Comment by cubefox

> we can decode any embedding vectors, given we have at least a few thousand of them.

Do you think that could be sufficient to translate the Voynich manuscript?