First with high-level, basic information as shown. The gap between the results is where we focus a human review — and in the production system, a change to the technical spec document automatically triggers a review for shift.
Gnomon points from evidence, so the reading is only ever as good as what you bring it. Have a real product in mind, and these six things to hand — the kind of detail a sound classification actually rests on:
A manufacturer’s datasheet sharpens all six — and the engine reads it against what you typed. If the document and your description don’t agree, the confidence softens, and it says so. Thin or mismatched in is never confident-wrong out; it’s a flag.
Have these to hand? You’ll see how far the engine gets — and where it stops and flags what’s worth checking.
Don’t? That scramble is the problem we exist to take off your desk. Begin anyway with our worked example, or talk to us.
“It points; you tell the time.”
The demo shows that the engine identifies products from a description with a measurable confidence score — and that a real datasheet sharpens (or honestly contradicts) the reading. It does not show the HS code, the export-control flags, or the rationale — those are what the product is for. They are what a paying customer pays for, and what a reviewer carries the audit trail of. The kernel: it points; you tell the time. The demo points. The full engine, in your team’s hands, is the reading.