This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend
🍽️ Just Pick One.
An AI restaurant decision-maker for people who say “anything is fine” and then reject every option.
There is a very specific kind of group-chat problem:
“Where should we eat?”
Someone says “anything.”
You suggest pizza.
“I don't feel like pizza.”
You suggest burgers.
“Something healthier?”
You suggest a restaurant.
“Too expensive.”
At this point, the group has spent 25 minutes deciding where to eat and somehow nobody is eating. 😭
So I built Just Pick One for a friend who regularly gets stuck in exactly this situation.
What I Built
Just Pick One is a restaurant recommendation tool that takes the things people actually care about when choosing a place:
- 💰 Budget
- 🥗 Vegetarian-friendly options
- 🍜 Cuisine / food preference
- 📍 Distance / location
- 😋 What you're actually craving
Instead of dumping a giant list of restaurants on the user, the goal is to help narrow the decision down to options that actually fit.
Less scrolling. Less arguing. More eating.
🎥 Demo
Video Demo: https://drive.google.com/file/d/1CtKR_HJyUgUFbJMqcJNFNX-i5ir-CKag/view?usp=sharing
The video shows the complete flow from entering preferences to getting restaurant recommendations.
How I Built It
The project uses open-source/open-weight AI as the decision-making layer, combined with openly available restaurant/location data.
The basic flow is:
User preferences → Restaurant candidates → AI reasoning → Recommendations
The AI isn't just there to generate a fancy paragraph about food.
It helps reason about the user's constraints and rank the available options based on what they actually asked for.
I also kept the project intentionally lightweight rather than building a giant stack of APIs, databases and authentication just to answer the question:
“Bro, where are we eating?”
🤖 Why Open Innovation?
I wanted the AI component to be based on an open-weight model rather than making the entire project dependent on one closed AI provider.
That gives the project more flexibility:
- Models can be swapped.
- The system can potentially be run locally.
- Open models can be adapted or fine-tuned.
- There is more control over how the AI layer is used.
- The project isn't locked into a single proprietary model.
For a recommendation tool like this, that flexibility matters because the AI is being used as a reasoning component rather than simply as a chatbot.
🧑🤝🧑 Built for a Friend
The whole idea came from a very real problem.
My friend doesn't need 100 restaurant recommendations.
They need someone to finally say:
“We're going here. Stop searching.”
So that's what I tried to build.
A tiny AI-powered decision-maker for one of the most difficult problems known to humanity:
getting a group of friends to agree on food. 🍕
And honestly, if this saves even one group chat from 47 messages of “you decide”, I'll consider it a success.
What I Learned
The biggest lesson from this build was that a useful AI project doesn't necessarily need to be huge.
The interesting part was figuring out how to use an open-weight model for a specific decision-making problem, instead of just making another general-purpose chatbot.
Sometimes the best product idea is hiding inside the most annoying sentence in your group chat:
“Where do you guys wanna eat?”
Top comments (0)