“To remain competitive in a zero-latency market, we must deploy high-velocity autonomous agents. But the legal liability of autonomous decisions was an existential risk. We needed a platform that wasn't just a standard login tool, but a structural risk discovery primitive. By investing in the Transformation Package, we gained the agent licenses and the structural audit to map our Trust Debt. It gives us the definitive, decidable proof we need to legally justify our compliance and secure meaningful human oversight.”
— Chief Risk Officer, Fortune 100 Organization Deploying Autonomous Systems
What they call "the structural audit" above is today's Readiness Investigation, below.
This is not an IAM.
Identity & Access Management
This is the hardware-verified signal that feeds into your IAM to make high-velocity decisions legally defensible.
Enter →The drift-gate · free to run
One install turns any agent into a verifiable, bounded one. If you can use a chatbot, you can deploy it.
Advisory custody by default: the agent proposes, the gate measures, you sign — the keys never enter the model's context. It is an attestation oracle, not a custodian and not a market-maker. Running it to test — to verify competence, to enter the market — is free, and we want you to reach for it. The price never goes up — the term each license covers shrinks going forward, so the earliest licensees lock in the most coverage per dollar, permanently.
Free to test — licensed to insure. The gate's output only becomes a number a carrier will underwrite against once it's under license — and it takes two: the deployer running the agents and the carrier pricing the risk each hold one, to the same patent. Verifying competence is free; making it insurable is the license.
The one price · volume unlocks the tiers
5,000 licenses (5,000 × $20 = $100K) also unlocks the Readiness Investigation audit
See the top tier →Certs for the humans who sign come with your license tier (1 per 10). Reading is free.
Read the manual →Not the decision maker? Forward this to your CEO. They delegate, you advance.
You're Liable for AI.
IAM proves who logged in. It can't prove what your agent actually did once it was in — that's FIM.
When it breaks—and it will—they'll ask what you did to prevent it.
You knew. You had a chance. What did you do?
Anyone who fixed AI reliability fixed competence verification at silicon speed too — by Rice (1953), same problem. They didn't. We did. We patented it.
The vendor selling you a software fix for AI containment has claimed Rice does not bind them — a claim any diligence falsifies. The only proof Rice does not bind is a substrate receipt of this shape, and that same receipt clears a human into a verified role. Whoever solves your AI side at the substrate has built the labor market's instrument too, whether they meant to or not.
The wild implications: no job search (the receipt IS the match, at cache-line speed); no separate verification step (stay-in-lane attestation IS the proof); a dignity pixel per operator and the next axis to grow into. Max income becomes a navigable trajectory.
Why believe? The same XOR that prices an AI agent's liability prices a human's role-fit. The 12-cell lattice the chip checks against is the same 12 coordinates the nine ThetaCog rooms operate from — Strategy × Tactics × Operations at three altitudes (silicon · terminal · operator). One architecture, three scales.
First implementation, not metaphor: every commit in this repo writes a receipt against that lattice; the rooms aren't marketing — they're the operating system this organisation runs on. iamfim names the liability; the rooms hold the daily proof. Schema · Argument · The nine rooms · The chip side.
Not a promise — proof you can browse: every commit's live drift receipt → (here's this page's own). Green in-lane · amber bleed · red drift, recomputable to the same panel on your machine.
No model in the loop — the same input recomputes to this exact panel on your machine.
This validates competence — it doesn't replace it
If you're a lab, a research group, an incumbent with real depth — this isn't the thing that disrupts you. It's the thing that finally makes your edge legible and defensible. The undecidable-AI flood commoditizes confident-but-wrong output; a decidable receipt is what lets genuine competence prove itself at machine speed instead of drowning in it.
It complements the theory, it doesn't compete with it. The labs and the research frontier own the ideas; we own the on-chip receipt that grounds one of them. The receipt doesn't manufacture skill — it can only witness it. Your judgment, your depth, your record become provable at machine speed — the moat you already earned, now legible to anyone who has to trust it.
And it widens the door without lowering it: real skill that never had the pedigree finally carries a receipt a stranger can check, so your best people are recognized wherever you find them. The same coordinate that prices an agent's lane prices a person's — it hands the expert a sharper proof and the newcomer a decidable place to stand.
Before you build it yourself
Good instinct — you should assume you can. So here's the calibration we publish, so you don't have to take our word for it. On sealed, blind, cross-domain inputs the sensor separates in-domain from out-of-domain at 0.90 and rejects out-of-domain ten times out of ten. Against a scrambled null — same bytes, meaning destroyed — the signal stands at 4.48σ. And we publish the inch we haven't closed: paraphrase-invariance sits at 0.30 — surface rewording still moves it more than we'd like, and we say so. Those are the numbers a weekend clone has to beat, on held-out data it has never seen.
