Diavie

The market thesis

Counting carbohydrates is the main daily pain for people with diabetes, and the leading apps are American: they don't know a French baguette portion from a bagel. A French-first companion, built on trust, has room to exist.

In health, the fastest way to fail is to build first and read the regulation later.

Rules before code

• Stay outside medical-device rules. Every feature was checked against the EU Medical Device Regulation (2017/745) before it was written: no dose calculators, no alerts, no interpretation of readings.
• Local-first by design. Health data never leaves the device. That keeps the v0 outside French health-data hosting obligations (HDS), costs nothing to run, and earns trust.
• Official routes only for sensors. File import first, then Apple Health or Health Connect; no unofficial APIs.

The product

• A journal for meals, readings, activity and notes.
• 207 French foods with home-style portions, plus barcode scanning through Open Food Facts.
• Descriptive statistics only, JSON export and a printable report to bring to a doctor.
• An installable app that works offline.
• A content site with 7 guides and a “first 90 days” kit, generated by a zero-dependency static builder.

Validation gates

Nine strategy documents cover the market, regulation, MVP scope and SEO (7 topic clusters, 50 planned articles). They set six testable hypotheses and three gates. No heavy build happens until the evidence clears the gate in front of it.

Honest status: the gates haven't been passed yet, so I don't claim validated demand. What the project shows is the method: thesis, constraints, smallest useful product, then evidence.

Numbers