A photo of the fridge, and dinner is decided

Build a photo of the fridge, and dinner is decided.

Context: Point your phone at what's in the fridge and get one answer: cook this tonight.
Not a list, not a search, not twelve options ranked by rating. One meal, from
what's actually in there, starting with whatever goes off first.

Everybody answers "what's for dinner" about 365 times a year, and most of us
answer it badly at 18:40 while tired and already annoyed. The decision is the
problem, not the recipe — there are ten thousand recipes online and none of them
help at the moment you're standing in front of an open fridge.

The hard part is restraint. The moment it becomes a scrollable list it's just
another recipe app, and everyone has already deleted three of those. It has to
commit to an answer, with a small "something else" if the answer is wrong.

Recognition doesn't need to be perfect either. It needs to be honest — "this,
but you'll need garlic" is more useful than confidently inventing an ingredient
that isn't there. Missing one item is fine. Pretending is not.

Probably starts with one photo a week rather than one per meal, since nobody
will photograph the fridge daily. Track what gets used and what doesn't.

Requirements:
- Ship the smallest version that delivers the core value above, nothing more.
- Pick a stack you can deploy in a day, and state the choice before writing code.
- Handle the empty state, the error state and the offline case explicitly.
- Pricing to build toward: Free for a few photos a week, which is genuinely enough for most households.
$4/mo unlimited, plus a shopping list built from what's nearly gone. Not
subscription-first — people abandon cooking apps quickly and a monthly charge
for something you use twice makes them cancel angry.

Start by listing the screens, then build them one at a time.