Every build proves its math. Each model ships with its cited corpus, its recipe, its beat-base evaluation, and its receipt — so you never have to take our word for it.
Models flow down · Receipts flow up · Your health data never flows up
No black-box. Every model here ships with four things you can check yourself: its cited corpus, its recipe, its beat-base evaluation, and its receipt. A model isn't "proven" until it beats its base model on a held-out test — and we publish that math, win or lose.
Models flow down to your devices and the on-device helpers. Your private health data never flows up. The facts behind every model are published openly at DiabeticTruth. Models organize and explain — they do not diagnose.
Honest status, in the open. A build only graduates to a download after it beats its base on a held-out eval — not before.
A small open model that organizes and explains diabetic foot care — grounded in cited guidance, tuned to never diagnose. Destined to run on the LocalDiabetic edge box, on your own hardware.
Honest status: not yet cooked. The cited fact corpus is staged (20 facts from NIDDK & IWGDF 2023, every one sourced). No weights exist yet. When this model beats its base on a held-out foot-care eval, the receipt and the download land right here — not before.
Download — pending beat-base ✕ See the facts it learns →
The next builds, once the foot-care cook proves the recipe. Same spine: cited corpus → cook → beat-base → host.
Facts already published: Neuropathy · A1C / targets.
Five gates, every one published. No source → not in the corpus. No beat-base receipt → no download.
"The AI trust-me-bro is over." A model you can't verify is a model you shouldn't trust. We show the corpus, the math, and the receipt so you never have to take our word for it.