Inspiration
As three college students in our junior year, we know how difficult it can be to balance school, time, and money. We don't always have the time to cook elaborate meals, but we also don't always have the budget to order takeout or make another grocery trip in the middle of the week.
That inspired us to create a simple tool that answers a common question: "What can I make with what I already have?"
Instead of planning a meal and then buying all of the ingredients for it, our project starts with what is already in your pantry and works from there. We wanted to make it easier to spend less money, reduce unnecessary grocery trips, and spend less time figuring out what to eat.
What We Built
Our demo is a locally hosted website where users can enter any ingredients they currently have available. The AI then generates recipe suggestions based around those ingredients.
To make the suggestions more practical, we added quality-of-life features including:
- Ingredient substitution suggestions for ingredients the user doesn't have
- Ingredient progress bars showing how much of each recipe the user can already make
- Dietary filters to help users find recipes that fit their needs
- Multiple recipe suggestions so users can choose what sounds best
Our goal was to make the experience feel less like searching through a recipe database and more like having a personal chef work with whatever happens to be in your kitchen.
How We Built It
We used AI throughout the development process to rapidly turn our initial idea into a functional web application. We made use of Google Antigravity to establish a framework for the website, develop functionality, and help identify and fix bugs as we iterated.
This allowed us to spend more of our limited hackathon time thinking about the user experience and the actual problem we wanted to solve rather than getting stuck on setup and implementation details.
What We Learned
One of our biggest takeaways was seeing firsthand how powerful AI can be as a development tool.
Starting with a barebones project and an idea, we were able to use AI to quickly turn that idea into a functional, cleanly formatted website within a very short amount of time. The difference between our initial concept and having a working application was surprisingly large, and it showed us how much faster prototyping can become when AI is used as part of the development process.
We also learned that generating an answer is not necessarily the same as generating a good answer. Getting the AI to produce useful and varied recipes required us to think carefully about the constraints we were giving it and how we could guide its responses.
Challenges We Faced
One of our biggest challenges was getting enough variety in the suggested meals. We noticed that when given certain combinations of ingredients, the AI would repeatedly fall back on similar types of dishes, particularly salads and broths.
Figuring out why this was happening and how to encourage more creative suggestions required some brainstorming and iteration as a team. We experimented with how we structured our requests and what constraints we gave the AI so that it would consider a wider range of possible meals.
This became an important lesson for us: good AI applications aren't just about getting an AI model to generate something—they're about designing the right constraints and user experience around the model.
Log in or sign up for Devpost to join the conversation.