Witnessed events · agriculture
Everyone is instrumenting the animal. Nobody is instrumenting the transfer.
Collars, ear tags, biometrics, virtual fence — there is real money in sensing what an animal does. But the moment ownership actually changes hands, at a set of pens forty minutes from pavement with no service, it is still a paper bill of sale and a handshake.
That moment is where the theft, the disputed count and the premium-program money all live.
Sensing produces claims. Claims are only worth money if they survive an audit.
A collar knows an animal grazed a paddock for 180 days. By the time that reaches a premium buyer it is a vendor’s assertion about its own customer’s animal, checked by a person with a clipboard.
The proof is a bill of sale
Ownership at the moment of transfer rests on a document written by one of the two interested parties. In a theft investigation or a contested delivery that is exactly the evidence you do not want to be relying on.
Nobody will share the underlying data
Stocking rates, weights, head counts and death loss are competitive information you do not hand to a neighbour, a buyer or anyone who might bid against you. This is the wall that kills most agricultural data products, and it is a rational wall.
Six parties, six organisations
Rancher, hauler, buyer, brand inspector, auditor, lender. No shared system, no shared login, and no reason any of them should have to trust one vendor’s account of what happened.
What a custody record would have to do to be worth adopting.
The same primitive we built for witnessed clinical and evidence transfers. Different payload, identical structure.
Both parties sign one record
Releasing and receiving, each bound to their own key, co-signing a single fact rather than producing two documents that can later disagree. Each transfer points at the one before it, so the chain from pasture to sale barn is continuous.
Prove one fact, withhold the rest
Every field is hashed separately, so a single claim can be proven on its own while head count, origin premises and the rest of the operation stay withheld — not redacted by us, simply never sent. This is the only shape of agricultural data product that gets adopted.
No shared enrollment authority
Records are self-contained and carry each signer’s full public key. A brand inspector or a lender verifies without an account anywhere and without trusting the rancher, the buyer or us.
It completes where the work happens
Sealed on the device at the pens, queued, and uploaded when there is signal again. The seal time governs, so a transfer recorded out of coverage is not a transfer recorded late.
Who actually needs this.
Probably not the individual producer, and we would rather say that than pretend otherwise. The value concentrates wherever every animal passes through.
The sale barn and the order buyer
One facility touches tens of thousands of head. A verifiable custody chain arriving with the animals is worth more to the party accepting delivery than to the party sending them.
The ag lender
Livestock secures operating loans, and inventory is verified today by counting or by trusting an affidavit. A borrower proving the ownership chain on a schedule — while withholding everything else — is a recurring need rather than a seasonal one.
The verified-program auditor
Age-and-source, never-ever, non-hormone-treated, grass-fed. We are not trying to replace the auditor — we are trying to make the evidence they already review verifiable instead of a binder.
Where this actually stands.
This is the earliest of our witnessed-event cases. We’d rather be straight about that.
What exists today
- The two-person seal is built, tested, and proven on live infrastructure — a two-signer record sealed by two real passkeys, synced, and independently verified
- Cross-organisation verification from day one — six parties from six organisations need no shared enrollment authority
- Selective disclosure — prove one field, withhold its siblings, mathematically bound to the same seal
- Offline-first sealing with a local queue; the seal timestamp governs
- A published
livestock_custody_transferrecord schema, riding the same envelope as every other witnessed event
What does not exist yet
- Any capture flow at a chute, and no RFID or EID tag reading of any kind
- Long-range presence for parties on opposite sides of a trailer — designed for, not built
- Any grazing, premium-program or verified-claim logic
- A single deployment, or a single head of cattle recorded
One thing we are deliberately not building on: the USDA electronic eartag rule is under active legal challenge and never covered feeder cattle or intrastate movement. Premium programs, theft and lending do not depend on a regulation a court can vacate, so that is where this is aimed.
We are asking a question, not selling a product.
If you run a sale barn, inspect brands, or lend against livestock: what does a disputed transfer or a bad inventory count actually cost you today, and who eats it? That answer decides whether we build this at all.