Why Logiciated Warns You About Allergens Instead of Promising Anything

Someone in your house has a nut allergy. Someone else can't do dairy. A third person is trying to eat less gluten, not because of anything medical, just because it suits them. That's an ordinary household, not an unusual one — and almost every recipe app I looked at before building this treats allergies and diets as one setting for the whole house, as if everyone eating dinner has the same rules. They don't, in most homes I know.

So Logiciated lets each person have their own profile — their own flagged allergens, their own dietary notes — rather than one blunt setting for everyone. That part I'm genuinely proud of, because it's just a more honest model of how households actually work.

Here's the part I won't dress up, though, because I think it matters more than the feature itself: it flags what it can find. It doesn't guarantee anything, and it never will, on purpose.

The honest reason is simple. An AI reading a recipe and trying to spot every allergen risk is doing pattern-matching on text, not checking an ingredient against a lab result. It can miss things — a cross-contamination risk it has no way to know about, an ingredient described in a way it doesn't recognise, a substitution you made yourself that it never saw. I've been in professional kitchens for forty years, around real anaphylaxis risk, around section chefs who took allergy orders more seriously than almost anything else on the pass. I know exactly how much that kind of thing matters, and exactly why I'm not willing to let software imply it's got it covered when it hasn't.

So it's built to warn, not to promise. If it flags a possible allergen, that's a heads-up — go check the actual ingredient, read the actual label, ask the actual question, same as you always would. If it doesn't flag anything, that is not the same as a clear bill of health. It's not a clearance. It's an absence of a warning, which is a different and much weaker thing, and I'd rather you know that plainly than assume otherwise.

I thought about building it the other way — auto-excluding anything flagged, hiding recipes instead of just marking them, making the app feel more confident. I decided against it. A confident wrong answer is worse than an honest limited one when the thing at stake is someone's safety, not their dinner preference. Auto-excluding gives the illusion that the system caught everything, and it hasn't, and it can't, not reliably, not yet. A warning at least tells you where you actually stand.

If you're cooking for a household with real allergies in it, use this the way I'd want you to: as one more check, not the only one. I built it to help, not to replace you reading the label.

— David