ProofField
Camera-first field notes that turn photos, audio, and corrections into structured evidence.
The workflow
ProofField is for inspections where stopping to fill a form is slower than documenting the problem. One press opens capture. A photo creates the record, audio adds context, and vision proposes fields while the user is still on site.
The user corrects the proposed details, attaches supporting files, and chooses a template. The final record can produce a consistent PDF and a QR-linked version for later access.
Evidence before prose
The machine is allowed to help with structure, not invent certainty. Proposed fields stay visibly editable. The original photo, audio, timestamps, and corrections remain attached to the record.
That matters because an inspection note can become evidence. A polished paragraph is harmful when it hides which parts came from the user, which came from a model, and which remain uncertain.
Product shape
The app is camera-first rather than form-first. Templates define the output without forcing the capture flow to match every eventual document field. The prototype is being built in React Native with mobile camera, audio, vision, and PDF surfaces.
Current state
The core flow is defined, but the product still needs field testing. The hard questions are practical: how quickly someone can correct a bad prefill with gloves on, how the app behaves with weak connectivity, and whether the generated report keeps a clear chain back to the original evidence. Those decisions matter more than adding another model-generated summary.