TouchGrass: Leave the screen behind

Vijay Kumar — Mon, 12 Oct 2026 06:31:39 +0000

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

What I Built

TouchGrass is an anti-distraction, screen-minimizing outdoor companion powered by open-weight AI.

Most mobile and web apps are designed to trap user attention with infinite feeds, algorithmic notifications, and gamified loops. TouchGrass does the opposite: the screen is engineered to be the shortest part of the entire experience.

What TouchGrass Does:

Synthesizes Ambient Micro-Quests: Powered locally by Google's Gemma 2 (2B) via Ollama, it generates a personalized outdoor walk tailored to your available time (e.g., 10m, 20m, 45m, or a custom duration) and mood. Each quest contains 3 gentle, sensory-focused invitations (e.g., "Spot small moss or weeds growing in pavement cracks", "Pause for 30 seconds to listen to wind or birds", "Examine the bark pattern of two mature trees").
Pocket Mode & Screen Sleep: A dedicated full-screen dimmed black screen mode allows you to put your phone in your pocket with zero distractions and low battery usage.
Paper Slip Printing (A6): For a truly unplugged experience, users can click "Print Paper Card" to generate a minimalist paper slip, leave their phone at home on their desk, and head outside.
Offline Procedural Chimes: Built with native browser Web Audio API synthesizers—delivering serene singing bowl and bell harmonics upon starting, completing steps, or finishing the walk—without needing external audio assets.
Private Field Journal: When you return from outside, you can record your observations, mood shift (Before vs. After), and physical discoveries in a private, local IndexedDB notebook.

Who It Is For:

Developers, designers, and students suffering from screen fatigue and doomscrolling burnout.
Remote workers looking for a mindful, restorative break between meetings or sprint tasks.
Anyone who wants an encouraging, gentle reason to step outside and reconnect with the physical world.

Demo

Repository: https://github.com/Vijaykr35/touchgrass
Live Local Demo Experience: The app runs locally with Ollama + Next.js to ensure an air-gapped, zero-telemetry experience.

Key Screenshots & Workflow:

Step 1: Calibrate Your Horizon
Select your available time window (10 min quick break, 20 min stroll, 45 min deep walk, or fine-tune custom minutes).
Step 2: Departure Protocol & Pocket Mode
Receive your three sensory invitations. Choose to print an A6 paper slip or activate the blackout Screen Sleep Mode to tuck your phone away into your pocket.
Step 3: Return & Field Journal
Log what you noticed outdoors into your private naturalist journal, track your calming mood transition, and reflect on the world beyond the screen.

Code

The complete source code is open source and hosted on GitHub:

Vijaykr35 / touchgrass

TouchGrass 🌱

Leave the screen behind. Let curiosity guide your steps.
Built for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass.

🌿 Overview

TouchGrass is a screen-minimizing outdoor companion that uses open-weight local AI (Ollama with Gemma 2 2B / Qwen2.5 3B / Llama 3.2 1B) to generate personalized, ambient outdoor micro-quests.

In an age of endless feeds, screen fatigue, and algorithmic doomscrolling, TouchGrass inverts the paradigm: the screen is designed to be the shortest part of the experience.

Once your departure card is generated:

📖 Memorize 3 gentle sensory prompts (e.g., “Find moss growing in pavement cracks”, “Listen for bird calls or wind for 30 seconds”).
🖨️ Print an A6 pocket slip or tap Screen Sleep / Black Screen Mode.
🚶 Step outside with zero tracking, zero mandatory GPS, and 100% data privacy.
📝 Return and log your observations in an air-gapped local…

View on GitHub

Repository link: https://github.com/Vijaykr35/touchgrass

How I Built It

The Open-Source AI Architecture:

TouchGrass is architected around open-weight, local inference using Ollama and Google Gemma 2 2B (gemma2:2b), with automatic runtime fallback support for qwen2.5:3b and llama3.2:1b.

Local Model Execution:
- The application interacts with an on-device Ollama instance over http://localhost:11434.
- We specifically selected Gemma 2 2B because its parameter efficiency, fast time-to-first-token on consumer CPU/GPU hardware, and adherence to structured system instructions make it ideal for generating creative, evocative outdoor observation prompts without sluggish latency.
Structured JSON Output & Schema Recovery:
- The Next.js API route (/api/quests/generate) leverages Ollama's native JSON mode with strict token bounds (num_predict: 260, num_ctx: 512, temperature: 0.1) to ensure ultra-fast generation times (< 5-10 seconds on consumer hardware).
- A resilient JSON parser handles bracket recovery and sanitization, validating the resulting payload against a strict Zod schema (QuestSchema) before serving it to the client.
Front-End & Ambient Experience:
- Framework: Next.js 16 (App Router) with React 19 and TypeScript.
- Styling: Tailwind CSS with a warm, grounding naturalist color palette (#244B36 deep pine, #F5F2E8 warm paper, #D9E66F moss accent).
- Web Audio API: Real-time sine-wave frequency synthesis simulating singing bowls (392Hz, 587Hz, 783Hz) and meditation chimes directly in the browser.
- Client Storage: Browser IndexedDB through the idb library to keep all quest history, reflections, and specimen notes private on the user's machine.

