It's not a big data set that lends itself primarily to analysis, it's more like content. For example, a list of all US Presidents with a lot of metadata or text content fields about them collected/combined from different sources, cleaned, corrected, annotated, etc. (Pretend Wikipedia has only a subset of these fields and considers broadening them out of scope.)
As for Github, the data would still be under "my" account and I'm thinking about more of a platform that doesn't depend on one person. Maybe I would manage day to day version control in Github but I'd want to promote occasional releases to be more official and not reliant on my account.
In the GPT-2 era I created CouldReads, a big data set of generated book titles/synopses trained on thousands of e-books. It was a fun project in the naivete of 2020 but it's less amusing now.
A while back I wrote up a way to turn the big Wikipedia XML dump into a database. Not a generic table with articles but thousands of tables, one for each article "type". I'm not sure if this is still the best way to go about it.
That looks like a crowdsourced project for turning arbitrary sites into RSS which is very cool, but I don't see a way to get a large RSS data set out of it. And with about 5000 sources (I think) it's not as large as what I was hoping for, but it could be a good complementary source.
And also a new project to fetch all links seen in the Bluesky firehose and gather metadata to build a database of sites and pages at a more granular level than the domain. For example, is account X posting video links from one YT channel or many?
As for Github, the data would still be under "my" account and I'm thinking about more of a platform that doesn't depend on one person. Maybe I would manage day to day version control in Github but I'd want to promote occasional releases to be more official and not reliant on my account.