← the studio
Opportunity [Agency AI]

An agent organization
of agent organizations.

Where a single Hunter AI — like Crypto Hunter AI — is a multi-agent organization that hunts opportunity in one domain, Opportunity is the layer above them: it reads across every independent Hunter engine and decides how your finite time and money should actually be spent today, across every domain at once. Built on Veritas — the deterministic-scaffold framework every engine, and Opportunity itself, are made of.

The stack

Three layers. Veritas is infrastructure, not a product.

Veritas is the tool stack — the org registry, the hub, the propose → gate → debate mechanics. Hunter engines are independent org instances built on it, one per domain: each owns its own scouts, its own deterministic gate, its own debate, its own memory. Opportunity reads every engine's verified output through read-only bridges and runs its own gather → weigh → learn agents on top — not deciding what's true (each engine's gate already settled that), but what's worth this user's next dollar and next hour.

Crypto Hunter verified queue Collectible Hunter verified queue Free Money Hunter verified queue OPPORTUNITY gather · weigh · learn arbitrates priority, not truth today's allocation
Each Hunter engine's gate already settled truth. Opportunity settles priority — one shared budget, many legitimate claims on it.

Two principles, every layer

The arc is the validation

Scouts and debaters run unconstrained; the gate fires at discrete points only — once on filing, again only on new evidence. Never continuously. True inside an engine and at the Opportunity level.

Gate-hard vs. debate-argued differs per domain

Decided by how deterministically checkable a risk is. Crypto's scam-safety is gate-hard. Collectibles' arbitrage math is gate-hard; authenticity is debate-argued instead. No universal split.

Two kinds of debate. Within an engine, Advocate/Skeptic/Strategist contest one opportunity's own merit — never changing the verdict. Across engines, at the Opportunity level, the debate isn't about truth at all (each engine's gate already settled that) — it's about priority: independently-legitimate opportunities from different domains competing for the same finite time and money.

Constitution

Hard invariants, every layer.

Never holds keys

Never signs, executes, or submits anything on the user's behalf — at any layer.

Deterministic trust, at every level

A gate decides what's verified inside an engine; one deterministic allocation decides what reaches the user at the Opportunity level. Arbitration, not another opinion.

Provenance, always

Every trusted claim traces all the way down to the engine and gate check that produced it.

Intelligence, not advice

Opportunity ranks and arbitrates what's real; the user decides what to act on — across financial, legal, and personal-risk domains at once.

Hunter engines

Named, ranked by how cleanly each domain gates.
EngineStatus
Crypto Hunter AIshipped — the reference implementation
Collectible Hunter AIshipped — comp-arbitrage gate, live-validated
Free Money / Unclaimed-Benefits Hunter AIshipped — registry gate, live-validated end to end
Grant Hunter AI, Scholarship Hunter AIplanned — cheap to add, no urgency
Vehicle, Developer/Bounty, Government Opportunity Hunter AIplanned
Stock Hunter AIplanned — advice-liability posture needs its own care
Property Hunter AIflagship vision, deliberately deferred — MLS licensing, county-level data fragmentation
Startup/Equity Hunter AIunder consideration — too high-variance to gate meaningfully today

Where it stands

✓ proven  the thesis, end-to-end, in Crypto Hunter AI — 47 tests, daily autonomous run
✓ extracted  the shared ~80% became hunter-engine; two independent consumers (Collectible + Free Money Hunter AI) proved every extension point live
✓ shipped  Collectible Hunter AI + Free Money Hunter AI — both live-validated against real APIs
✓ shipped (v1)  Opportunity's own meta-layer — read-only bridges into all three engines, a deterministic shared-budget cross-engine allocation, its own keystone record and store. First real run merged 18 real Crypto Hunter AI opportunities with 5 real Free Money Hunter AI ones into one genuine mixed allocation.
→ next  the LLM-driven cross-org debate — arguing priority across engines, not truth — the genuinely novel piece this vision was built around; v1's greedy allocation proves the mechanism, the debate layer goes on top of it
Remaining Tier-1 engines (Grant, Scholarship) are cheap to add whenever wanted — no urgency, the meta-layer proof was the priority.