Where you are
Tuesday tomato pasta
One pan, twenty minutes, familiar ingredients, no recipe needed.
A meal you can already navigate
The bridge from the meals you know to the meals you want.
An example path
One familiar meal becomes a way forward.
Illustrative, not a recommendation.
A reachable next step
Tomato and lentil pasta
The same pan and sauce logic, with one new ingredient doing a clear job.
New enough to create movement
Your direction
More plant-forward weeknights
A wider repertoire without rebuilding the way Tuesday evening works.
A goal, not a prescribed menu
The proposal preserves what is already easy, changes one meaningful thing, and points in the direction the person chose.
Try it → respond → update the map
Discovery is not progression.
Finding an exciting dish is easy. So is receiving a perfectly organised recipe plan. The hard part is knowing what has to change between the way you cook today and the way you want to cook next.
Right now, that translation lives in your head. You have to judge the technique, equipment, ingredients, time and confidence each recommendation assumes. When the jump is too large, the recipe is saved, skipped or abandoned.
The missing product is not another destination. It is a bridge from your actual starting point.
Fingerprinting taste
Names the destination
Useful for discovering restaurants, dishes and flavours you may enjoy.
Meal plans
Hands you an itinerary
Useful when the prescribed recipes already fit your skills, kitchen and week.
The missing layer
Understands the distance
Where you are, where you want to go and which next move is realistically yours.
Before Culinae can guide you, it has to locate you.
Not with a personality label. With evidence from the way food already works in your life.
- 01Your anchor meals
- The dishes you can make without instructions, including shortcuts and improvised versions.
- 02Your cooking moves
- The techniques you trust, the substitutions you make naturally and the places you hesitate.
- 03Your real kitchen
- The ingredients, tools, shops and leftovers that shape what is possible at home.
- 04Your actual week
- The time, energy and occasions that decide whether a meal is realistic on a given day.
A goal gives the map direction. It should not erase the starting point.
You may want more vegetables, more confidence with another cuisine, less delivery, less waste or a technique you have never trusted yourself to try.
Culinae looks for an adjacent move: enough familiarity to be doable, enough difference to create progress.
How Culinae judges whether a next step is reachable.
01Aroma compounds
An example person likes garlic and dislikes onion. Onion has 42 compounds that garlic does not, so each one is a suspect. Every new answer clears some of them until only a few are left.
- Still a suspect
- Cleared by this answer
- Cleared earlier
Likes garlic. Dislikes onion.
Onion has 42 compounds that garlic lacks.
42suspects
Dislikes leek.
Only compounds leek also has can explain both dislikes.
14suspects
Cleared 28 that leek does not have.
Dislikes shallot.
Only compounds shallot also has can explain all three.
6suspects
Cleared 8, including propenyl propyl disulfide, ethyl acetate and furfural.
Likes beef.
A compound in a liked food is unlikely to be the cause.
3suspects
Cleared 2-methyl-2-pentenal, 2-tridecanone and methyl mercaptan.
Likes apple.
Same test with a second liked food.
2suspects
Cleared hexanal.
Two suspects are left
- Dipropyl trisulfide
- Methyl propyl trisulfide
Each one is found in only 11 of the 1,525 ingredients.
What the profile now predicts
- Likely dislikes, with both suspects
- Chive and onion oil.
- Possible dislikes, with one suspect
- Cocoa, durian, french-fried potato, nobiru (a wild onion), peanut butter, roasted cocoa, roasted peanut and scallion.
Best next question
Ask about chives.
Chives contain both suspects. If the person likes chives, both are cleared, and compounds do not explain the dislike.
Data: Ahn et al. (2011), flavor network compound lists. The person and their answers are an example, not real data. The lists show which compounds are present, not how much, so the result is a hypothesis to confirm by asking. Roasted onion contains all 42 of the onion-only compounds, so liking roasted onion but disliking raw onion cannot be explained by compounds here.
02Culinary networks
Ingredient combinations recur in different patterns across cuisines. These networks help Culinae find changes that remain legible within a food tradition, or form a sensible bridge between traditions.
A network describes recurring structure. It is context, not a rule about authenticity.
Italian
7,493 recipes. Herbs, Additives and Spices touch 16 of 19 links.
Hover, tap or focus an ingredient type to see its numbers here and in the other graph.
Mexican
3,137 recipes. Vegetables touch 15 of 19 links.
Hover, tap or focus an ingredient type to see its numbers here and in the other graph.
Highlight an ingredient type in both graphs
Each graph is a maximum spanning tree: the strongest set of links that connects all 20 ingredient types without loops. A link that is not drawn can still exist in the recipes. The recipe sets are samples and do not define either cuisine.
Data: Caprioli et al. (2025), The networks of ingredient combinations as culinary fingerprints of world cuisines, doi.org/10.1038/s41538-025-00588-4. Code and data: github.com/claudiocaprioli/ingredient-combination (MIT license).
03Meal vectors
Recipes can be placed on a path from the simplest form to more complex ones. Here every bread keeps one dough, 500 g of flour to 300 ml of water plus salt, and each branch adds or changes one thing. Each change asks for a new ingredient, technique or skill.
Flatbread (you bake this)
roll, cut, bake, crisp
Cracker (1 step away)
+ raising agent, quick rise
Soda bread (1 step away)
+ butter, richer
Scone (2 steps away)
+ yeast, slow rise
Yeast loaf (you bake this)
+ extra fat, richer
Focaccia (1 step away)Next step
+ milk, egg, butter, sugar, softer, richer
Bun (1 step away)Next step
+ much more egg and butter, much richer
Brioche (2 steps away)
+ eggs on top of the milk, then a syrup soak, wetter, sweeter
Baba (3 steps away)
Your goal: richer. Each step further is a bigger jump.
- A bread you bake
- Next step toward your goal
- Another bread
- Rise
- Richer
- Finish
Steps away means changes from the nearest bread you bake. Example household, not real data. Bread steps follow the bread chapter of Lateral Cooking by Niki Segnit.
The first session has one job: find a useful next step.
Culinae starts with meals you really make, learns the conditions around them, and asks what direction matters to you. Then it offers one proposal you can accept, reject or correct.
Success is not “that looks delicious.” It is “that is new, and I can see myself making it.”
Locate me · 1 of 4
Begin with meals that are already yours
Add things you can make without much thought. Shortcuts and improvised meals count.
- Tuesday tomato pasta
- Egg on toast
- Rice with leftover vegetables
Usually 20–30 min
One pan is easiest
Weeknight energy
Screen 1 of 4
First build the path. Then help the household move along it.
Once Culinae understands current capability and direction, planning can become personal. Shopping can support the next move. Leftovers can become part of the path instead of a separate problem.
We are starting with single-living adults in Berlin, where time, space and energy are limited but the desire to cook at home is real.
We are testing the bridge before we build the system around it.
Early research in Berlin
Help us test the first crossing.
We want to know whether Culinae can locate how you cook today and suggest a next meal that feels both genuinely new and genuinely possible.
- Bring three to five meals you genuinely make
- Choose a direction you actually care about
- Judge whether the proposed next step feels both new and possible