Siru

Considering the other comment, yes, this does seem like a better option.

This isn’t entirely true, some of these systems are simply not speed optimized, and even then there still always has to be one source of ground truth.

That’s not always great either though because if the AI miss-localizes things, end users in that language wont even know that their documentation is incorrect. Better to have incomplete than wrong documentation. At least then the end user will know they have to try to read a different language. It might be good to use AI to create an unpublished first draft that is then polished by a person before publishing, though it seems the AI is not good enough to be useful in that manner.

You can try to help fund ladybird maybe? But it is still in development.