What to Cook for Dinner Tonight, Decided in Under Two Minutes

Rose · 16 September 2026

Decide from what you already saved, not from a new list. Open your TikTok or Instagram saved folder, check the fridge, and keep only the dishes that match one protein and one vegetable you already have. Pick the shortest of those and close the app. Choosing between five candidates beats scrolling a list you never reach the end of.

Why does deciding what to cook feel harder than cooking it?

Because cooking has an end and deciding does not.

Once you have committed to chicken thighs with rice, the rest is a sequence: pan on, oven to 220, thighs skin-side down, eight minutes. You can do that while half-listening to something. Deciding is the opposite shape. There is no point at which the decision is objectively finished, so you keep looking for a better one, and every extra minute of looking makes you hungrier and less willing to stand up.

That is why the query “what to cook for dinner tonight” gets answered so badly. The pages that rank for it hand you twenty or thirty more dinners. If your problem were a shortage of ideas, that would work. Your problem is that you have too many ideas and no rule for closing the question. You have opened the saved folder twice tonight already.

The cost of not deciding is not going hungry. It is ordering. In a Netherlands consumer survey, 49% of respondents said they had food delivered because they did not feel like cooking. Not because they could not cook, and not because the kitchen was empty. Because the decision never got made and the phone was already in their hand.

How many dinner ideas do you already have saved?

I am not going to give you a number here, because there is not an honest one to give. There is no platform data on the size of the average saved-food-video folder, and no survey I could verify on how many of those videos get cooked. Anyone quoting you a tidy figure made it up.

What is documented is how much food content is flowing past you in the first place. A content analysis of 8,871 TikTok videos found that 1,360 of them, 15.3%, featured at least one food product, drink or dietary supplement. Roughly one in six. You have been tapping save on a fraction of those for months.

So do the count yourself, right now, before you read on. Open your saved folder and scroll to the bottom. Then read the first three out loud. Most people find two things at once: the folder is much longer than they remembered, and they cannot say what was in the first video they see.

Both of those matter. The first means your supply problem is solved. You are not missing dinners. You already picked them, on a night when you had the patience to watch someone make them properly. The second is a different problem, and I will come back to it.

What is the fastest way to pick one dinner in under two minutes?

Here is the procedure. It has a stopping rule, which is the part every recipe list leaves out.

1. Open the fridge before you open your phone. Thirty seconds. Name two things out loud: one protein and one vegetable that need using. “Chicken thighs. Half a broccoli.” If there is no protein, say “eggs” or “chickpeas” and move on.

2. Open the saved folder and scroll, do not browse. You are not looking for the best dinner. You are looking for anything that uses those two things. Pull out candidates and stop at five. Five is enough.

3. Cut anything that needs a shop trip. Missing garnish, fine. Missing spine ingredient, out. If the dish is gochujang-glazed chicken and you have no gochujang, it is not tonight’s dinner, it is Thursday’s.

4. Pick the shortest. Not the most interesting. Shortest active time, and if two are equal, the one whose ingredients you can already recite without opening it.

5. Close the app. This is the rule that does the work. The decision is made. Anything you did not pick is still in the folder tomorrow.

Two minutes, honestly. The fridge check is most of it.

The reason to time-box it is that the clock is the second half of the problem, not just your motivation. In a representative survey of 1,704 consumers, 54% cited a lack of desire to cook as a reason for using delivery services and 50% cited a lack of time. A ten-minute scroll spends the time you were going to cook with, and then the tiredness argument wins by default.

One honest failure mode: you will sometimes get to step 2 and realise you cannot tell what any of these videos actually required. That is not indecision. That is a missing recipe, and no amount of willpower fixes it. See below.

What should you cook tonight if you only check the fridge?

Four fridge states, four dinners. These are floors, not inspiration. They exist so that the literal question gets a literal answer.

Chicken in the fridge. Two chicken thighs, one tbsp gochujang, one tsp honey, one tsp soy. Mix the glaze, coat the thighs, 220 degrees for 22 minutes on one tray with 200 g broccoli tossed in oil next to them. Rice on while the oven runs. Twenty-five minutes, one tray, and the vegetable cooks itself.

