Lynch: 49ers' injury conspiracy 'big nothing burger' https://www.espn.com/nfl/story/_/id/48340939/lynch-49ers-injury-conspiracy-burger
Scouting Report: KC Concepcion has WR1 traits the Raiders need https://raiderramble.com/2026/03/29/scouting-report-kc-concepcion-has-wr1-traits-the-raiders-need/
How a Republican state lawmaker tried to let Holocaust deniers hijack history lessons (Tom Dreisbach/NPR)
https://www.npr.org/2026/04/29/nx-s1-5708847/holocaust-denial-conspiracy-theory-politics
http://www.memeorandum.com/260429/p78#a260429p78
And LLM is literally autocomplete beasedmon statistics in training data.
That’s it.
It doesn’t know anything or think about anything (even if companies call some of the inference steps “thinking”), neither does it have concepts or any grasp on reality.
I really like the concept of noticing "greyed out options": perfectly reasonable if sometimes wacky things that you can do in a situation but that you accidentally treat as unavailable, like a greyed out option in a dialog box in software.
Once you start noticing them, you unlock a whole lot of self-awareness.
An example tonight from a moderation decision: someone who's kind of a jerk and often gets their messages blocked on the forum, and it's never quite over the line but often borderline. Do we ban the person or not?
Wait: there's a third option! We can send them a message and tell 'em to knock it off and contribute better.
Unlocking that not only gets you a lot more information as a moderator to decide whether to keep someone, but also gives them the opportunity to change their behavior.
Court gives DHS conspiracy theorist access to 2020 election data (Matt Cohen/Democracy Docket)
https://www.democracydocket.com/news-alerts/court-gives-dhs-conspiracy-theorist-access-to-2020-election-data/
http://www.memeorandum.com/260428/p119#a260428p119