🌱 TouchGrass AI — An AI That Tells You to Stop Using AI

This is my submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass.

What I Built

TouchGrass AI is a web app that uses local open-weight AI to generate personalized outdoor missions.

The user chooses their available time, preferred activity, energy level, mood, and company. The AI then creates a simple outdoor mission based on those preferences.

The goal is simple:

AI gives you the mission.
You put your phone away.
You go outside.

The app also includes Touch Grass Mode, a countdown timer, Grass XP, daily streaks, and mission history.

Demo

The complete flow is:

1. Choose your preferences.
2. Generate an AI mission.
3. Start the mission.
4. Put your phone away.
5. Go outside.
6. Complete the mission.
7. Earn Grass XP and maintain your streak.

Code

GitHub Repository:

https://github.com/ranjanxcode/touchgrass-ai

How I Built It

The frontend uses HTML, CSS, and JavaScript.

The backend uses Node.js and Express.

For AI generation, I used Ollama running the Llama 3.2 3B open-weight model locally.

The flow is:

User → Frontend → Express Backend → Ollama → Llama 3.2 3B → Personalized Mission

The generated mission is returned as structured JSON and displayed in the frontend.

Mission progress, XP, streaks, and history are stored using browser LocalStorage.

Why Does Open Innovation Matter?

I used an open-weight model because I wanted the AI to run locally instead of depending completely on a cloud AI API.

This gives the project:

• Local AI inference
• Better privacy
• No per-request cloud AI cost
• More control over the AI
• The ability to experiment with different open models

The open approach also makes it easier to experiment with how AI can be used in different ways.

For this project, AI is not meant to keep users on the screen.

It is meant to generate something useful and then get out of the way.

What I Learned

While building TouchGrass AI, I learned how to connect a web application with a locally running open-weight AI model.

I worked with HTML, CSS, JavaScript, Node.js, Express, Ollama, Llama 3.2 3B, JSON responses, LocalStorage, Git, and GitHub.

I also learned that AI does not always need to be something users spend a lot of time interacting with.

Sometimes the best AI experience is one that helps you leave the screen.

Future Improvements

• Weather-aware missions
• Location-aware missions
• Voice interaction
• More open-weight models
• Achievement system
• Community-created missions
• Mission sharing
• Better offline support

Final Thought

We usually build technology to keep people on screens for longer.

I wanted to try the opposite.

What if AI could help you leave the screen?

That is TouchGrass AI.

Get the mission.

Put the phone away.

Touch grass.