Open standards, and beyond them

How thinqOS speaks the open web, and what it adds on top

There is a quiet piece of plumbing sitting on more than half the world's websites. It is called JSON-LD, and it is a little island of structured data inside ordinary pages. No vendor owns it. No database contains it. Humans read the prose, machines read the island, and because the islands all point at a shared vocabulary, they connect into one graph without any central platform holding them. That is not a theory of portable meaning. It is the working state of the web.

We think agent memory should join that graph rather than hide from it. So thinqOS is built to speak those standards at its edges, and to go past them where they stop.

The standards we speak

A thinqOS Mind is not a proprietary blob. It exports as open linked data: JSON-LD, the format under that half of the web; YAML-LD, the same model in a more human-readable form, now advancing on the W3C standards track; and a Vault-LD folder, a directory of plain Markdown notes with linked-data frontmatter that opens in Obsidian and belongs to you. The vocabulary is not invented in a vacuum either. It composes schema.org and SKOS, the same terms the rest of the web already uses.

And the vocabulary is not hidden behind a login. thinqOS publishes a resolvable context at thinqos.com/ns/context.jsonld. Any linked-data tool can fetch it right now, resolve the identifiers, and read a Mind the way it would read any other island of structured data. Your memory can leave. Another system can understand it. It can join a wider graph without asking our permission. That is the test of ownership, and thinqOS is built to pass it.

Where the standards stop, a mind begins

Here is the honest limit of the open format. A vocabulary like schema.org can say this thing is that kind of thing and this note links to that one. It has no way to say I believe this at confidence 0.6, you believe it at 0.95, and here is where each of us got it. The standards serialize facts. They have no vocabulary for a point of view. Give an agent a folder of Markdown files and you have a wiki, not a mind. It remembers; it does not weigh, doubt, forget, or hold an opinion of its own.

That gap is exactly where thinqOS lives. On top of the open vocabulary it adds a small, documented layer for the things a mind needs and a file cannot hold: how sure a belief is, whose view it is, where it came from, how much it currently matters, and whether it has been superseded. A plain linked-data reader simply skips those extra terms and still gets clean, standard triples. A thinqOS reader picks them up and gets the whole point of view. One document, two readers, no fork.

Open memory, closed cognition

Put the two halves together and you get the position we actually hold. Your memory is yours, in an open standard you can walk away with at any time. The mind that reasons over it, the thing that forms beliefs, attaches evidence, tracks confidence, lets stale beliefs fade, surfaces contradictions, and picks the right slice for the task in front of it, is the product. We are not trying to trap your data. Data captivity is a weak thing to defend, and the open standards are coming for it anyway. We would rather give portability away on purpose and earn the stay on the quality of the thinking.

JSON-LD made the web a knowledge graph. YAML-LD lets an agent write to that graph in plain sight. A Vault-LD folder gives an agent a set of notes that is also a graph. thinqOS is the mind that reasons over the graph. A vault stores triples. A mind holds a point of view.

If you build with AI coding agents, the shortest path into all of this is thinqOS for Developers. And the fuller argument for why a mind beats a memory is our point of view.

thinqOS is a product of AI4Outcomes.