I’m about halfway through a proper OCR run with segmented layout regions to extract actual metadata from the emails. I hope to post a proper update to the corpus tomorrow along with graphs between named entities.

And the iMessage transcripts are about to get an update unmasking a few names.

Or, I guess, people and emails would be different node types, and the connections would represent to, cc, etc?

Most of my intrusive thoughts are just database schemas

I really want to see this dataset in a graph database. Nodes are people, links are emails, go nuts.

I think all the time about hobbyist-level machines that were sold circa late Industrial Revolution that included a user manual with full disassembly instructions.

It certainly does! I think LLMs are incredibly powerful tools and I use them every day, and I think the extreme hype is extraordinarily delusional, and that the market is in a bubble.

"Technology should be invisible, benefitting me while I don't think about it at all" - sure! I adore this type of technology, when it works well it feels like magic!

But it still works better when you're aware of it and know how it works.

Yeah, I’ll cop to some of this. “Illness is way down so what do we need vaccines for?” thinking

“Girlfriends” (statutory)

It also depends somewhat on who we think "trump's base" exactly is, but at least among the hard right evangelical NAR cohort, yeah, all kinds of child abuse are not just fine, but expressly permitted, since children are believed to be the material property of their parents.

Plus, having the transcript allows you to feed it into an AI to help with homewo--wait a second

The admins of the deep internet are all perverts (reverently) long may they reign

Put all 3 names on the ticket. President, vice president, and attorney general.

Nah I used the generator to make the words I wanted and then I rearranged it in pixelmator.