Paper Walk: Print a Card, Leave Your Phone, Go Outside

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

What I Built

Paper Walk is a small web page that makes a printable walk card. You choose how much time you have (10 minutes, 30 minutes or an hour), where you are (home, street, park, water or forest), your area, and how you feel. A small open-weight model running on your own computer writes a few short prompts such as "Find three different shades of green" or "Listen for the farthest sound you can hear."

You print the card, put your phone in your pocket, and go. The screen part takes about a minute and the rest happens in the world, so the screen really is the shortest part of the experience.

After the walk you can save a one-line note about what you noticed. The next card builds on your earlier notes and avoids repeating them.

It's for anyone who is bored at home or stuck in a screen habit and wants a small, low-effort reason to look around. I built it for myself first, in Patna, Bihar.

Demo

Code

https://github.com/jeetsingh008/PaperWalk

It's a single HTML file with no dependencies. Run it with Ollama and a small model; the steps are in the README.

How I Built It

• Model: gemma3:1b, an open-weight model, run locally with Ollama.
• App: one HTML file with plain JavaScript. It sends your choices (time, place, area, mood, and your last five walk notes) to Ollama's local API and turns the reply into a printable card.
• Fallback: if no model is running, the page uses a small built-in list of prompts, so it still works offline.
• Memory: walk notes are saved in the browser's local storage and never leave the device.

The hard part was the model, not the code. My first version produced prompts with invented details, like a faded tiger mural on a wall near me, because a 1B-parameter model can't see my surroundings and simply guesses. It also ignored my "bored at home" setting at first. Telling it what not to do barely helped, but giving it a few examples of the style I wanted worked much better. The final card gave plain prompts like "Find something smooth, gray, slightly damp" that work anywhere.

Why Does Open Innovation Matter?

• It runs on my laptop with no account and no cost. I don't need an API key or a subscription to use it.
• My walk notes stay on my device. Notes about where I am and what I do each day are personal, and with a local model they never go to a server I don't control.
• I could change the model. When the 1B model gave weak results, I could try different settings and prompts, and a bigger model would be a one-word change in the page. With a closed API I would be stuck with whatever the provider offers.
• The tradeoff is real. A tiny local model is private and free but sometimes awkward, as with "the distance of a distant, slightly rumbling vehicle." For this project that was an acceptable price.