Tutorial · 11 min · reviewed September 2026
Publishing intent without a meeting
A procurement team spends weeks finding who can supply a thing. The discovery layer lets an organisation publish what it wants next, one document at its own domain — with a visibility no caller can override and an expiry no caller can set.
The meeting the format replaces
A procurement process begins with a search that is almost entirely manual: someone asks around, reads directories, takes meetings, and slowly assembles a list of organisations that might be able to supply what is needed. The discovery layer of the agentic internet, intent/1, is a format for the other side of that search — an organisation publishing what it intends to do next, so that a machine looking for a supplier can find the intent directly rather than by convening a meeting to ask.
An intent is a small document: what is wanted, in structured terms, at the organisation’s own domain, mergeable across a network. A buyer’s agent reads it, a supplier’s agent publishes it, and the discovery step that used to be a fortnight of relationship work becomes a fetch. The value is not that it removes the human relationship — the trust layer is where that lives — but that it removes the blind first phase where nobody knows who to even talk to.
One document per repository, at your own domain
The unit is one intent document per repository, served at your own domain and merged by anyone who wants to read across a set of organisations. You keep a private working record of everything you might want; you publish only the part you have chosen to make public. The format is deliberately small, so that publishing an intent is a routine act rather than a project.
$ node vendor-intent.mjs verify example.com
published # one of: published · absent · unreachable · invalidFiled is not published
The refusal that makes the format safe for a real organisation is that filing an item publishes nothing. Every item is private until a named human promotes it, and the reference constructor has no visibility argument at all — there is no parameter a caller could set to make an item public on its own. A public item with no promotion record is refused, and a promotion whose author is not a person is refused. Agents suggest; humans consent, and here the consent is a fact on the wire rather than a line in a policy document.
The served copy is a projection of the public tier only. A partner-tier item behind no gate is treated as a public item with a misleading label, and a checker refuses a non-public item found at a public URL. For a procurement team, this means the discipline you need — "we have not decided to say this out loud yet" — is enforced by the format, not left to whoever runs the publish step.
Expiry is derived, never declared
The second structural refusal is around staleness. Every item carries an expiry, and the checker refuses any value that is not exactly the assertion date plus the window for that kind of item — a blocker lives fourteen days, a decision thirty, a task ninety, an idea a hundred and eighty. A publisher cannot set their own expiry, which means they cannot keep a dead intent looking current forever, and cannot quietly shorten one either. The only way an intent stays alive is that somebody re-asserts it — looked at it again and said it again. A directory full of intents that have not been re-asserted is visibly a directory of things nobody is still pursuing.
Reading someone else’s intent without lying about it
When your agent reads other organisations’ intents, the consuming rules matter as much as the publishing ones. A consumer fetches over https only, with a timeout and a size cap, follows redirects only within the same publisher, and never reads a network failure as "they publish nothing". Unreachable is not absent — treating your own outage as evidence that another organisation intends nothing is a claim about them made from a fact about you. The verdicts are kept distinct — published, absent, unreachable, invalid — so a buyer’s system can tell "they want nothing" from "we could not reach them", which need opposite responses.
- The lab’s view (planned)FlashyLabs is preparing an engineering companion to this piece at https://flashylabs.com/insights/publishing-intent-without-a-meeting — planned, not yet published.
- The studio’s view (planned)The 4 Ventures thesis desk is preparing an investor-lens companion at https://4.ventures/thesis/publishing-intent-without-a-meeting — planned, not yet published.
Terms used here
Author
Name pending · engineering, agent infrastructure practice. Reviewed by the practice lead.
Cite
MLG Blockchain, “Publishing intent without a meeting,” 2026. TechArticle, machine-readable. https://mlgblockchain.com/insights/agentic-internet/publishing-intent-without-a-meeting
Prints cleanly, with URL and date in the running head.