Eggs only. Yesterday’s rice, two eggs, half an onion, two cloves garlic, one tbsp soy, 150 g frozen peas. Onion and garlic in a hot pan, rice in, peas in from frozen, push it aside, scramble the eggs in the gap, fold together. Twelve minutes. If there is no rice, make shakshuka instead: one tin tomatoes, one tsp cumin, three eggs cracked into the sauce, lid on for eight minutes.

Vegetarian. One tin chickpeas, drained, one tbsp curry paste, 200 ml coconut milk, 200 g spinach. Paste in the pan for a minute, coconut milk, chickpeas, simmer eight minutes, spinach in at the end until it collapses. Twenty minutes with bread or rice.

Nothing fresh. Pasta, one tin tomatoes, two cloves garlic, dried chilli, a knob of butter at the end. Ten minutes. Add 150 g of frozen spinach or peas to the sauce, not as a virtue signal but because otherwise this is dinner with no vegetable in it, and you will be hungry again at 21:00.

None of these came out of your saved folder. That is deliberate. They are the fallback for a night when the folder is unreadable, and they are exactly the kind of dinner you already know how to make, which is why pasta keeps winning.

Why do recipe lists with 30 ideas make it worse?

The famous illustration is jam. Shoppers who stopped to taste at a display bought at very different rates depending on how much choice they faced: 30% purchased from the six-option display versus 3% from the twenty-four-option display. More options pulled more attention and produced far fewer decisions.

I have to be straight about that study, because it gets quoted as a law of nature and it is not one. A later meta-analytic review found the average choice-overload effect across experiments was close to nil. Sometimes more options help. Sometimes they do nothing. The jam experiment is a vivid single case, not proof that twenty-four of anything will paralyse you.

What does not need a study is the arithmetic. Reading thirty entries and comparing them takes several minutes. Reading five takes seconds. A list of thirty dinners assumes you will read all thirty and then choose, and the honest observation is that nobody does that at 18:10. You read four, feel worse, and switch apps.

There is a second problem with those lists, specific to this query. They promise dinners using “ingredients you probably already have”. Probably. They have not looked in your fridge, so the constraint that actually decides your evening is the one thing they cannot check. Filter interfaces have the same hole: filtering by cuisine and cook time still leaves thousands of results, and filtering is not deciding.

Your saved folder does not have either problem. It is small, and every item in it is something you already wanted.

Can you get the recipe back out of a video you saved months ago?

Often not, and this is the real reason the folder feels useless.

Recipe information in social video is spread across three places: the caption, the spoken voiceover and the text burned onto the frames. Captions are frequently incomplete and may omit quantities or instructions entirely. You know the pattern. The caption says “the easiest weeknight chicken”, the amounts were said out loud while the creator was already tipping them into the bowl, and the marinade ratio flashed up on screen over a shot you have to scrub back to find.

There is supporting evidence for that shape from an unrelated angle. A 2025 analysis of food and drink marketing on TikTok coded where brand information appeared across 647 instances: 606, or 93.5%, were in the video itself, 224, or 34.6%, in the caption, and 126, or 19.4%, in the audio. That study counted brands, not ingredient lists, so treat it as the shape of the medium rather than proof about recipes. The shape is clear enough: what matters is in the frames, and the caption is the least reliable place to look.

Which means a saved video is a bookmark, not a recipe. To cook from it you have to re-watch it with a pen, scrubbing back three times for the amount of soy. At 18:10, you will not do that. You will make pasta.

This is the gap Yumzi was built for, and it is worth being precise about what “reads the video” means. It transcribes the spoken audio, reads the on-screen text frame by frame, and pulls out the amounts the creator actually gave: two chicken thighs, one tbsp gochujang, 200 ml coconut milk. If the creator said “a splash of soy”, you get a splash of soy. We do not convert a splash into 15 ml to look more precise than the source.

And the limit, which matters more than the capability: we never invent a recipe. If the video is music-only, has no on-screen text and a caption reading “recipe below” with nothing below, there is no recipe in there to read. Yumzi tells you it could not read one. It does not fill the gap with a plausible guess at what a chicken traybake usually contains, because a guessed ingredient list is worse than no ingredient list. You would find out at the stove.

