What's the Deal with Zohran's Millionaire Tax? Here’s a 2-pager I wrote about Mamdani’s “millionaire tax.” Sources are at the end.

Phenomenal explainer of Zohran’s proposed millionaire tax from @economixcomix.bsky.social

My fat pasty white ass rode the metro in Chicago confidently last time I was there. They may not like it, but this is what peak performance looks like.

I kind of think about Citizen's United more than Dobbs for cases where the SCOTUS is like "and furthermore..." and then just does something terrible.

Maybe I'm mis-remembering it but wasn't that a significant overstep of the question at hand?

I just can’t believe our Reichstag Fire is the Big Balls Beatdown.

You wanna get roasted, go somewhere that "Truck Guys" congregate and point out they're paying 90k for a truck that should cost a third that and then call it the "Redneck Aesthetic Tax" and make sure you've got your asbestos undies on because you're gonna catch some heat.

Speaking from experience.

I used to think it was just the folks I support not bothering to try to learn but while that happens sometimes a lot of it is also just... stuff isn't designed to be learned a lot of the time.

Total agreement.

Also your screen name is hilarious. I regret I only have one updoot to give.

Stick figure with pony tail talking to another stick figure. Pony Tail: "Silicate Chemistry is second nature to us geochemists, so it's easy to forget that the average person probably only knows the formulas for olivine and one or two feldspars." Stick Figure: "And Quartz, of course." Pony Tail: "Of course." Quote at the bottom of the comic: "Even when they're trying to compensate for it, experts in anything wildly overestimate the average person's familiarity with their field."

I'm pretty good at figuring out user interfaces in apps and websites, but some programs if I go 2-3 weeks without using the program again I'm completely lost for a while.

This is also a problem with technical writing as well and reminds me of the classic XKCD comic.

I work with laymen users all the time and the UX of so many things is so painful that even after 10, 50, even 100+ uses of the app or website, they still can't remember what to do to get something done. Primary job loops never sink in and become intuitive, even after a year of daily use.

The conventions that came out of the mid to late 00s, otherwise known as Web 2.0, are kind of shit! This is when the hamburger menu reared its ugly head and gratuitous Flash and Javascript took over the Web.

You can still go see the first website ever made, but every website from the 00s is dead.

As opposed to Cuomo who... *Checks* picked a fight with Chicago???

True: @zohranmamdani lives in 1 unit of rent-stabilized housing.

Also true: Andrew Cuomo oversaw the loss of 96,064 units of rent-stabilized housing. And killed NYC’s program helping people move out of shelter.

Leaders must show moral clarity. Andrew, time to move out. 🧵
x.com/andrewcuomo/...

@thelouvreof.bsky.social

It's a cheesy film but he wanted to make an indie cyberpunk art film off of a foundational text and the studio tried to make it a tent pole so....

Johnny Mnemonic. Bad but fun movie. He was fun in it.

This is an easy block.

That's kind of my point. There's technology approaches underlying the "AI" that the product is advertised as having, and it can be dozens of different technologies. We don't know but we house it all under the term "AI".

You're the type of person Ken was originally referring to. The technology can't fail you, you can only fail the technology.

I'm done. Enjoy fucking with the bullshit engine.

Most of the coders that I respect that have dug into coding copilots are deeply skeptical about the benefits it can provide. It provides poor code. It also enables vibe coding which is arguably worse because when something breaks the vibe coder doesn't know how or why things broke

Are we talking about chatGPT or are we talking about the more traditional solutions that have fallen under the category of AI for decades?

Because open AI has absolutely muddied those waters to the point where calling it AI is pointlessly vague

Like, you yourself said that judging models for their ability to provide "factual" information is a mistake.

Coding copilot is fraught and it's turning out generates vulnerabilities constantly.

Same with personalization of generalized content- It's unreliable because it just makes shit up 15-40% of the time.

So pattern recognition isn't reliable- It hallucinates there too. GPT-5 apparently can't count how many "b"s there are in "blueberry".

tens of billions of dollars a year in operating costs for "lorem ipsum" is a hell of a take.

Data extrapolation isn't reliable by your own admission.

Let me put it this way. If you can't trust what an LLM generates since it's wrong 15-40% of the time, so trusting it to provide you *factual* information is a mistake by your own admission, what the hell is it good for?

. Businesses are making business decisions based on that bloviating. I watched my management sign up for a GPT-powered agent after being told that at some point in the near future the product would replace swaths of technical workers with scripts generated off their work documentation.

Sam Altman literally said that ChatGPT should be "solving all of physics" right around now.

My management just spent 45 minutes of our hour long town hall gushing about AI and how much of a "fact-producer" it is, to use your term.

That's it's use case right now in business.

Statistically the number is anywhere from 15-40% depending on the subject being tested. I don't think there's numbers out for GPT-5 at this point but it'll be interesting to see if it's better or worse. New models have not improved their error rate, and in some cases have gotten worse.

But that hammer costs tends of billions of dollars a year to maintain and like, you have 8 nails a year you have to drive. And 20% of the time it smashes your foot and then assures you that your foot is actually a nail.

I mean we have to first specify what underlying technology we're talking about when we talk about AI. Otherwise it's like talking about "chemicals" and saying there are genuine use cases where "chemicals" are great. It's vague enough to be a basically useless statement.

Harvard Business Review shows that in anno domini 2025 the #1 use for LLMs is apparently therapy and "relationships". As the technology "matures", it's use cases become more and more soft/humanities driven. Which is terrifying in it's own way.

bsky.app/profile/spav...

AI is such a vague catch-all term that it's almost useless to use. Machine learning? GPT? LLM? Neural network? AGI? There's a ton of different fields inside AI.

"AI" products have been used for decades and can be immensely powerful. LLM/GPT models have *far* more limited use cases.