TouchGrass AI: An Open-Source AI That Turns Screen Time into Green Time

Devansh Yadav — Sun, 11 Oct 2026 10:14:31 +0000

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

🌿 TouchGrass AI — Turn Screen Time into Green Time

What I Built

TouchGrass AI is an AI-powered web application that encourages people to take a break from their screens, step outside, and reconnect with the real world.

We spend so much time scrolling, gaming, studying, and working on our devices that we sometimes forget how refreshing a little time outdoors can be. I wanted to build something that uses AI not to keep people online longer, but to help them spend more time offline.

TouchGrass AI turns the simple idea of “going outside” into a fun, personalized experience. Instead of giving users generic advice like “take a walk,” it generates outdoor quests based on their mood, preferred activity, and available time.

🌱 Key Features

AI-Powered Outdoor Quests: Generates personalized outdoor activities using an open-weight language model.
Personalized Experiences: Users can select their mood, preferred activity, and available time to get a suitable quest.
Quest Checklist: Breaks activities into manageable steps that users can complete outdoors.
Built-in Timer: Helps users dedicate time to their outdoor adventure.
XP and Progress Tracking: Rewards completed activities with experience points and tracks progress.
Streaks and Journal: Encourages consistency and lets users record their outdoor experiences.
Fallback Quest Generator: Provides predefined activities when the AI model is unavailable.
Privacy-Friendly Design: Progress and journal entries are stored locally in the browser using LocalStorage.

The project is designed for students, developers, gamers, remote workers, and anyone who wants to build healthier digital habits and spend more time in the real world.

The main idea: Use AI to help people spend less time using AI. 🌳

Demo

🚀 Live Website: https://devanshy0820.github.io/TouchGrass-AI/

Code

💻 GitHub Repository: https://github.com/devanshy0820/TouchGrass-AI

The project is built with HTML, CSS, and JavaScript, making it accessible to developers who want to explore, learn from, or contribute to the codebase.

To run it locally:

git clone https://github.com/devanshy0820/TouchGrass-AI.git
cd TouchGrass-AI
python3 -m http.server 8000

Then open http://localhost:8000 in your browser.

Contributions, suggestions, bug reports, and new outdoor quest ideas are welcome!

How I Built It

I built TouchGrass AI using standard web technologies and an open-weight language model that can run directly in the browser.

Tech Stack

HTML5 — Structures the application and its interactive components.
CSS3 — Creates the responsive, nature-inspired user interface.
JavaScript — Handles quest generation, user interactions, timers, checklists, and progress tracking.
Transformers.js — Enables running compatible machine-learning models directly in the browser.
SmolLM2-360M-Instruct — The open-weight language model used for generating personalized outdoor quest ideas.
LocalStorage — Stores progress and journal entries on the user's device.

How the AI Works

The application uses the onnx-community/SmolLM2-360M-Instruct model through Transformers.js.

The general flow is:

The user selects a mood, an activity preference, and an available time.
JavaScript combines these preferences into a prompt for the language model.
The model generates a suitable outdoor quest.
The application presents the quest and its suggested steps in an interactive interface.
Users can complete the steps, use the timer, and track their progress.
If model loading or generation fails, the app can fall back to predefined quests.

The model and its required files may need to download the first time the application runs. Browser compatibility, memory requirements, and network access can affect the experience.

I wanted to keep the architecture lightweight and approachable rather than relying on a large backend or making every quest-generation request to a hosted AI API.

Why Does Open Innovation Matter?

Open innovation made it possible for me to experiment with AI in a project focused on a real-world problem without making a paid, closed AI API the foundation of the application.

Using an open-weight model and browser-based inference gives developers more flexibility to experiment with prompts, customize experiences, study model behavior, and build applications without sending every generation request to a third-party inference service.

It also makes the project more accessible to students and independent developers who want to learn how AI-powered applications work.

For TouchGrass AI, this approach supports three important goals:

Accessibility: Developers can explore the implementation and adapt it to their needs.
Privacy-conscious design: Quest generation can happen locally after the model loads, reducing the need to send user preferences to an external inference API.
Experimentation: The model, prompts, and quest-generation logic can be modified as the project evolves.

Open innovation is about more than using an existing model. It is about making technology easier to explore, improve, and build upon together.

Note: The model and its dependencies remain subject to their respective licenses. Initial downloads still require network access, and browser-based inference does not automatically mean the entire application works offline.

My Agent Session

I don't have a public DevRelay agent-session link to share for this submission. If I publish one, I'll add it here.

Prize Categories

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

I'll add any applicable partner-specific prize categories according to the challenge's official rules.

Final Thoughts

TouchGrass AI started with a simple question:

What if we used AI to encourage people to close their laptops, step outside, and experience the world around them?

This project is my attempt to turn that idea into something practical, interactive, and fun.

I'm excited to keep improving it, explore more open-weight models, and add features that make spending time outdoors a rewarding habit.

If you like the idea, check out the live demo and GitHub repository. Feedback and contributions are welcome!

🌿 Build with AI. Step outside. Touch grass.

DEV Community: Devansh Yadav