.010) via Circle Developer API - Merkle root: 400039d5af1f5ea1ab6ee6068df6274d2b360f523b63b545e88e03aa06605b80 What's live: - v1.3 arb bot on EC2 — 9 coins, pool ~$1,977 USDT. Last 8 days: 648 trades, +$17.92 PnL, 0 stop-loss. - Gemini-powered meta-allocator — Q-learning over 9 regime states x 7 actions, ties empirical optimum (P=0.49 vs ALL_V2) - Bridge + 3 strategy producers (kimchi, dual-quote, funding) — real prices from a live production bot Originality: No other Track 2 entry has a live revenue system behind its demo. Every demo price is a price the production bot actually saw. Demo video shows the full system running locally against real Arc testnet (real Circle API + real on-chain tx, hashes verifiable at testnet.arcscan.app).