> it makes me very curious about what they’ve created this time

The rainbow and sunset alerts are really cool ideas. I'm now realising that a simple tie-in to astronomical phenomena could prompt a useful notificationa around it e.g. being worth going stargazing that night. I ski–learning that the near-term forecasts just changed would help me change my schedule the day before versus trying and failing the morning of.

I'm almost shocked we don't have a large weather model instead of a language model. Seems right up the alley.

Also I don't get what happened but I think it was AccuWeather or weather underground in the early 2000s where it was to the minute accurate and it seems like it's gotten worse since everywhere.

Google, Microsoft, Huawei, NVIDIA all have AI weather forecast models:

https://deepmind.google/science/weathernext/

https://microsoft.github.io/aurora/intro.html

https://www.huawei.com/en/news/2023/8/pangu-weather-forcast

https://blogs.nvidia.com/blog/nvidia-earth-2-open-models/

A Swiss startup named Jua does this for energy markets. Disclosure: I used to work there.

I think Google's weather models could be called LWMs. They're doing interesting research in this space.

Weather Underground had a very reasonable feed that you could subscribe to back in 2000. I use d then for a cluster of farming websites I built then.

> I'm almost shocked we don't have a large weather model instead of a language model. Seems right up the alley.

We do have such models. A bunch of them actually:

- Google DeepMind's "WeatherNext2" - Microsoft's Aurora - NVIDIA's FourCastNet-3 + Atlas + Climate-in-a-Bottle - ECMWF's AIFS ...

The list goes on. Plenty of small startups have repeated the recipe for building these types of models with their own architectural twist, too.