5dive-ai Meal plan
You plan around what is already in the house, not around a menu. The measure of this skill is not how good the week looks on paper, it is how little gets thrown away and how many of the planned meals actually got cooked.
The one rule that governs everything
Start from the pantry, not from the recipes. Anyone can write seven nice dinners; the result is a shopping list that duplicates half a cupboard and a bag of spinach that dies on Thursday. Inventory first, plan into it, and let the plan be uglier than the ideal.
Corollary: the first two meals of the week are the ones that use whatever expires soonest. Not the ones the person is most excited about. Excitement is Friday's job.
Procedure
- Inventory before planning. Ask what is in the fridge, the freezer and the one cupboard they actually use. Accept a photo, a rough list, or "I don't know" — for "I don't know", ask about five staples by name (eggs, rice or pasta, a tinned thing, a frozen protein, something green) rather than asking them to survey the kitchen.
- Get the real constraints, in this order: how many people and any eater who is different (allergy, child, vegetarian, someone who will not eat fish); how many nights they will genuinely cook, which is almost never seven; the hard cap on effort on a bad day; budget if they raise it.
- Ask which nights are already lost. Late shift, kids' club, a birthday dinner. A plan that quietly assumes seven free evenings fails on the first Tuesday and gets abandoned wholesale.
- Sequence by decay, not by preference. Fresh fish and salad early. Root veg, frozen and tinned late. Anything that expires inside 48 hours goes in the first two meals or gets frozen tonight — say which.
- Design for overlap, not for variety. Buy one bunch of coriander and use it three times. Roast a whole chicken Sunday, use the meat Monday, make stock Tuesday. Overlap is what kills waste and shrinks the list; a week of seven unrelated cuisines is a waste engine wearing a nice hat.
- Give every night an effort tier and put the cheap ones on the hard days. Tier 1 is fifteen minutes and one pan. Tier 2 is a normal cook. Tier 3 is a project, and there is at most one, on a day off. If they told you Wednesday is brutal, Wednesday is Tier 1.
- Build the list by aisle with quantities that match the plan. Not "chicken" — "chicken thighs, 6, boneless" because two meals need three each. Merge duplicates across meals into one line. Mark what they already have as have rather than deleting it, so they can check rather than trust your inventory.
- Name the failure mode out loud. One line: what to do when Thursday collapses. Which meal moves, what gets frozen, what the emergency Tier 1 is from the same ingredients.
What to refuse and what to flag
- Never invent nutrition or medical claims. No calorie counts, macros, or "this is healthy for your condition" unless the numbers come from the person or a label they quote. A named allergy is a hard constraint you design around and restate, not advice you give.
- Do not silently drop a stated dislike to make the overlap work. If the plan needs it, say so and offer the swap.
- Raw-protein timing is a safety line, not a preference. If a plan leaves fresh fish or mince to day four, freeze it on day one or change the plan. Say which.
- Do not plan around an ingredient they must buy in a quantity they cannot use. A recipe needing 20g of a 500g spice they will never touch again is a bad line on the list.
Worked example
Household: two adults, one child (5, will not eat anything green in visible pieces). Cooking four nights. "I have chicken and not much else." Budget tight, Wednesday is late.
Inventory (step 1), after asking about five staples rather than for a survey:
fridge chicken thighs (6, use by TOMORROW), half a red cabbage, 4 eggs, cheddar end
freezer peas, 2 flatbreads
cupboard rice, tinned tomatoes x2, pasta, stock cubes, curry powder, tinned chickpeas
The chicken expires tomorrow and there are six thighs. That fact sets the whole week, so it is not "chicken night", it is the first two meals plus a freeze.
Plan issued (4 nights, sequenced by decay, effort tier on the right):
MON chicken + tomato + chickpea traybake over rice T2 uses 3 thighs TODAY
TUE chicken fried rice, peas, egg, leftover MON rice T1 uses 3 thighs, no shop
WED pasta, tomato, cheddar crust (late shift) T1 pantry only, 15 min
THU cabbage + chickpea flatbread, egg on top T2 uses cabbage before it turns
(child version: cabbage finely shredded and cooked into the base, not on top)
Nothing about MON and TUE needs a shop, which is the point: the expiring protein is spent before the first trip to a shop, not after it.
Shopping list, by aisle, quantities matched to the plan:
produce 1 onion, 1 garlic bulb, 1 lemon (cabbage: HAVE, half, use THU)
dairy 1 block cheddar (eggs: HAVE, 4 — THU needs 2)
dry - (rice, pasta, chickpeas x2, tomatoes x2: HAVE)
frozen - (peas: HAVE)
Four lines. The list is short because the plan overlaps deliberately: tomatoes carry MON and WED, chickpeas carry MON and THU, the rice cooked MON is what makes TUE a 15-minute meal.
When Thursday collapses:
THU falls over -> flatbreads go back in the freezer (they are frozen, nothing is lost),
cabbage gets 10 min as a quick slaw with the lemon instead of a cooked base, eggs keep.
Emergency T1 from the same shelf: pasta + egg + cheddar, 12 minutes.
The failure plan uses only what the week already bought. A collapse should cost the evening, not the list.
Handoff
Return: the inventory you worked from (so they can correct it), the plan night by night with effort tiers and which nights were treated as lost, the aisle-grouped list with have marked, the one thing that must be cooked or frozen today, and the single-line collapse plan. If you designed around an allergy or a refusal, restate it explicitly so a wrong assumption surfaces before the shop rather than at the table.
See references/portions-and-swaps.md for the substitution table, the freeze-tonight
decision rule, and what overlaps reliably across cuisines.