RIP, razetime.
[0] https://codegolf.meta.stackexchange.com/questions/26416/in-m...
In theory you should be able to define entire neural networks with the help of a handful lines of APL. You wouldn't even bother with complex frameworks offering you pre-built architectures. You'd just copy paste the 10 lines of fully self contained APL code that describes the network from the documentation, because even the idea of downloading a library is overkill.
My issue with APL is I was never able to turn the corner to "generic problem solving" in APL (or other array langs). It feels like learning written Chinese, like 50,000 individual techniques but if you know them you can do incredible things quickly. For the problems I know how to solve, I can solve them quickly. And you CAN do amazing things with inner products in APL.
On the other hand, studying APL, even if you don't master it, is not without benefits. LLM transformer architecture and GraphBLAS algorithms are junior APL level implementation problems (at least conceptually, operationalizing them is a different story).
Adam Brudzewski has one of the most criminally underrated YouTube channels[2]. It would be great to solve problems that elegantly in any language, and Adam has always been very friendly in answering questions if you ever get a chance to speak with him. I just seem to be a lost cause lol.
(It also has a few other benefits over Dvorak, optimising for a few more factors than Dvorak does)
Part of that is because unlike other APL-likes it uses a stack (sort of) and I can't explain exactly how but it made it much easier for me to picture how the data flows from one operation to the next (I have to admit I like concatenative languages a lot so I'm obviously biased here too).
On top of that none of the glyphs are overloaded with monadic and dyadic versions, they're one or the other, which reduces ambiguity a lot when trying to read/write code.
There's lots of other little ergonomic tweaks to it that make it really neat, but those were the big ones for me.
Also worth noting is that it has lots of multimedia support - you can generate pictures, gif animations, sounds. So it's easy to "play" with for fun!
However, I felt that writing anything not closely related to solving mathematical matrix problems made no sense to me. Unfortunately, APL is too niche; I don't know anyone in my industry with whom I could share the tools. Nevertheless, it was a valuable learning experience.
Anyway, under the assumption that I'm correctly guessing what you have in mind when using the words "serious language", Uiua certainly qualifies. The author is very passionate about exploring and discovering "the good parts" of the design space of the array language paradigm, and has put a ton of work into making it accessible and practically useful within the constraints of being an interpreted language that autoformats its source code to at-first exotic looking maths symbols.