forgot to post this one in time

youtube removed sort by new on the search for literally no reason at all

hello hello, working on the next video, sry for the wait, will be out soon!!! (for real don't worry i won't make you wait 4 months this time around)

Stanford researchers just managed to prove how you can easily extract near-verbatim dumps of books from the training data of production LLMs by simply asking them to do that in a very specific way LMAO I can smell lawsuits coming

https://arxiv.org/pdf/2601.02671v1

every day i am cursed/blessed with the knowledge of some new kind of old abandoned hardware that looks kinda cool and i want but i also have no space or use for

spending 200€ a month to have an autocomplete on steroids trained on reddit interact with other bots pretending to be somewhat conscious in a reddit clone regurgitating what humans in their training data supposed bots would sound like in a reddit-like forum is now an actual hobby some people have

(this is not a dig at openclaw i am just shocked by how invested some people are in moltbook)