I don’t want to sound like a downer, but I’m skeptical about Armada: never heard of it before, repo created last June, tons of features implemented in that short span and already “ready to go”… Seems like it has been heavily vibe-coded.
It’s not a LaTeX successor, but rather an alternative compiler. It’s more pleasant to use than other LaTeX compilers, but I’ve noticed discrepancies between its output, and the output of whatever standard compiler I used, so I had to drop it.
The term you’re looking for is GPGPU (General Purpose computing on GPU)
There were some controversies due to the toxicity of Hpyrland’s community (namely its discord server). See this post for example
Edit: I should have read OP’s linked post first, which actually also refers to Drew’s post.
700+ preprints is a lot and takes time to be ingested, especially when the text is generated. For example, Communication in Mathematics had less than a hundred submissions in 2025 and accepted a fourth of them.
So, the small correction rate might be due to the bottleneck at the verification stage, which OpenAI should have done in the first place, rather than throwing everything at mathematicians and expecting them to do the ungratifying work whilst they reap the fame.