Do you worry about $1.2B in AI liability?
If the property is trivial, software can check it — and why are you paying to check trivial properties? If it isn’t trivial, Rice’s theorem says nobody can. So we fixed the math.
press enter
Who did this make you think of? We’d love to know.
Free to run · license the patent per agent
Installing and running the drift-gate costs nothing — that is the lighthouse, and it stays lit. Paying is the bonus. When you run agents in production under the patent, you take a license: one per agent you run, billed annually.
What you're buying: a license, not a login.
The entry point is the $100K Governance Harness — the Readiness Investigation, a structural audit of your execution boundary as it actually runs today, plus the infrastructure to run the gate across your fleet. It carries 5,000 agent-years of licensing with it.
The $20 unit is how you scale out after the harness is in — not how you start. We don't sell five-agent pilots. A five-agent trial proves nothing about a five-thousand-agent fleet, and we'd rather tell you that now than six months in.
It's a lease, and it has a mileage cap
An agent-year is 365 days or 10,000 attestations — whichever comes first. An attestation is one signed receipt over one bounded piece of agent work, and you can count them yourself. For scale: this repository burns about 56 a day, so a hard-working agent runs through roughly two agent-years in a calendar year. We publish that number because it is the only fleet we can honestly measure — ours.
“We'll just self-insure”
You can, and at small scale you probably should. But staking, wagering, skin in the game — every version of putting your own capital behind an agent's behaviour is self-insurance wearing a newer word. The cost is not the reserve you set aside. It is the ceiling: you can only run as many agents as you are willing to lose, so your fleet size gets set by your risk appetite instead of by your market. That is how you keep the cars in the garage.
A recomputable record changes what there is to argue about later — not whether the work was good, only where it landed. Whether that moves anyone else's underwriting is their decision and not ours: we are not an insurer, we bear no risk, and we promise no outcome.
The only arithmetic that matters: what are your agents already authorized to do?
Don't ask what this saves you on insurance — we can't answer that, and anyone who does is selling you a carrier's decision they don't control. Ask the question you can answer yourself, today, without leaving the room: how much authority have you already handed your agents?
Worked example
Two tenths of one percent of what's already moving, to make all of it auditable — a signed, recomputable record of where every agent action landed against the rules it was given. That is not a software purchase you have to justify. It is the cheapest line on the risk register, and it comes out of the risk budget, not engineering's.
The drift-gate
Free, forever
One install turns any agent into a verifiable, bounded one. 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. Every action mints a signed receipt that recomputes on your own machine; nothing to trust.
What it buys today — and what it unlocks later
Today, with nobody else's permission: when an agent does something expensive, you stop paying people to reconstruct what happened. The signed receipt already says where the action landed and against which rule — so the audit hours, the engineering archaeology, and the legal discovery that currently follow every agentic incident mostly go away. That value doesn't need a carrier, a broker, or a committee. It lands the day you turn it on.
Later, and this part is a direction, not a promise: a unit an underwriter can rate is what turns AI from an exclusion into a line. We are not an insurer, we bear no risk, and we will not tell you what a carrier will charge you — that is their decision. What we can say is that they have never had a recomputable unit to price against, and now one exists.
Running agents — license below, self-serve. Underwriting, GRC, or carrier-side, and want a say in the standard rather than a login? Start at The Table →
Why both sides hold a key
You license to run. The deployer key is one per agent — it mints receipts you can recompute yourself, and that value stands alone with nobody else in the room.
A carrier that wants to rate against those receipts registers on the other side. Not because we require it — because a receipt it can recompute and stand behind to its own reinsurer is the only kind it can price. The key is a crypto attestation that the receipt was minted under the patented method, so neither side has to trust the other; each only has to recompute.
We sell the same recomputable unit to both sides and impose no condition that either accept the other — the two-sidedness is the carrier's admissibility standard, not a gate we control. We are not an insurer, we bear no risk, and whether a given carrier honors a given key is their decision, not ours.
Which budget this comes out of
The objection is rarely the price — it's “whose budget is this?” It doesn't need a new one. This license draws from lines you already fund:
You don't have to win a new line-item fight to run this. Point your broker at the loss-control line; point counsel at the reserve.
One flat price per production agent, billed annually. Free to run below production; the license covers each agent you run under the patent.
Per-agent annual license
$20 / agent / year
A fleet is just the unit times your agent count — 10 agents is 10 × $20 a year. No tiers, no negotiation, no surprise.
License an agent — $20/yr →Locking in now gets you the most coverage per dollar you'll ever get — that's the reason to prepay. A license starts covering a full year, and that coverage window only shrinks from here; not a SaaS subscription, a prepaid engagement. You buy the license count, any volume, any time, at the flat rate above — no threshold required, no audit attached.
One threshold: ~5,000 licenses (≈$100K)
Crossing it auto-attaches the Readiness Investigation
Below the threshold, you're just buying licenses at the flat rate — no audit, no consulting, nothing bundled in. Cross it and the purchase automatically triggers a structural audit of your execution boundary: we find out how exposed your current systems already are — not hypothetical future risk, the liability that obtains today, under your practices as they actually are right now — and hand your counsel something concrete to act on, not a vendor pitch to push back on.
$100K is the anchor a 5,000-license rollout has landed at, not a rate card — every deal at this scale is scoped.
Everything above is the license a deployer buys to run its own agents. If you run agents for clients — an integrator, an MSP, a consultancy — you can buy licenses before you know which client will end up running them, and re-register them once you do. That is a provisioning convenience, not a channel. A receipt only peers when both endpoints are registered, so whoever ends up running the agent ends up registered either way — there is no position to hold.
Transferable for use — not tradeable
A license can be re-registered to the client, vendor, or portfolio company that will actually use it, so you don't have to name the final deployer on day one. Re-registration runs through our registry — a flat $2.00 administrative fee per agent-year — and the license keeps the caps it was minted with.
What this is not: a trading position. Licenses may not be re-registered above the unit price plus that fee, we don't operate a secondary market, and nobody should buy one expecting it to be worth more later. It is sold for use. If you want the margin, charge for your work — not for our paper.
What this does — and doesn't — decide for you
We don't decide whether your agent gets to act without waiting on a human. That call is your systems', your policy — we stay out of it. What the license buys is the record: a signed receipt, grounded in the Unity Principle (S=P=H), that makes whatever your systems already decided provable after the fact.
First, what it gets you: access to the zero-latency economy — letting an agent move now, at machine speed, because you can prove after the fact where it moved. That access is the reason to buy.
Second, what it protects you from: officers are increasingly on the hook for silent AI decisions with no record at all — the same Caremark-style oversight duty that already applies to every other system you run. That's the reason to buy now rather than after the first suit names someone.
What you are buying — and what you are not
An Authorized Broker keeps 90% of every license they place; we keep 10% at re-registration to run the neutral ruler. The margin is compensation for enforcing the standard of care — which is exactly why a Fortune 500 procurement team can put an Authorized Broker on a vendor form and buy with confidence. The measurement itself is open source (MIT); only the financialization is licensed.
The Authorized Broker program →thetadriven.com/pricing • Trust Physics • US Patent App. 19/637,714 — 36 claims, filed Apr 2, 2026 (Track One)
