At its core, AI models just have a score that their objective is to maximize, and uses historical data to identify what factors to give weight to in order to maximize that score in the future. And if not tuned properly, sometimes they'll take shortcuts to get that score. It's not really forecasting things with precise physical equations like the older models. Now of course its not *that* simple, since the Euro is experimenting with hybrid models mixing physical, AI, and human forecasting capabilities. And there are ways to suppress those biases too. But it is still vulnerable especially when its lacking data.

IMO that late month window has great potential!

lol, I asked AI about AI, and it said:

Hoodoo still closed this weekend with only a 19” base after picking up 2” of snow today.

That's a good theory. AI can most certainly be subject to seasonal biases. AI uses statistics, which can result in them falling for the "correlation but not causation" issue - and giving weight to the wrong factors implicitly (especially in the long range).