The hard part isn't the metric — it's that the metric has to answer where is what: every input has to land as a coordinate, not a scalar, so “a little different” and “out of its lane entirely” are distinguishable. Getting that coordinate space to hold under paraphrase and adversarial camo is the work, and the map that does it is the patented part (US Patent App. 19/637,714). We don't claim it makes your AI correct — Rice's theorem says no one can decide that. We claim it makes placement decidable, and we publish the σ so you can audit the claim.
Browse 400+ live receipts — each one ringing exactly where the work landed, green in lane, amber bleed, red drift — run the demo on your own trace, and hold our numbers to your own null.
“My engineering leads took one look and said they could clone it — a compressor, a seeded reef, a weekend of work. Suppose they were right. When we brought a homegrown gate to our carrier to underwrite the deployment, they asked for the one thing a clone can't produce: a license to the positional-geometry patent the policy prices against. We couldn't. We hadn't built a compliance instrument — we'd built a very efficient way to compress our own liability. And there's no version of telling your board your AI exposure is covered by something you reverse-engineered from someone else's patent. We licensed it instead.”
— VP of Engineering, Enterprise AI Integrator
A receipt you can license is defensible; a clone you have to explain is the exposure.
The answer is one command away · this is what caps your exposure
iamfim names the liability — you signed for the AI; when it drifts, you owe. The receipt at thetadriven.com/air-receipt is the artifact that turns drift into a priced, provable event and a mathematically defensible safe harbor — not a warranty, not insurance, but the record that caps unbounded balance-sheet exposure. The pipeline that produces the receipt is publicly runnable. On your laptop. In 30 seconds.
npx thetacog-mcp pmu-report --file your-agent-trace.txt
Receipt lands at ~/.thetacog/pmu/receipts/<id>.json · HTML report auto-opens in your browser · signature verifiable at /verify-receipt · the receipt body has 12-cell heatmap + Δ-map + σ-floor + market-match against a built-in compliance-officer job.
The beating heart: iamfim says "you're liable" → /air-receipt shows the schema → pmu-report produces the artifact → /verify-receipt closes the trust loop. Four URLs, one pipeline, no central server, no trust point. The artifact IS the proof.
What you're getting
One thing, one price: a license for every agent you run — $20 per agent, per year, under the patent. Volume unlocks the tiers.
The $100K you'll see below isn't a second price — it's 5,000 licenses × $20, the same rate at scale. That top tier is the one that also unlocks the Readiness Investigation (details below).
License the patent · $20/agent/year
Tier: TeamPatent Pending · US Patent App. 19/637,714 (Track One) — application, not a granted monopoly
One currency: the agent-year — $20 buys one. The price is fixed forever; the term each agent-year covers shrinks for later buyers, so early buyers lock in the most coverage per dollar. Why the term shrinks — the physics, the rules, the contract →
The top tier — $100K (5,000 transferable licenses) unlocks the Readiness Investigation
You push thousands of automated decisions an hour. Your logs prove a write happened — not that the agent stayed in its lane when it did. If a regulator or a class-action looks under the hood, that gap is what they find — and no bigger model closes it. Standard vendors can't either: by Rice (1953), software cannot ground its own semantics.
$100K buys 5,000 agent-years of licensing (at $20/agent/year, locked at today's price) and auto-attaches the Readiness Investigation: a time-boxed structural audit of your execution boundary — not a pilot, not a resilience program, a readiness investigation into where your digital intent already decouples from your hardware execution. We hand your General Counsel or Chief Risk Officer a concrete, deterministic map of where you're exposed. Not a legal opinion, not a guarantee — a baseline you can act on.
- • 5,000 transferable agent-years — a sublicensable patent right under US Patent App. 19/637,714, locked at today's price.
- • The investigation — where your digital intent decouples from your hardware execution, today.
- • The baseline — the exposure map for your board, before a plaintiff's attorney draws it for you.
Secure your whole supply chain, not just yourself
The agent-years are fully sublicensable. Use 2,000 internally; transfer the other 3,000 to your ten biggest vendors — now their agents run the gate under your license and hand hardware-verified receipts straight back to your compliance team. Under the EU AI Act and third-party risk rules you're already liable for the AI your vendors point at your data; this is the instrument that lets you enforce that boundary on them.
