The Weather at Home

nixgeek 13 points 13 comments November 17, 2026
alexhowells.com · View on Hacker News

Discussion Highlights (6 comments)

MostlyStable

Cool! I also have an ecowitt station (third in line of stations, behind a SkyWeather2, and an Argent WS1). Years ago I made myself an R Shiny Dashboard, and since then had Claude make a replacement (I really like my dashboard but R Shiny is slooooooooow). I focused more on pure data. Mine is pretty much nothing but graphs without description or text, but I've got record low/high for selected date range, various aggregation levels (from the native 5 minute data all the way up to yearly), integration with my rain gauge records, and a climate tab showing rainfall/temperature records for the past 8 years of data. Pretty much only the rain/wind data is actually useful for us, but I've had a lot of fun over the years thinking of how I want the data displayed, etc. It's a fun hobby.

luckydata

please do a little bit of edits after you get a draft from Claude, the writing is so grating and clearly ai-sloppy.

johnz

I did something very similar to this a few weeks ago. https://johnzanussi.com/weather

lukeify

Maybe just me but I've developed an aversion to any of the obviously Claude-like designs an LLM puts out, which this project obviously uses. They work for proofs of concept, but it really isn't hard to tell the agent to create your own design language for a project.

trvrprkr

You mentioned pressure readings not entirely making sense. I had a similar experience across two Tempest units (different generations of their hardware, even). They tout doing a lot of data quality QA and adjusting things on their end, but nothing ever settled in a good place for me. And various sensors seemed to drift over time as well. Maybe they’ve figured things out and you have better luck.

envoked

Super cool. Earlier this fall, I got annoyed with my sprinkler system and how it couldn't accurately adjust to the rainfall at my house. I had Sonnet build MCP servers on top of Tempest and Hydrawise API, all orchestrated by Hermes. Works like a charm.

Semantic search powered by Rivestack pgvector
13,334 stories · 126,150 chunks indexed