My Project

Vaibhav Bodakhe — Fri, 09 Oct 2026 10:18:03 +0000

🌿 TouchGrass AI — Your AI-Powered Offline Adventure Companion

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

What I Built

Meet TouchGrass AI — an open-source AI agent designed to help people spend less time scrolling and more time exploring the real world.

We all know the feeling: you open your phone to check one notification, and suddenly an hour disappears. TouchGrass AI turns that endless scrolling habit into an opportunity for real-world adventure.

Instead of recommending another video or app, it suggests activities that get you moving, exploring, and connecting with people around you.

🌱 Key Features

AI Adventure Generator: Get personalized outdoor challenges based on your mood, available time, and interests.
Digital Detox Quests: Complete challenges like walking for 20 minutes, visiting a nearby park, or watching a sunset without your phone.
Mood-Based Activities: Feeling stressed? Get a nature walk. Feeling bored? Discover a photography challenge.
Eco Explorer: Receive ideas for litter-picking walks, community gardening, and other environmentally friendly activities.
Offline Achievement System: Earn virtual badges for completing real-world activities instead of spending hours online.
Privacy-First Design: Keep personal activity data on your device whenever possible.

The idea is simple: AI should help us experience more of the real world, not trap us inside a screen.

My target audience is students, remote workers, digital wellness enthusiasts, and anyone who wants to build healthier technology habits.

Demo

🌍 Imagine this scenario:

You open TouchGrass AI and enter:

"I'm bored, have 30 minutes free, and want to feel refreshed."

Your AI agent responds:

Your mission: The 30-Minute Reset

Leave your phone in your pocket.
Walk to a nearby green space.
Find three interesting things in nature.
Take five minutes to sit quietly and observe your surroundings.
Return and reflect on how you feel.

Reward: Nature Explorer badge 🌳

Demo status: Concept demonstration. A deployed demo or recorded walkthrough can be added here when available.

Code

The project is envisioned as an open-source application with a modular architecture.

Planned repository: touchgrass-ai

The codebase would contain:

frontend/ — React and Vite user interface.
backend/ — Node.js and Express API.
ai/ — Open-weight model integration and activity-generation prompts.
models/ — Activity schemas and recommendation logic.
docs/ — Setup instructions and contribution guidelines.

Repository: Add your GitHub URL here once the project is published.

How I Built It

The imagined implementation uses a lightweight full-stack architecture with an open-source AI model at its core.

Technology stack:

Frontend: React, Vite, and Bootstrap.
Backend: Node.js and Express.
Database: MongoDB for optional activity history and achievements.
AI: An open-weight instruction-following model, such as Qwen or Mistral, accessed through a compatible local inference runtime.
Privacy: Local-first storage for preferences and progress wherever practical.

How the AI Agent Works

The user describes their mood, available time, and interests.
The backend passes those preferences to the AI model.
The model generates a practical, time-limited offline activity.
The application checks the response against a structured activity format and basic safety rules.
The user completes the activity in the real world and optionally records the result.

For example, a request for a 15-minute activity should produce a short, achievable mission rather than a complicated day-long itinerary.

The AI does not need access to a user's precise location. Location-based recommendations could be added as an optional feature with explicit permission.

Why Does Open Innovation Matter?

Open innovation makes TouchGrass AI more than a single application.

A closed API could generate activity suggestions, but an open ecosystem offers additional possibilities:

Model freedom: Developers can experiment with different open-weight models and choose one that fits their hardware and requirements.
Privacy and control: Local inference can reduce the need to send personal preferences to external services.
Community-driven challenges: Contributors can add activities tailored to different cultures, climates, accessibility needs, and age groups.
Transparent development: The community can inspect the recommendation logic and help improve safety and usefulness.
Accessible experimentation: Students and independent developers can build and test ideas without making a proprietary AI service the foundation of the project.

The larger goal is to make AI a tool that encourages healthier habits, environmental awareness, and real human connection.

Open AI should empower people to live better offline, not just interact more online.

My Agent Session

The project could use DevRelay to capture an agent-assisted development session, showing how the agent helps create the activity-generation logic, build the interface, and test the application.

Agent session: Add a real DevRelay session link here when available.

Prize Categories

Potential categories, subject to the challenge's official eligibility rules:

Open-Source AI
AI Agents
Developer Tools and Community Innovation

Final Thoughts

TouchGrass AI explores a question worth asking:

What if the best thing an AI assistant could do was encourage you to put your phone down?

Technology doesn't have to compete with the real world. It can help us rediscover it.

🌿 Less scrolling. More exploring. More living.

Hacktoberfest #OpenSource #AI #AIChallenge #TouchGrass #DigitalWellness

DEV Community: Vaibhav Bodakhe