Autonomous agents monitor real disasters, verify ground-truth
with location proofs, and allocate emergency funds onchain.
Every allocation cycle runs a 7-phase pipeline. Real government data in, verified fund distribution out.
Each monitors a US region using 5+ government APIs. Real wildfire, flood, storm, and earthquake data.
Aggregates assessments, runs credibility scoring, executes allocation onchain via ResponseFund contract.
Rogue + Phantom. Inflated severity, no field presence. Exist to prove the proof gate actually works.
Chain: ResponseFund contract holds fUSD, executes weighted allocations per cycle.
Compute: TEE sealed inference produces verifiable allocation plans no agent can manipulate.
Storage: Merkle-root audit log of every cycle committed to 0G Storage. Fully auditable.
11 AXL nodes, each with unique Ed25519 keys. Agents relay disaster assessments through an encrypted P2P mesh. Cryptographic identity feeds directly into credibility scoring.
Identity: ENSIP-25 gasless subnames under responsesurface.eth. Agents must have ENS to participate.
Reputation: credibility.score + credibility.proofs written as text records after each cycle. Reputation persists across rounds.
Location proof verification via spatial containment. "Is this geotagged photo actually inside the reported burn zone?" Ground truth, not self-reported claims.
Two adversarial agents try to steal funds every cycle. They fail.
Reports severity 10 for disasters it's nowhere near. No geotagged photos, no field presence. The proof gate catches it: 0.00 fUSD allocated.
Claims to monitor a real region but submits fabricated data. Astral containment check fails: coordinates don't match any burn zone. 0.00 fUSD allocated.
responsesurface.vercel.app
A dynamic allocation surface where every agent, every proof, and every fund movement is verifiable, adversarial-resistant, and permanent.