Pulled datasets off libgen for a corpus once and it's overwhelmingly in-copyright textbooks, the public domain framing doesn't really hold up.
hn1rig3rak · · focus · HN ↗