David

Happy birthday Annie!!!

"And on occasions kill them."

February 15, 2026 The Trump administration’s white nationalist project was on full display this weekend at the 62nd Munich Security Conference that took place from February 13 to 15, 2026.

February 15, 2026
open.substack.com/pub/heatherc...

Interiors.

Yes. And Yes. But you can't say that's "the most power". Even appellate and circuit judges have significant power at the state and federal level.

But I think arguing this point obscures the real problem. The country is seriously divided, and thus, so is Congress. Thus its powerlessness.

There's a really cool scene with a young Fisher Stevens on the 2 train that's one of my favorite scenes of all times. The punchline is so prescient for NYC at the time.

No, really it isn't. Mcgolrick is right. If the Congress passes a law and the Supreme Court rules it's unconstitutional, it is null and void. If the President vetoes a law, it is null and void (unless Congress overrides it). Checks and balances. Coequal, with different realms of authority.

I guess you think that song is about you.
Don't you.
Don't you.

The latest episode of Fallout is so Trumpy. Down to the ICE rank and file.

A classic!

David Byrne was interviewed a few decades go and the topic turned to how Europeans reacted to his music. He said they loved it because they thought it was so "happy". He realized they didn't understand the lyrics. He doesn't do happy. Snarky and biting, yes. Happy, no.

Annie Lennox: Cold (From Diva)

One of my favorite lines from of one my favorite songs.

I do. It's amazing. Possibly my favorite game of all time.

Horny old bugger.

Where Winds Meet. It's FTP.

😬

I’m sure Russia knows what we are going to do before the American people since they have so many people serving in the Trump admin sharing intel with them.

WATCH: The 60 Minutes CECOT Segment I was sent the CECOT segment anonymously in a group chat.

BREAKING: WATCH the full 60 Minutes CECOT segment here. This was sent to me anonymously. It appears to be the segment CBS' Bari Weiss killed. www.muellershewrote.com/p/watch-the-...

Nope.

He's as fake as his eyeliner.

This is consonant with the last 30 years of the tech industry. Move fast and don't worry if you do harm, as long as you get rich.

Bari Weiss in the Conference Room with the TV Remote.

NAILED IT!!!

Machine learning is real. Real "AI" may be one day, but it isn't right now. Everything OpenAI, Microsoft, Google, etc., are selling is smoke and mirrors.

That's what happened in the 80s. AI was oversold as a panacea, investors poured boatloads into it, and it collapsed.

Because it wasn't real.

Eliza did fool some people some of the time. ChatGPT is simply a much more expensive (and climate unfriendly) version of Eliza. It is not intelligent in any way, shape or form. It doesn't actually know anything and certainly can't reason.

It's a parrot on a perch saying, "You're an idiot."

9/x

Except they're not. In the 1960s, Joe Wiesenbaum, a professor at MIT, wrote a program called Eliza. It was a rudimentary pattern matching program that would take the role of a psychiatrist. Users would make statements or ask questions and the program would respond with a saccharine response.

8/x

In the last 10 years, researchers and companies started to figure out how to "reverse" statistical models to mimic understanding by churning out responses based on some input (type into ChatGPT). As they trained on larger and larger datasets, these GenAI models appeared to become "smarter".

7/x

Then the Internet happened. Fast forward to 2010 or so, companies started collecting all sorts of data. Wikipedia happened, websites happened, StackOverflow happened. It was a data goldmine. Finally, researchers (and companies who wanted to use ML to create "AI" products) were in their heyday.

6/x

Statistical models showed great promise as research went on. There was just one problem. They required training data. And there wasn't enough of it for general application (think SIRI/Alexa). Most experiments were based on small domains and (relatively) tiny datasets.

5/x

In the early 90s, there was a shift toward statistical models – those that were not explicitly "taught" by humans, but taught by feeding training data into statistical models. At that point, neural networks were well understood computationally. But there were others (HMMs in particular) in use.

4/x