Bring along
Check the weather and local conditions yourself. This is an activity idea, not verified route guidance.
LESS SCROLLING. MORE SKY.
Your next adventure
can start small.
Turn a little free time into a reason to step outside.
A personal outdoor plan, powered by AI inside your browser.
01 / MAKE ROOM FOR OUTSIDE
What feels right today?
02 / YOUR NEXT SMALL ESCAPE
A little fresh air.
A different perspective.
Your plan will appear here. Choose your mood, your time, and a place you already know.
03 / BE HERE, OUT THERE
Your screen can wait.
Read your plan, then put the screen away. Come back when you're ready.
Time is measured by the clock, not GPS or activity tracking. No background alarm.
YOUR FIELD JOURNAL
Small moments, kept.
No completed sessions yet.
Setup, privacy & how this works
This app uses three files: index.html, style.css, and script.js. Keep them in the same folder. It uses WebLLM instead of Ollama: an open-source engine that runs Qwen2.5 on your browser’s GPU. The model weights are downloaded separately on first use, not embedded in these source files.
Open in a current Chrome or Edge browser. Open index.html using your editor’s localhost preview or an HTTPS static website so the JavaScript module can load. You do not need an AI server, Node.js, Ollama, an account or an API key.
The first load retrieves JavaScript, WebAssembly and model weights from external hosts (jsDelivr, GitHub and Hugging Face). These hosts see ordinary download requests; this app does not send your planning preferences or journal to them. WebLLM normally caches the model, but the browser can remove cached data. Offline reload is not guaranteed: this app does not bundle the runtime or install an offline service worker.
The small model may ignore constraints or make mistakes. Review its plan. No live maps, weather or location verification is included. The timer does not track physical activity or produce a background alarm. Active sessions are not saved on reload. The journal retains up to 200 entries in this browser and is erased by clearing browser data.
Model integration is implemented against the WebLLM API; actual GPU inference must be tested on your device. Source references: WebLLM documentation and Qwen2.5 model.