The first mover in a supply chain sets the standard everyone else has to meet — and the term each agent-year covers only shrinks from here. Whoever licenses first locks the deepest coverage and the position: your vendors adopt your receipt, not a competitor's.
This is a Risk & Legal line, not an IT line — you're buying a block of legal clearance to the IP, not software. Same license as the checkout above, at the scale that unlocks the audit. Clear it live on a card, or generate an invoice — net 30.
Set the checkout to 5,000 agents — pay by card →$100K clears live on a procurement card — no multi-week wire. Prefer a wire anyway? Invoice instead. Don't take our word for it — run the free proof first: npx thetacog-mcp attest-demo. Deeper reading: the manual · the Rice argument. Certifications for the humans who sign are bundled with your license tier.
The physics underneath · why you can trust the receipt
You just priced the baseline. Here's why it's real.
Everything below is the evidence — the mathematics that makes the receipt legally defensible, the 12-cell grid that decides in 10µs, the undecidability proof no software vendor can answer, the manual your officers read, and the standard forming around it. Scroll if you need to convince yourself; it only gets more grounded.
The Readiness Investigation is your Gap Analysis — the audit that lets you tell the board:
"We measured exactly where we're exposed. Here's the structural fix."
Your Certified AI Trust Officers (CATOs) conduct it — the humans who sign the receipt, included with your agent-years, not bought separately.
If You're Asking These Questions, This Is For You
CEO / Founders (especially mid-market)
"AI is my responsibility. I can't afford a governance team—how do I delegate this liability?"
Fortune 500s have departments for this. You have CATO—build an officer core internally, not a headcount line item.
CISO / Security Leaders
"How do I prove our AI isn't a liability?"
Gap Analysis gives you documented proof for the board.
Compliance / Risk Officers
"EU AI Act fines are €35M. Are we ready?"
Your CATO maps your gaps before regulators find them.
CIO / Chief AI Officers
"We can't promise it'll do better next time."
Without a Gap Analysis, you don't know what went wrong.
Board Members / Executives
"We need AI governance. Where do we start?"
Train your own CATO. Internal expertise, external validation.
See industry leaders facing this challenge →
Insurance & Reinsurance
- • Joachim Wenning (Munich Re) — €35M EU AI Act exposure
- • Oliver Bäte (Allianz) — €90B under AI management
- • Andrew Witty (UnitedHealth) — 150M lives affected
Financial Services
- • Jamie Dimon (JPMorgan) — $3.7T in AI-assisted decisions
- • Larry Fink (BlackRock) — $10T AUM AI allocation
- • Jane Fraser (Citi) — AI lending discrimination risk
Healthcare
- • Dr. Tom Mihaljevic (Cleveland Clinic) — AI diagnosis liability
- • Gail Boudreaux (Elevance/Anthem) — Prior auth AI
Tech & AI
- • Satya Nadella (Microsoft) — Azure AI attribution
- • Dario Amodei (Anthropic) — Constitutional AI limits
Not a Fortune 500? That's the point.
Mid-market CEOs are personally exposed. You can't hire a governance department—but you can certify your sharpest people as CATOs. Delegate the liability to an officer core you built, not a team you couldn't afford.
The Convention: license volume dictates standing
We are not selling consortium seats. Your standing in the standard forming around agentic-AI attestation is set purely by the volume of your transferable agent-years — your execution scale is your skin in the game.
Lead Co. — 5,000+ agent-years ($100K tier)
Hold 5,000+ transferable agent-years and you automatically become the reference implementation. You co-author the findings. You shape the standard everyone else implements.
Observer Co. — under 5,000 agent-years
Deploy a smaller fleet and you audit the methodology in action. Your voice is proportional to your active license count — grow it any time; the agent-years sum into your standing.
If you don't hold the IP licenses, someone else becomes Lead Co. — they shape the standard, you implement their rules.
Want to weigh in before you license? Sign the Snowbird Declaration or endorse the standard.
The human layer · an allocation, not a product
CATO is not something you buy — it's what your agent-years yield
CATO certifies the named human officer who signs the receipt. There is no CATO price. You receive one certification allocation for every 10 agent-years you license (10 → 1, 5,000 → 500). Certify your own officers with them — or, like the agent-years themselves, transfer the allocations to your vendors' compliance teams to extend the boundary across your supply chain. The manual is the reading that grounds it; certifications arrive with the volume.
This is a patent license and an attestation instrument, not an investment or security.
No financial returns implied.
Not ready to commit? Show your support another way:
One price, one decision — license your agents
↑ Back to the checkout