This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
What I Built
🌿 TouchGrass AI — Less scrolling. More exploring.
We spend a significant amount of time looking at screens, often struggling to decide how to take a meaningful break. I wanted to build something that uses AI not to keep people glued to their screens, but to encourage them to step away from them.
TouchGrass AI is an AI-powered outdoor adventure companion that generates personalized outdoor quests based on four inputs:
- Activity: Walking, running, nature observation, or gardening.
- Available time: 10, 20, or 30 minutes.
- Energy level: Relaxed, moderate, or energetic.
- Environment: Park, garden, neighbourhood, or trail.
Based on these choices, the app generates a short outdoor mission with achievable tasks and a reflection question.
The idea is simple: use AI for a moment of inspiration, then put the phone away and experience the real world.
Demo
🌐 Try TouchGrass AI:
https://touchgrass-ai-mgxtxn7ybxnffaenhsuyyd.streamlit.app/
Code
💻 GitHub repository:
https://github.com/Bodhi14/touchgrass-ai
The project is built with Python and Streamlit, with a hosted inference API powering quest generation.
How I Built It
I built TouchGrass AI using:
- Python for the application logic.
- Streamlit for the interactive web interface.
- Open-weight language models for generating personalized outdoor quests.
- Hosted inference APIs to make model inference possible without requiring every user to run a large model locally.
- Streamlit Community Cloud to deploy the application.
The app collects the user's preferences, turns them into a structured prompt, sends that prompt to the hosted model, and displays the generated quest.
One of the interesting parts of this project was learning the difference between hosting an application and hosting an AI model. Deploying the interface was only one part of the problem; the model also needed an inference endpoint and appropriate API authentication.
Why Does Open Innovation Matter?
Open innovation makes AI experimentation more accessible to developers who may not have the hardware or resources to train and run large models themselves.
Open-weight models offer opportunities to inspect model choices, experiment with different inference providers, and build applications around models whose weights are available under their respective licences.
For a project like TouchGrass AI, this flexibility is valuable. I can experiment with different models and hosting arrangements while keeping the user experience focused on the same goal: encouraging people to spend less time scrolling and more time outdoors.
I also learned that using an open-weight model is not the same as running it entirely locally or having unlimited free inference. Hosting, licensing, privacy, and usage limits all matter when building a real application.
My Agent Session
This project was an opportunity to explore how AI can be used for something deliberately different from the usual productivity or screen-time-heavy applications.
Rather than generating more content for users to consume, TouchGrass AI aims to give them a reason to put their devices down.
What's Next?
Some directions I'd like to explore include:
- Location-aware quests based on nearby parks and outdoor spaces.
- Weather-aware mission suggestions.
- Quest difficulty that adapts to user preferences.
- A stronger local-inference option for users who prefer running models on their own devices.
The long-term vision is to make AI a bridge between the digital world and the physical one.
Build less screen time. Experience more real life. 🌱
Top comments (0)