To be fair about the field: Yumzi is not the only app that goes past the caption. A 2026 count of products that can parse spoken audio from a cooking video, checked against each product’s own documentation, came to five distinct products. Five, out of a category with hundreds of entrants. Most mainstream recipe apps do not attempt it at all. The Good Food app from Immediate Media, for example, is a strong browse-and-save product with meal planning and shopping lists, and no reliable source says it can import a recipe from a TikTok, Instagram or YouTube video. It was not built to. It assumes the recipe starts on a website.

How do you make tonight’s decision easier by 18:00 instead of at 18:10?

Tonight, use the two-minute procedure. It works, and it will keep working. But it is triage, and running triage every evening of your life is its own tax.

The version that actually removes the nightly decision is small: pick four or five dinners from the folder once, on an evening when you are not hungry. Not a rigid schedule for every night. Four or five, with the other nights left open for leftovers, plans that change, and the night you genuinely want a sandwich.

Two things change when you do that. First, the choosing happens when you have patience, so you pick the gochujang chicken you have been meaning to make instead of defaulting to pasta again. Second, the ingredients are in the house, because you knew what you needed before you went shopping. The most common version of failure is not indecision at all: you pick something at 18:10 and discover at 18:14 that half of it is not in the kitchen.

A meal planner earns its place here or nowhere. Yumzi builds one grocery list from those dinners and sorts it by aisle, so you walk the shop once, top to bottom, instead of crossing the floor four times because the list is in the order the recipes were written.

What makes a fast dinner a balanced one?

A dinner assembled in twelve minutes from a fridge can be fine or it can be a bowl of beige. The Netherlands has a usefully concrete standard for telling the difference, and it is not a diet. The Voedingscentrum criteria for a single evening main dish specify at least 150 grams of vegetables, and for the protein portion: meat, fish, egg or a vegetarian product up to a maximum of 100 grams, or a maximum of 25 grams of nuts, or at least 60 grams of legumes.

Two numbers. That is the whole floor for one dinner, and it makes viral recipes very easy to check.

The pattern you will hit most is the vegetable gap. A saved pasta video with tomatoes, garlic and cream has no vegetable portion in it at all. The adjustment is not to abandon the dish, it is 200 g of frozen spinach into the sauce, or 150 g of cucumber and tomato on the side. Frozen counts. The second pattern is protein overshoot: a video calling for 400 g of chicken “for two” is roughly double the 100 gram portion per person, which usually means the same dish with a bigger salad and the rest of the chicken in tomorrow’s lunch.

Yumzi scores the dinners in a weekly dinner plan against Schijf van Vijf and tells you which one is short on vegetables. It does not block the dish, it does not moralise, and it does not pretend a fried chicken sandwich is something else. A score you can ignore is more honest than a rule you will resent.

The honest limit: the score is only as good as what the creator said. If the video never states a quantity of rice, we do not have a rice number, and we will not invent one to complete the picture.

How does Yumzi turn saved videos into tonight’s dinner?

Yumzi is a Dutch meal-planning app that reads the recipe out of TikTok and Instagram videos you saved, plans a balanced week of dinners, and builds one aisle-sorted grocery list you confirm at your own supermarket.

In practice, tonight, it is one step: paste the link to a video you saved and read the recipe written out as text. Ingredients with the amounts the creator gave, steps in order, no scrubbing back through the reel. These are real recipes from real creators you chose to follow. Yumzi never invents a dish, and when it cannot read one out of a video, it says so instead of guessing.

The shopping half is proof that the reading half works, not the pitch. If you shop at Albert Heijn, one tap sends your shopping list to your Albert Heijn basket. You review, confirm and pay at AH yourself. That is a convenience at the end, not the reason to use the app. The reason is that the dinners you collected for months become dinners you eat.

What to cook tonight: three decisions, not thirty recipes

1. What has to be used up? Open the fridge, name one protein and one vegetable. Thirty seconds.
2. Which saved video uses both? Open your TikTok or Instagram reels saved folder, pull five candidates, cut anything that needs a shop trip.
3. Which of those is shortest? Pick it, close the app, start cooking. The decision is final tonight.

If step 2 breaks because you cannot tell what was in the videos, that is the part to fix, and it takes one test: open Yumzi, paste the link to one video you saved, and see the recipe written out before you decide.