AI weekly meal plan — stop deciding what's for dinner, start eating it

It's Sunday evening. Mia is standing in front of the fridge trying to work out what the family will eat next week. Olle, 7, won't touch fish on Mondays. His brother Hugo, 11, has handball Wednesday and needs something quick. Dad is vegetarian half the week. Mia wanted 30 minutes to herself — and she's just spent them on a spreadsheet.

This is one of the most exhausting parts of parenting: not the cooking itself, but the planning before it. Today we're shipping AI weekly menu on FamilyBoard for web.

What it is

At /app/menu you'll find a single text box. Write one sentence about how your family eats:

"We're four — two adults and two kids (7 and 11). No nuts. One vegetarian. Easy recipes during the week, something a bit more fun on Friday."

Hit Generate. Within ten seconds you get back:

• Seven dinner suggestions, one per day
• A consolidated shopping list where duplicates are merged (eight tomatoes plus two tomatoes = ten tomatoes, not two rows)
• Quantities scaled to your family size
• Ingredients grouped by store aisle so you don't zigzag the supermarket

Tap Save as list and a shared shopping list appears in FamilyBoard Lists — visible to the whole family. Whoever shops on Saturday can tick items off, and the other parent sees updates in real time.

Why we built it

We spoke to 47 parents over the autumn. The question we asked wasn't "what's hard about cooking?" but "what do you do before you even start?". The answers were almost identical: 30 to 40 minutes on a Sunday afternoon, a mix of fridge inventory, calendar checks and a negotiation with a partner. For some, a co-decision. For many, a solo job. For nearly everyone — a mental tax.

AI weekly menu takes off the first 80% of that work. You get a serious starting point, not a blank page. You can still swap Thursday's pasta for something else — but you don't have to invent the whole week from scratch.

How it works under the hood

There's a Supabase edge function called generate_weekly_menu. It sends your prompt to gpt-4o-mini with a structured output schema — so we can parse the response into real lists rather than a blob of text. The shopping-list consolidation happens in the same pass: the model is told to merge identical ingredients and normalise units (turning twelve tablespoons of olive oil into a cup, for instance).

We don't store your prompt or the menu on our servers beyond the duration of the call. No tracking dataset, no model training on your dinners.

The free plan gets 5 generations per month — enough for once a week plus a little experimentation. Premium has no quota.

Trade-offs we made on purpose

No photos. Images would make the menu prettier but would double generation time and triple cost. A family that's already decided on sausage stroganoff for Wednesday probably doesn't need a mood board.

No recipes baked in. We suggest what you'll eat, not how to cook it. There are already a thousand recipe apps that do that better than we ever will. Need the method? Google "sausage stroganoff kids" — you'll find twenty good options.

Web first, iOS later. This kind of input is just easier on a keyboard than a phone. The iOS version will follow once we've made the flow as smooth there as on the web.

Try it

Sign into familyboard.io and head to Menu in the sidebar. Write how you eat, hit generate, save as list. Ten seconds from empty fridge to shopping list.

We didn't build this to be first with AI in a family app. We built it so Mia gets her Sunday evening back.