QuarterMill · Est. 2026
Knowledge, kept whole in the age of AI.
The internet gave us access to knowledge. AI gave us answers. Neither held the knowledge together. QuarterMill is the missing layer between the world's knowledge and the machines that now speak for it: connected across every source, controlled by those who own it, and always traceable to where it came from.
For institutions, researchers, and press — whether you want to pilot QuarterMill with your own collection, see a demo, or just follow along.
Recognized by
The problem
Why this matters now
The world's knowledge is scattered across billions of documents, and the answer you need usually lives in no single one of them.
AI's answer was to melt all of it into a model, and lose the thread back to where any of it came from.
So it sounds certain, connects nothing you can check, and hides what it doesn't know. Where being wrong has a cost, that isn't good enough.
The shift
One question. The whole record behind it.
Ask about a single person today, and the answer is blended from whatever sources a model happened to swallow, with no way to see which ones it leaned on or whether any of them were right. QuarterMill works the other way around. It draws together everything that touches that person, a court filing, a census record, a newspaper column, a manuscript page, across every collection that holds them, keeps each piece tied to where it came from, and hands you the picture no single document ever held. And it never freezes into a snapshot. When a new source surfaces, the picture grows with it: fresh context is added, a claim the record no longer supports is struck through instead of quietly overwritten, and the places where it stays silent stay marked as silent. Not an answer a model once memorized, a living record that keeps its receipts.
What QuarterMill is
A different way to work with knowledge
QuarterMill doesn't absorb knowledge and speak over it. It holds knowledge living and connected, so you can move through it, follow a thread across everything that holds it, see the ground under every claim, and know where the record stops. The people who hold the knowledge retain control of it. AI can draw on it, but never rewrite it.
Connected.
The same person, place, or event, linked across every source that mentions them. One question, answered from all of them at once.
In its owners' hands.
The knowledge stays under the control of whoever holds it. It's drawn on, never taken.
Always traceable.
Every piece of an answer stays tied to the exact source it came from, and can be checked against the original.
The living record
The record never stops growing. Neither do we.
The world doesn't stop making knowledge when a model stops training, when a database stops updating, or when the last source is filed. Neither does QuarterMill. As the record grows, and as more of what people and machines produce goes on the line, the layer that holds it connected, sourced, and honest has to grow with it. That's what QuarterMill is built to do, especially where being wrong carries a real cost.
“AI learned to talk before it learned to cite. We're inverting that: building the layer that makes knowledge accountable, from the ground up, for the future of how we learn, decide, and remember.”
The foundation
Not a wrapper. Not a search box. A new foundation.
A chatbot and a retrieval tool do the same basic thing: fetch some text and let a model summarize it. The model still does the talking, still decides what to say, and still fills the gap when it's unsure. Nothing underneath it changed. A plain database has the opposite problem: it holds records but can't connect them, reason across them, or tell you what's missing.
QuarterMill is a third thing. It's a living layer of knowledge beneath the model, where the connections between sources, the control held by the people who own them, and the record of what's missing are built into the foundation, not added by the model at the end. The model draws on that layer. It never gets to overwrite it. Everyone else makes the model more convincing. We change what sits underneath it.
Positioning
Where we fit
Who it's for
Anywhere the answer has to be right, and has to be checkable.
We're starting with the people whose whole profession is keeping knowledge trustworthy: archivists, librarians, curators, and scholars. For more than a century they have set the standard for knowing where something came from, and we're building to that standard rather than around it. Their collections carry what most systems overlook: handwritten pages, many languages, fragile formats, and long silences in the record that deserve to be respected, not smoothed over. We're building for them first because they have been right the longest, and the same principles then carry to every field where being wrong has a cost.
Cultural heritage institutions.
The stewards of the primary record, with their collections kept in their hands and made answerable.
Researchers and historians.
Follow one question across many collections at once.
Law and public interest.
Show what the record says, and prove what it doesn't.
Science and medicine.
Trace a claim back to the study it came from, with nothing smoothed over.
Journalists.
Every claim resolves to a source, and contradictions surface.
Students and the public.
An answer you can check for yourself, and the habit of asking where every claim came from.
In early work with research libraries, public-interest legal organizations, and cultural heritage collections.
For institutions, researchers, and press — whether you want to pilot QuarterMill with your own collection, see a demo, or just follow along.