Engram
Shared food data is not intelligence. FoodBlocks let knowledge travel. Something still has to turn what arrives into a picture of one particular world, and that thing has to be private.
Shared food data is not intelligence.
A protocol lets knowledge travel. It does not make anyone understand anything. A record of every batch, price, delivery and review in a county is, on its own, a very large pile of true statements that nobody has read.
Something has to turn what arrives into a picture of one particular world. That is the Engram: the private, adaptive intelligence of a person or a food organisation.
What it does
It receives observations and blocks. From them it builds a living model of its owner's world — who they buy from, what moves in what season, which claims have held and which have not.
It learns from what actually happened rather than from what was declared. It anticipates: this arrives Wednesday, that certificate lapses in a fortnight, this customer has come back four times and will probably come back again. It simulates, because most useful thinking is about things that have not happened yet, and simulation is cheap when the world is a model.
Then it proposes. Information, or an action, or a question.
And it acts only with the authority it has been given. Nothing else.
Why it has to be private
The chain is shared. The model is not, and this is not a privacy feature bolted on afterwards — it is what makes the thing work at all.
Two people can read the same record and correctly reach different conclusions, because they want different things. A wholesaler and a household looking at the same seller's history are asking different questions of it. A shared model would have to answer both, so it would answer neither, and it would need to sit somewhere — which means somebody would own it, and then everybody's understanding of everybody else would run through one party.
So the Engram sits inside the boundary. Blocks cross in. Proposals and deeds cross out. What is never shared is the interpretation.
Travel and understanding are different jobs
This is the whole division of labour, and it is worth being exact about it.
The protocol's job is that knowledge can move. Signed, linked, verifiable by anyone, owned by nobody.
The Engram's job is that knowledge means something to one owner. Local, partial, opinionated, and answerable only to the person whose world it models.
Conflate them and you get one of the two failures we already have plenty of: a standards body that tries to decide what things mean for everyone, or a platform that holds everyone's understanding on their behalf and rents it back.
Proposes, does not decide
The last clause is the one doing the most work: acts only with appropriate authority.
Thinking is free. It can model, predict and simulate as much as it likes, because none of that touches the world. Acting is not free, and the boundary between them is not a setting — it is the shape of the thing.
An Engram that can only propose is one you can afford to let be wrong. That is the trade, and it is why the interesting question was never how clever it is. It is what it is permitted to do while being clever.