We build AI for decisions where a confident guess is expensive.
Language models are fluent, and the fluency is the problem. Ask one how much protein was in your lunch, or what a holding is worth, and it will answer — smoothly, specifically, and sometimes from nothing at all.
In most software a wrong number is an annoyance. In health and in money it is the whole product. So we don't ask the model to be careful. We build systems where inventing a number isn't something it can do.
Figures are resolved from real data and filled in after the model has written its sentence. A gate inspects the output as it streams and removes anything the model tried to state on its own.
A value read off a label is not the same kind of fact as a value inferred from a photograph, and the interface never lets them look alike. Provenance travels with the number, and a total inherits the weakest source it was built from.
An estimate is labelled as an estimate and shows its range. When the honest answer is that we don't know, the product says so instead of picking something plausible.
We don't claim to be the most accurate. That claim is unfalsifiable and everyone makes it. We claim something narrower and checkable: you can always see where a number came from.