TouchLess AI 🌱 — AI That Wants You to Stop Using It

A developer built TouchLess AI, an open-source outdoor mission generator that uses Google's Gemma 3 4B model via Ollama to turn screen time into real-world activity. Users pick their available time, environment, goal and difficulty, and the model produces a personalized offline mission; a deliberately minimal "Phone-Away Mode" then encourages them to put the phone down, with a second Gemma call generating a grounding response to a post-activity reflection. The stack is Next.js, React, TypeScript and Tailwind on the frontend with Python, FastAPI and Pydantic on the backend, storing everything in browser localStorage with no authentication, database or user tracking.

🌱 What is TouchLess AI? TouchLess AI is an AI-powered outdoor mission generator that uses Google Gemma 3 to turn screen time into real-world activity. Instead of giving you another reason to stay on your screen, it generates a personalized mission that encourages you to leave it. github- https://github.com/ashab683/touchless-AI https://github.com/ashab683/touchless-AI The user chooses: ⏱️ How much time they have 🌳 Their environment 🎯 Their goal ⚡ Difficulty Gemma then generates a personalized outdoor mission. And then comes the most important part: Put the phone away. 💡 The Problem A lot of technology is designed around keeping us engaged for longer. More notifications. More scrolling. More screen time. I wanted to explore the opposite idea: Can AI be useful while encouraging us to spend less time using technology? That became TouchLess AI. 🔄 How It Works The experience is designed as a simple loop: Choose your preferences ↓ Gemma generates an outdoor mission ↓ Start Phone-Away Mode ↓ Put your phone away ↓ Go outside ↓ Complete the mission ↓ Write a short reflection ↓ Gemma generates a grounding response ↓ Save the experience locally The philosophy is simple: The better TouchLess AI works, the sooner you should stop using it. 🧠 Gemma at the Core Gemma is at the core of the TouchLess AI experience. Before going outside Gemma receives the user's preferences and generates a structured outdoor mission. Instead of simply saying: "Go for a walk." the application can create a more specific activity based on the user's available time, environment, goal and difficulty. 📸 Insert your actual mission screen showing the Gemma-generated mission here. This is where the AI turns the user's preferences into an actionable offline activity. After coming back The user writes a short reflection about what they noticed or experienced. Gemma then generates a brief grounding response based on that reflection. So AI is used at two important moments: Before: Help me decide what to do. After: Help me reflect on what I experienced. The middle part is intentionally human + offline. 📵 Phone-Away Mode This is one of the most important parts of TouchLess AI. Once the mission starts, the interface becomes intentionally minimal. IMAGE 3 — PHONE-AWAY MODE It includes: ⏱️ A timer 🌿 Minimal visual feedback 📵 A reminder to put the phone away 👀 An optional way to briefly check the mission The feature is designed around one principle: The application should become less interesting once the mission begins. The goal isn't to keep you inside the app. The goal is to get you out of it. 🌳 The Real-World Activity This is where TouchLess AI becomes different from a typical AI application. The AI generates the mission. Then the user leaves the screen behind and actually completes it. 📸 Insert your outdoor photo/video screenshot here. This could be a photo of you completing the mission, a park/nature scene from the activity, or a frame from your demo video. The important part is showing that the AI-generated task leads to something outside the screen. ✍️ Reflection After completing the activity, the user comes back and records a short reflection. For example: "I noticed how quiet the park was once I stopped checking my phone." The reflection becomes the input for the second AI interaction. Gemma then responds with a short grounding reflection rather than trying to keep the user engaged in another conversation. 🛠️ How I Built It Frontend Next.js React TypeScript Tailwind CSS Backend Python FastAPI Pydantic AI Google Gemma 3 4B Ollama for local inference Storage Browser localStorage The current application has: No authentication No database No user tracking No unnecessary data collection The project is designed to keep the experience simple and local-first. 🏗️ Architecture The current local architecture looks like this: ┌──────────────────────┐ │ Next.js + React │ │ Frontend │ └──────────┬───────────┘ │ ▼ ┌──────────────────────┐ │ FastAPI │ │ Backend │ └──────────┬───────────┘ │ ▼ ┌──────────────────────┐ │ Ollama │ │ Local Inference │ └──────────┬───────────┘ │ ▼ ┌──────────────────────┐ │ Gemma 3 4B │ └──────────────────────┘ I used structured responses and validation so that the frontend receives predictable mission and reflection data from the AI service. 🔓 Why Open Innovation? The interesting question for me wasn't just: "What can AI generate?" It was: "What should we use AI for?" Open-weight models make it possible to experiment with AI in different directions and build applications around ideas that don't necessarily fit the usual engagement-driven model. TouchLess AI is my small experiment in that direction. Instead of using AI to keep someone on a screen longer, I'm using it to help them leave the screen. 🎥 Demo The demo follows the complete experience: Configure ↓ Generate with Gemma ↓ Phone-Away Mode ↓ Go Outside ↓ Complete Mission ↓ Reflect ↓ Gemma Response ↓ History VIDEO / GIF — OPTIONAL 📹 Insert your 60–90 second demo video or GIF here if DEV supports your chosen format. The demo should show the actual application rather than only the source code. 💻 Source Code The complete project is open source. 🔗 GitHub: https://github.com/ashab683/touchless-AI https://github.com/ashab683/touchless-AI The repository contains the frontend, FastAPI backend, AI service layer, prompts, configuration and documentation. 📚 What I Learned Building TouchLess AI helped me learn more about: Integrating an open-weight AI model into a real application Building a FastAPI backend Connecting a Next.js frontend to an AI service Working with structured AI responses Validating AI-generated data Designing a local-first experience Building an interface where less interaction is actually better Thinking about AI from a human-behavior perspective The biggest lesson was probably this: Good AI doesn't always need to give you more to do. Sometimes it should help you do less on your screen. 🚀 What's Next? Some improvements I'd like to explore: 🌳 More diverse outdoor missions 🧠 Better personalization ♿ More accessibility improvements 💾 More local-first functionality 🤝 Community-contributed missions 🤖 Additional open-model inference options 📱 Better mobile/PWA experience 📊 More ways to measure meaningful offline activity 🌍 Final Thought We already have plenty of technology competing for our attention. Maybe we also need technology that knows when to let us go. TouchLess AI is an experiment in exactly that. 🌱 If you had 30 minutes away from your phone, what outdoor mission would you want an AI to give you?