Overview

Cobble is a consumer food application that reduces the friction between deciding what to eat, finding a recipe that fits, and actually cooking it.

Instead of starting from a recipe search, someone can start from the food they already have. Ingredient recognition, recipe discovery, meal planning, saved recipes, shopping lists and dietary preferences are joined into one continuous experience rather than five disconnected features.

The problem

Food applications tend to treat recipe discovery, shopping and meal planning as three separate problems. The person using them does not experience three problems.

The real sequence is continuous: what food do I have, what can I make with it, what am I missing, when am I going to cook it, and what do I need to buy. Every break in that chain is a point where somebody gives up and orders a takeaway instead.

Cobble was designed around the whole sequence, which is a harder product problem than any single step in it.

The product

The application lets someone photograph what is in a fridge or cupboard, turns that into structured ingredient data, and uses it to surface recipes that are actually cookable right now.

From there, a recipe can move into a weekly plan, missing items can become a shopping list, and the recipe itself can be followed step by step while cooking. The catalogue behind discovery holds over 25,000 recipes.

Key capabilities

Ingredient intelligence

Capture ingredients by photograph rather than typing each one. Images are turned into structured ingredient data that the rest of the product can reason about.

Recipe discovery

Search and recommendation across a catalogue of more than 25,000 recipes, ranked against what is actually available rather than what is popular.

AI-assisted meal creation

Where no catalogue recipe fits the available ingredients, a recipe can be generated to suit them, within the dietary constraints already on the account.

Pantry management

A persistent record of what is in the kitchen, so discovery does not start from nothing on every session.

Weekly meal planning

Recipes move into a planner across the week, which turns discovery into a decision that has already been made by the time somebody is hungry.

Shopping lists

Missing ingredients from planned meals become a consolidated shopping list, rather than something to reconstruct by hand from several recipes.

Guided cooking

Step-by-step instructions designed for use with wet hands in a kitchen, not for reading like an article.

Nutrition and personalisation

Dietary requirements, food preferences and nutrition information are applied across discovery, generation and planning rather than sitting in a settings screen.

Engineering

The backend is built on Supabase, providing the database, authentication and storage behind user accounts, saved recipes, pantry state and meal plans.

Computer vision sits behind one step of the workflow — turning an image into structured ingredients — and its output feeds a deterministic recommendation and planning system. The model is a component inside a larger product, not the product.

Production concerns took a substantial share of the work: account state, offline and poor-connectivity behaviour, deep links into shared recipes, subscription entitlement, application lifecycle, notification permissions, platform policy compliance and error handling.

The application runs across iOS and Android from shared application logic, with the platform integrations each store requires.

Product experience

The central design decision is that people should be able to show the application their ingredients rather than describe them. Typing fourteen items into a form is the reason ingredient-first cooking apps historically go unused.

AI is deliberately not the headline. A model that suggests a recipe nobody can cook is worse than a search box, so generated output is constrained by what is actually in the pantry and by the dietary rules already on the account.

Outcome

  • Cobble is live and publicly available at cobble.food, on iOS and Android.
  • Pantry, discovery, recipe, planner and shopping operate as one connected system rather than requiring someone to manage each stage in a different application.

Our role

  • Product strategy
  • UX and interface design
  • Mobile application development for iOS and Android
  • Backend architecture and data modelling
  • Computer vision integration
  • Subscription and payment integration
  • Production delivery and store release

Have a comparable requirement?

Send the brief or the procurement reference. We will tell you whether this is work we can do well.