This is a submission for the
Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
What I Built
We keep telling people to go outside, but many outdoor apps still keep them staring at a screen.
A map app asks for constant checking. A fitness app asks for another dashboard. A nature app asks users to frame, upload, wait, read, and repeat.
The phone becomes the destination instead of the doorway.
That observation led me to build Nature Go: a screen free outdoor exploration app that turns real world nature into small, verifiable quests.
Nature Go gives an explorer a mission such as:
- Find a green leaf.
- Photograph tree bark.
- Spot a pinecone.
- Find moving water.
- Discover a wildflower.
- Look for moss, dewdrops, a feather, or a spiderweb.
The explorer goes outside, captures the discovery, and receives an AI verification result. The app checks whether the requested target is present, whether it is physically real, and whether the image shows genuine outdoor context.
If the discovery is verified, Nature Go saves it to the explorer’s journal, updates the daily streak and badges, and reads an educational nature fact aloud. If the image is a phone screen, printed image, indoor plant, artificial plant, or unrelated target, the app rejects it and explains why.
The screen is only needed for the shortest part of the experience: choosing a quest and confirming the result. The real activity happens outside.
The application includes:
- 20 nature quests
- 8 expedition challenges
- 5 achievement badges
- Custom quest creation
- Discovery journal with photos and confidence scores
- Daily streak and longest-streak tracking
- Explorer profiles
- Screen-free audio mode
- Camera and photo-upload flows
- Demo scenarios for real discoveries and spoofed images
- Admin tools for explorer profiles, official quests, and verification controls
Demo
Live application:
https://nature-go-outdoor-vision-quest.ai.studio/
Try this flow:
Open the live app.
Keep the default Green Leaf quest.
Select Real Nature — Green Leaf under the demo photos.
Observe the successful verification result.
Try Phone / Tablet Screen and compare the rejection.
Open Quests to browse the 20 nature missions.
Select Audio Walk to experience the screen-free mode.
The project workflow is:
Code
*GitHub repository: *
https://github.com/RabiaA-arif/NatureGo
The repository contains the complete React, TypeScript, Express, Firebase, quest, verification, audio, and admin-console implementation.
Local setup:
Bash
git clone https://github.com/RabiaA-arif/NatureGo.git
cd NatureGo
npm install
cp .env.example .env
npm run dev
Configure the required environment variables before testing the full AI verification and Firebase persistence flows.
How I Built It
Nature Go is built around an open-source AI verification layer rather than using AI as a decorative chatbot feature.
Gemma integration:
Model: [add the exact Gemma model name used in the submitted build]
Runtime/provider: [add the exact runtime or provider]
Inference mode: [local, Google Cloud, or another verified provider]
Source code: [add the exact public GitHub link to the Gemma implementation]
The model receives the current quest target and the captured image. It returns structured verification data:
The application uses that response to decide whether to:
- Play a success or rejection sound
- Narrate the nature fact
- Save the discovery to the journal
- Update quest completion progress
- Update streaks and authenticity score
- Unlock badges
- Guide the explorer toward another outdoor quest
The verification policy checks three conditions:
1.Target match: The requested object or scene must be visible.
2.Physical authenticity: Screens, printed images, wallpapers, fake plants, and indoor houseplants must be rejected.
3.Outdoor context: The image should show natural daylight, soil, open air, ground, or surrounding vegetation.
The rest of the stack is:
- Frontend: React 19, TypeScript, Vite, Tailwind CSS, Motion, and Lucide icons
- Backend: Node.js, Express, and TypeScript
- Authentication: Firebase Authentication with Google sign-in and explorer login
- Database: Firebase Firestore
- Local persistence: Browser local storage for cached profiles, quests, journal entries, badges, and progress
- Audio: Browser SpeechSynthesis API and Web Audio API
Why Does Open Innovation Matter?
The central problem is not simply identifying a leaf. It is protecting the purpose of going outside.
A closed image API can return a label, but Nature Go needs a transparent and adaptable verification policy. The project must be able to distinguish a real outdoor discovery from a photograph of a screen showing nature.
Using Gemma as an open-weight model makes that direction possible.
It gives the project a model layer that can be inspected, served in different environments, adapted for local nature vocabulary, and replaced as better open models become available. It also creates a path toward privacy-preserving or offline exploration instead of forcing every outdoor image through a closed service.
Open innovation matters here because different communities may need different quest sets, languages, accessibility features, safety rules, and local ecological knowledge. The open foundation makes those changes possible without rebuilding the entire experience from scratch.
Most importantly, open AI helped me build a tool where AI supports the physical world instead of competing with it for attention.
The best outcome is not a user spending an hour inside Nature Go. The best outcome is a user completing one quest, hearing one nature fact, putting the phone away, and noticing something else nearby.
Prize Categories
Best Use of Gemma — Gemma is used as the open-weight vision verification layer that powers target matching, anti-spoof checking, outdoor-context verification, and structured nature-fact responses.
Links
Live demo:
https://nature-go-outdoor-vision-quest.ai.studio/
Source code:
https://github.com/RabiaA-arif/NatureGo
Challenge:
https://dev.to/challenges/hacktoberfest-week1-2026-10-05
Top comments (0)