Here is your complete, polished submission post ready for DEV.to:
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
What I Built
I built TouchQuest — an AI-powered outdoor photography game designed to solve screen addiction with a simple philosophy:
“AI gives you the challenge. The real world gives you the answer.”
Rather than keeping players trapped in an endless loop of notifications, feeds, or chat bubbles, TouchQuest uses the screen as briefly as possible. The core loop is:
Start → Read Mission → Pocket Phone → Go Outside → Capture Discovery → Return → AI Verifies → Earn Badge → Leave Screen.
How the Game Works
The game has exactly 3 levels requiring a total of 9 photographs (4 + 3 + 2). Players do not choose the timer—the expedition sets fixed countdowns:
🟢 Level 1 — Explorer (40 mins, 4 photos, Easy): Broad awareness of immediate surroundings (e.g., natural green hues, circular geometry, marks of aging).
🟡 Level 2 — Observer (30 mins, 3 photos, Medium): Nuanced visual relationships and strict constraints (e.g., natural red that is strictly non-manmade, cast shadow silhouettes, repeating patterns).
🔴 Level 3 — Creative Explorer (15 mins, 2 photos, Hard): Compositional twists and pareidolia (e.g., everyday objects resembling faces, nature and human structures interacting). Includes a one-time emergency +3 minute extension if time runs low.
The Golden Rule: As the difficulty increases, Photos Decrease (4 → 3 → 2) while Creative Twist Increases (Easy → Medium → Hard).
Beat-the-Clock "Done" Button Players aren't forced to wait out the clock. If you capture all photos for a round ahead of time, a glowing "Done" button unlocks in the header and banner. Clicking Done banks your remaining time as a speed bonus and immediately unlocks the next round.
Accessible Everyday Environments
Difficulty never means requiring rare mountains, oceans, or 10km travel. TouchQuest is 100% playable in an ordinary backyard, college campus, city sidewalk, or neighborhood park.
Demo
Deployed Web App: https://trideep091.github.io/touch-quest/ Responsive Experience: Fully optimized for mobile phone browsers, tablets, and desktops.
Key Screens & Features
Distraction-Free Landing Hero: Complete with an outdoor aesthetic, camera viewfinder reticle corners [ + ], and a visual progression map.
Live Camera Viewfinder: Integrated with the WebRTC MediaDevices API for real-time camera streaming, shutter flash, and sound synthesis.
Interactive AI Vision Scanner: An animated green laser scanner inspecting real image properties with instant pass/fail critiques.
Collectible Embroidered Badges: 6 collectible scout patches (First Explorer, Sharp Eyes, Nature Lover, Visual Hunter, Challenge Master, Touch Grass Legend).
Field Journal & Archive: Player profile with level ranks, XP progression bar, day streaks, and an archive gallery of all past outdoor photos.
Code
The project is structured with zero heavy framework bloat into a clean, flat architecture:
Repository: [Insert Your GitHub Repo Link Here] Core Stack:
Vanilla ES Modules (JavaScript): Zero npm dependency architecture for instant execution.
Vanilla CSS Design System: Curated nature-inspired tokens (forest moss #2D6A4F, warm sandstone #FAF7F2, terracotta #D95C38, sunlit amber #E5A93C).
HTML5 Canvas Vision Engine: Offscreen pixel analysis (Excess Green Index 2G - R - B, contrast histograms, luminance gradients).
Web Audio API: Procedural sound synthesizer (camera shutter click, level complete chords, and fanfares without external audio files).
Canvas Particle Physics: Custom celebratory confetti engine for badge unlocks.
How I Built It
TouchQuest is built around a Dual Open-Weight AI Architecture:
User Privacy & Photo Sovereignty: Closed multimodal APIs require streaming every private photo of a player's backyard, street, and personal environment to centralized corporate servers. With open-weight models (like PaliGemma or Moondream2), inference can run on-device via WebGPU or on private self-hosted endpoints (via Ollama or vLLM). Your photos never become training data for closed models.
Freedom from Provider Lock-in & Pricing Shifts: Outdoor games shouldn't die because a proprietary API doubled its vision token costs or deprecated an endpoint. Open weights give developers and players permanent ownership of the experience.
Hyper-Local Fine-Tuning: Open weights allow communities to fine-tune the challenge generator on regional botany, local geography, or native wildlife without depending on a one-size-fits-all model.
My Agent Session
This application was architected and built in collaboration with Google Antigravity:
From planning the anti-screen philosophy and math of the 3-level progression curve (Photos ↓, Difficulty ↑) to designing the zero-dependency ES module architecture. Antigravity generated the embroidered scout patch emblems and outdoor artwork, wrote the canvas pixel inspection pipeline, implemented the Web Audio procedural synthesizer, and structured the anti-repetition memory engine.
Prize Categories
Grand Prize: Touch Grass (Best application getting people off their screens and into nature)
Best Use of Open-Source / Open-Weight AI (Swappable Qwen/PaliGemma architecture with local-first vision verification)
Most Creative Real-World Application