@m_ashcroft I used to be able to represent “what would visa say” pretty well, but now that corpus is large enough that the deviation can be really substantial

Posters: visakanv. visakanv

2 likes recorded by the source corpus.