Hmm, one application could be to use for a lossy text compression system. Start it off with a few hashtags and words, and then use entropy encoding like ANS to store the probabilities for the dynamic top-K.

You'd need a larger model to have the embeddings (to set how "lossy" it is).

My hanzi never looked beautiful, but we were mostly practicing stroke order just to help us read characters. We used pinyin first which I think is the way to go.

Wait, what? How did handedness come in?

I suppose I would have said "Zhongwen" if I meant the language alone. Whenever I try to say "I studied Chinese language (Mandarin) for a year" ("Zhonggua Hua") speakers think I said that I was in China for a year… but if that were actually the case my pronunciation and grammar would've been better!

Well, I took a year Mandarin in college, but that was 29 years ago. I was really into Taoism at the time.

1.239x, skeettag wakandaforever

My cat says hi but she stepped out of her box by the time I opened up the camera

Ah, there was the time of screen savers and the time of Matrix screen savers, and when you watched The Matrix again it felt like the equivalent of seeing the "DVD" logo bounce around every screen.

Is that why you leave my high fives hanging, because you can't stoop that low?

I don't know the character limit on Threads but it appears to be just right. At least, now that I'm experimenting with posts and cross posts