A variety of things: Books, movies, music, scientific journal publications, Slackware's "current" branch with all past packages since 2009 (only half a TB though), All Slackbuild sources, an almost-complete crawl of CentOS 6 packages, large language models and datasets (almost 8TB now), an old TankNet archive, a few wikipedia dumps about two years apart, chat logs, archived email, a lot of smaller archives of niche interests .. it's something of a mess.