I grew up in Kansas, where watching a thunderstorm roll across the prairie is its own kind of entertainment, and I've paid attention to the weather everywhere I've lived since. When I landed in Tucson I wanted to understand the desert's weather, so I built a small app around a forecast most people never see: the National Weather Service's Area Forecast Discussion, a technical, conversational document where meteorologists talk through their reasoning.
The app puts a layer of AI interpretation on top of it. I can highlight any sentence and have it explained in the context of the whole discussion, generate a summary anywhere from one line to three paragraphs, or pull out and unpack the key concepts. Highlight something like "split flow" and it tells me what that means for tomorrow.
What surprised me was that I started learning meteorology in the wild — not by reading about Tucson's weather patterns, but by meeting them in the day's actual forecast, right when I was curious about it. The AI isn't replacing the expert; it's adding a layer of interpretation exactly where I need it.
Later I came back to it from another angle, inspired by the chipper DJs reading the forecast through the clock radio in Groundhog Day: an automated job that gathers the NWS products, writes a script aimed at an amateur weather nerd, and reads it aloud in a synthesized voice, refreshed a few times a day.
It's on hold for now, but maybe there's more here — a lot of genuinely interesting weather is locked up in formats that are hard to get at.