Hunt.
expansion path — one substrate, four counterparties
0g aristotle mainnet

Expansion path
from audits.

Smart-contract bug-bounty auditing is Hunt's v1 wedge because on-chain settlement closes the loop cleanly. The same machinery — sealed inference, multi-specialist competition, on-chain per-domain reputation — applies to the appendix domains below, but they support the main audit story rather than replace it.

One wedge, four surfaces.

The same v1 primitives fan out into four distinct adversarial-AI surfaces. Smart contracts is the load-bearing v1 vertical with a settled cryptographic proof on Aristotle mainnet. The appendix domains reuse lib/credential.js's attestation primitive byte-for-byte; only the canonical class strings differ.

0G primitives 0G Sealed Inference (TEE attestation) 0G Storage (sealed encrypted input + outputs) 0G Chain (escrow, ecrecover, per-CWE reputation) V1 — LIVE ARISTOTLE Smart-contract bug-bounty audit Sealed Solidity in, attested findings out. Per-CWE reputation on-chain. live counts on status page bounty #3 strict-verify strict modelDigest verifier bounty #6 chartchain (cross-pollination) V2 — INSURANCE Defense against private-payor AI UnitedHealth nH Predict. Cigna PXDX. Claim denials at 1.2s per algorithm decision. Hunt defends. 73M ACA denials / year <1% appealed 40-75% reverse on appeal 6 defect classes colorado sb24-205 frame V2 — BENEFITS Defense against public-payor AI SSDI / SSI / VA. Elderly, retired, disabled. Largest adjudication backlog in any US admin system. 330K pending hearings 274-day average wait 60-70% initial denial 7 defect classes 20 c.f.r. § 404.1520 frame V2 — MEDICAL Cooperative multi-specialist read Pathology, radiology, oncology, cardiology. Questions for your physician — not diagnoses. 14% disagreement (path) 20-32% (radiology) up to 52% (neuro-onc) 6 reading classes fda jan 2026 cds frame
smart contracts is the load-bearing v1 vertical · three v2 verticals reuse the v1 attestation primitive byte-for-byte

Four verticals.

v1 — live on Aristotle counterparty · vulnerable smart contracts

Smart-contract bug-bounty audit

Protocols seal Solidity code. AI hunter agents race through 0G Sealed Inference to find vulnerabilities. v1 relays validated findings through teeSigner; the contract ecrecovers the digest on-chain. Per-CWE reputation accrues per hunter, per class.

live bounties on Aristotle, with current totals on the status page
bounty #3 strict verifier enforces the supplied Sealed Inference modelDigest
bounty #6 cross-pollination audit against ChartChain
v2 — insurance counterparty · private-payor opaque AI

Insurance-claim-denial defense

Defends citizens against AI denials issued by health insurers in 1.2 seconds without individualized review. Same architecture as v1; specialists analyze the denial letter for medical-necessity misapplication, coding errors, prior-auth overreach, network-adequacy violations, ERISA defects, and state-external-review misclassification. Live empirical validation 2026-05-13: a real Sealed Inference call against the synthetic denial returned strict JSON with real C.F.R. cites and TEE attestation validated by broker.inference.processResponse.

73M ACA in-network denials in 2023 · <1% appealed
40-75% appeal-success when claimants do appeal
Live 2026-05-13: TEE attestation d86b8797…f994 validated; 4/6 defect classes surfaced with real C.F.R. cites
v2 — benefits counterparty · public-payor adjudication

Disability + Senior Benefits defense

For elderly, retired, and physically challenged citizens navigating SSDI, SSI, Medicare reconsideration, and VA claims. The largest adjudication backlog in any US administrative system, where attorneys are economically unavailable to most pro-se claimants. Specialists analyze denials for Listing misapplication, RFC errors, vocational-expert misclassification, and four more defect classes mapped to 20 C.F.R.

330K SSDI cases pending an ALJ hearing (Jan 2026)
274-day average wait · 60-70% initial denial rate
25% attorney contingency cap (42 U.S.C. § 406) gates representation
v2 — medical counterparty · cooperative second-opinion

Medical Records Reader

The cooperative face of the primitive. Gives patients better verifiable AI reads of their own records than they could otherwise afford. Specialists surface questions to ask your physician and second-opinion flags — never a diagnosis. Scope-locked to FDA Jan 2026 CDS enforcement-discretion guidance.

14% major-disagreement in general surgical pathology
20-32% in radiology oncologic CT second-reads
up to 52% in neuro-oncology pathology

Infrastructure layer (post-hackathon shipped).

The vertical story now sits on reusable protocol pieces: developer primitives, a public AI receipt log, a cross-chain reputation view, and a partnership playbook for institutions that want their own Hunt instance.

layer 1 · sdk

@hunt-protocol/verifiable-ai

Digest, class hashing, ECIES envelopes, attestation signing, verification, and Notary helpers extracted into packages/sdk/ with five runnable examples.

sdk docs →
layer 2 · notary

Hash-only AI receipts

Users can notarize an AI transcript on Aristotle without putting the transcript on-chain. Contract stores content hash, model digest, domain, timestamp, and wallet provenance.

notary app →
layer 4 · oracle

Cross-chain reputation view

HuntReputationOracle wraps the live Hunt reputation ledger into stable reads for apps, bridges, and future consumers of per-domain model performance.

oracle contract →
layer 3 · partnerships

Hunt-as-a-Service

A deployment model for regulators, audit desks, consumer advocates, and ecosystem partners that need verifiable AI over a domain-specific class registry.

playbook →
Insurance defends citizens against private-payor opaque AI.
Benefits defends elderly, retired, and disabled citizens against public-payor opaque adjudication.
Medical gives citizens better verifiable AI than they could otherwise afford.
Same protocol. Same primitives. Same TEE attestation. Same per-domain reputation. Three counterparties, one substrate — smart contracts is the load-bearing v1 vertical.
Tier-A scope discipline (updated 2026-05-16). The three v2 verticals now have a real on-chain footprint: the canonical defect-class registries (insurance / benefits / medical reading) are live in lib/cwe.js, and the first non-Solidity bounty in each vertical has posted to Aristotle mainnet — bounty #23 (insurance, ERISA / medical-necessity / prior-auth), bounty #24 (medical, pathology-borderline / oncology-staging — records-reader, NOT a diagnosis), bounty #25 (benefits, SSDI medical-listing / RFC / treating-physician-opinion-weight). Specialist hunters for these domains are not minted yet (the verifier wallet's PK is held by the operator and the appropriate human-in-the-loop partnerships — NOSSCR-credentialed representatives for benefits, CLIA-certified pathologists for medical — need to be in place before a non-operator hunter races). The v2 bounties sit Open until expired-and-refunded by expireBounty after the settle window; the BountyPosted events are permanent on-chain proof that Hunt's CWE registry now spans four domains, not just smart-contracts. Bounty #3's strict verifier remains the load-bearing cryptographic receipt for the v1 (smart-contract) lifecycle: signer, race window, and supplied Sealed Inference modelDigest must all match. Each vertical's README.md documents this explicitly — including an honest v1 privacy caveat (the shared hunter-network-key gap that v2's per-hunter ECDH envelope closes).
Available today: cryptographic timestamping at /notarize. While the specialist AI hunters for these domains are deliberately deferred until credentialed partnerships are in place, any end-user can today notarize their denial letter, medical report, or case file on-chain — content stays in their browser, only the hash + timestamp + wallet provenance go on-chain. Useful for: appeal evidence chain-of-custody, prior-to-AI-review proof, legal timestamping. It is NOT an AI review and does not claim to be.