Forecast vs. reality
Use with: build-website. Resume line when done: Built {project_name}, a daily forecast-accuracy study with paired observations — live at {url}
Brief
Same spot, same time, every day: record the temperature at your window next to what the forecast promised for that hour, then chart the gap. It is a paired-measurement study anyone can run with a thermometer and a weather app, and the result is a page about how much to trust your local forecast.
Personalize
- Your station. One spot (windowsill, balcony, porch), one thermometer, one fixed time. Note the spot's quirks: afternoon sun on the wall, heat from a vent.
- Your forecast source. One named app or site, and the forecast captured the evening before for your reading hour. Screenshot it; forecasts change overnight and you want the version you actually saw.
- Your duration. Two weeks minimum; a month if you can hold the routine. State the dates on the page.
Milestones
- Set up the station and the log: date, forecast (with capture time), observed reading, conditions.
- Collect daily; screenshot each forecast when you capture it. Missed days get logged as missed.
- Structure the log into one data file; compute the daily gap.
- Chart forecast and observed as two lines, plus the gap, axes labeled with units.
- Publish: charts, full table, and a written conclusion covering average gap, worst miss, and why your window is not a weather station.
Done means
- Live URL with charts, the full paired table, and the conclusion
- Every forecast value backed by a screenshot taken before the observation
- Missed days visible as gaps
- The conclusion names the study's limits: one spot, one source, uncalibrated thermometer
- At least 5 commits across at least 4 distinct days, authored by you
- SESSIONS.md has an entry per session, written by you