This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass 🌿 Fieldcard — Make a plan. Step outside. Leave the scroll behind.
Modern digital platforms compete for our attention through endless feeds and notifications. Even when we want to spend time outdoors, planning an outing can become another reason to stay on our phones.
For Hacktoberfest 2026 Week 1 — “Touch Grass” — I built Fieldcard, an AI-powered outdoor outing planner designed to turn the intention of going outside into a practical, time-bounded plan. Fieldcard helps users prepare for an outdoor experience and then step away from their screens.
Here’s what it does:
Time-based planning: Choose an outing duration of 30, 60, 90, or 120 minutes.
Personalized outings: Enter a familiar park, neighborhood trail, or community garden and select your activity, pace, budget, and accessibility requirements.
Sensory challenges: Get screen-free activities that encourage you to notice the world around you, such as observing tree-bark textures or listening for different bird calls.
Packing checklist: Receive suggested essentials suited to your planned activity.
Closing reflection: Finish your outing with a short reflection.
Printable field card: Print the plan on A4 paper or save it as a PDF so you don't need to keep checking your phone.
Sample mode: Explore a clearly labelled sample plan when live AI generation is unavailable.
The goal is simple: use technology to make going outdoors easier, not to give people another feed to scroll.
🌍 Live application: https://fieldcard-touch-grass.vercel.app/ The application is deployed on Vercel. Try creating an outing plan and exploring its printable field card.
💻 GitHub repository: https://github.com/PAWANBHOWATE04/fieldcard-touch-grass The repository includes the application source code, API implementation, tests, configuration, and documentation.
I built Fieldcard using a React-based frontend and a server-side API for AI-powered plan generation.
Tech stack:
The application sends the user's outing preferences to the server-side API, which requests an AI-generated plan. JavaScript validation and normalization help keep the itinerary consistent with the selected duration.
I also added a sample-plan fallback so users can explore the experience when live generation is unavailable. Sample output is labelled to distinguish it from a live AI response.
The project passed 15 unit tests during my previous test run, and the production build completed successfully.
One important implementation detail: the model inference uses Google's hosted API; the model does not run locally. The API key is configured server-side rather than intentionally exposed in frontend code.
Open innovation makes it easier for developers to experiment with AI models, learn from available tools, and explore new applications beyond familiar chatbot experiences.
With Fieldcard, I wanted to explore a different use of AI: helping someone prepare for an offline activity rather than encouraging another extended online session.
Working with Gemma gave me an opportunity to explore an open-weight model in a practical, human-centered application. The model is only one part of the experience, though. Input validation, clear limitations, sample-mode transparency, and a useful interface are also important.
For me, open innovation is about giving developers the freedom to experiment, build useful tools, and discover new ways to apply AI to everyday problems. Best Use of Gemma — Fieldcard uses Google's Gemma 4 model (gemma-4-26b-a4b-it) to generate personalized outdoor outing plans.