How it works
How it works
Two kinds of thinking, working together. One reads and understands language. The other checks it against the record and the rules. Most AI only does the first. That's why it can sound right even when it's wrong.
It reads the material. Printed text, handwriting, and recordings, across many languages.
It connects the same people, places, and events across everything that mentions them, so one question can be answered from many sources at once.
It anchors every fact to the exact document and page it came from, so any answer can be checked against the original.
It catches contradictions. When two sources disagree, that surfaces, instead of being averaged away.
It marks what's missing. When the record is silent, the gap is shown, not filled with a guess.
The knowledge stays controlled by whoever holds it, and can be seen and withdrawn at any time.
documents, recordings, and records, in any format.
reads, connects, and anchors every fact to its source.
answers you can trace, cite, and check, with the gaps named.
Under the hood, this is a neuro-symbolic system: neural models read, a symbolic layer holds what's verified.
What using it looks like.
QuarterMill is one layer that meets you where you work, whether you open it directly, call it from your own systems, or connect it to a collection you hold.
A workspace you open
Ask a question in plain language and get an answer where every claim links to its source, disagreements are flagged, and gaps are named. Nothing is asserted without a receipt.
A layer your systems call
Your existing tools and AI systems call QuarterMill as infrastructure. It returns the same sourced, verifiable results behind the scenes, so answers in the tools you already use become checkable.
A connector for your collection
Connect your collection once. QuarterMill reads and links it while it stays in your hands. You see exactly how it is used, and you can withdraw it at any time.
What did the 1854 commission conclude about the district's water supply?
The commission found the supply unsafe and traced repeated outbreaks to a single public well1, and recommended its closure2. A dissenting member disputed the cause3.
Source 2 and Source 3 disagree on the cause. Both are shown, not averaged into one answer.
No measurements survive for the winter of 1853. The silence is marked, not filled with a guess.
- 1Commission Report, 1854Manuscript · p. 12 · City Archive
- 2Sanitary Recommendations, 1854Printed · p. 3 · City Archive
- 3Minority Statement, 1854Manuscript · p. 1 · City Archive