Why Does Open Innovation Matter?

Open-source and open-weight innovation was essential for building TouchGrass in ways closed APIs could never match:

True Privacy & Zero Telemetry for Wellness:
- When building a tool to unplug from the modern digital ecosystem, requiring users to send personal reflections, walk times, and location habits to a centralized proprietary cloud API defeats the entire ethos of digital mindfulness.
- Open-weight models like Gemma 2 allow TouchGrass to run completely offline on an airplane, in a remote cabin, or in an urban park without internet access.
Cost-Free Longevity & Community Sovereignty:
- Closed APIs incur per-token billing, rate limits, latency spikes, and eventual deprecation when providers sunset models.
- Open weights mean this tool belongs to the user permanently. Anyone can clone the repository, run ollama run gemma2:2b, and have an outdoor companion that functions forever with zero subscription fees.
Democratizing Edge AI:
- Open innovation empowers developers to run capable, expressive language models on modest laptops and local machines, proving that intelligent, personalized software doesn't need billion-dollar cloud clusters to make a positive impact on human life.

Prize Categories

Best Use of Gemma

TouchGrass defaults to and is optimized for Google's Gemma 2 2B (gemma2:2b). Gemma's balance of lightweight resource footprint and linguistic nuance makes it the ideal companion for generating contemplative, localized sensory walks that encourage people to close their laptops, put their phones in their pockets, and reconnect with nature.

Idea Board — Turn 3 AM Idea Overload into Actionable 30-Minute Tasks with Local AI

Vijay Kumar — Sun, 04 Oct 2026 18:59:14 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

I built Idea Board for my best friend who suffers from severe "3 AM idea overload." He constantly wakes up with ambitious side-project ideas, writes down vague goals like "Learn AI vectors" or "Build a search engine," but gets completely overwhelmed trying to figure out where to start. Within a few days, those ideas die in his notes app clutter.

Idea Board solves this by converting any vague goal or project idea into a structured, step-by-step Kanban task board in seconds. Each generated task is concrete, starts with an action verb, and is strictly capped at 15 to 60 minutes so he can make real progress in bite-sized sessions without feeling overwhelmed.

Demo

Video Demo

https://github.com/Vijaykr35/idea-board/blob/main/ideaBoard.mp4

Note: Video demo is also stored in the codebase at frontend/public/ideaBoard.mp4.

Screenshots

Dark Notebook Kanban Board View

Generated Task Breakdown

Code

github Vijaykr35/idea-board

Repository: https://github.com/Vijaykr35/idea-board

How I Built It

Idea Board is built with local open-source AI at its core:

Local AI Inference Engine: Powered by Ollama running the open-weight qwen2.5:3b model locally at http://localhost:11434.
Backend: Built with FastAPI (Python 3.10+), SQLAlchemy, SQLite, and httpx for async local REST calls to Ollama.
Frontend: A custom React (Vite) application styled with Tailwind CSS using a dark notebook aesthetic with hand-drawn Patrick Hand SC typography, interactive HTML5 drag-and-drop Kanban columns (To do, Doing, Done), progress bar tracking, and batch text import.

Why Does Open Innovation Matter?

100% Privacy & Zero Server Leakage: My friend doesn't want his raw startup ideas or personal notes sent to third-party cloud LLM APIs or used to train corporate models. Running qwen2.5:3b locally via Ollama keeps all data strictly on his machine.
Runs Offline on a Laptop: It requires zero internet connection. He can work on airplane rides or in cafes without relying on cloud availability or API latency.
Zero Subscription Fees: Closed API subscriptions add up fast. Open-weight models running locally cost $0 per prompt.
Full System Control: Open innovation allowed me to fine-tune the system prompt and strict formatting rules without API guardrail interference or rate limits.

My Friend's Reaction

"Man, this is literally what I needed. I plugged in 'Learn vectors for AI' with 1 hour/day, and having 30-minute micro-tasks right in front of me on a dark notebook board made starting feel effortless instead of intimidating."

My Agent Session

I used an agentic pair-programmer during the hackathon to build, test, and style this full-stack application.

Prize Categories

Best Project Built for a Friend
Most Impactful Open Source Learning Tool

DEV Community: Vijay Kumar

TouchGrass: Leave the screen behind

What I Built

What TouchGrass Does:

Who It Is For:

Demo

Key Screenshots & Workflow:

Code

Vijaykr35 / touchgrass

TouchGrass 🌱

🌿 Overview

How I Built It

The Open-Source AI Architecture:

Why Does Open Innovation Matter?

Prize Categories

Best Use of Gemma

Idea Board — Turn 3 AM Idea Overload into Actionable 30-Minute Tasks with Local AI

What I Built

Demo

Video Demo

Screenshots

Dark Notebook Kanban Board View

Generated Task Breakdown

Code

How I Built It

Why Does Open Innovation Matter?

My Friend's Reaction

My Agent Session

Prize Categories