Walking Challenge AI: An Open-Source AI That Sends You Outside 🚶

Description and use

Walking Challenge AI generates real‑world walking challenges using an in‑browser Qwen2.5 Instruct model, runs locally without servers or API keys, and encourages students to step outside.

Problem

Students spend most of the day on screens and need a simple, fun reason to step outside.

Outcome

The app provides a quick, privacy‑preserving way to generate walking challenges, encouraging users to step outside without relying on external APIs or servers.

Workflow

• Choose walk time (10‑60 min), place (park, city, college, neighborhood) and difficulty (easy, medium, hard).
• The app builds a prompt and sends it to an open‑weight Qwen2.5 Instruct model running in the browser via Transformers.js.
• If the model output is unusable, the app falls back to a hand‑written bank of challenges.
• Walk, tick off completed challenges, and finish.
• The first run downloads the model (~1.2 GB) and caches it for future use.

Evidence

Walking Challenge AI is a small web app that turns an ordinary walk into a game.

Source

An open-weight model (Qwen2.5 Instruct) runs in your browser with Transformers.js and writes your challenges.

Source

The first run downloads the model (about 1.2 GB), so use Chrome or Edge and give it a minute.

Source