[tptacek]

He doesn't like or believe in the economics of the frontier labs and then works backwards from that, which takes him to some very strange places.

[dragontamer]

And I'm sure Galileo didn't like the idea of geocentrinism either for irrational reasons. Seriously, good enough telescopes havent been made yet and he's all confident etc. etc.

But Galileo was still had a useful framework for his criticisms despite all of his faults.

----------

I see Ed Zitron in virtually the same spot. Especially now that the SPVs thing is publicly confirmed by 3rd parties. I agree he made a bunch of shitty arguments over the years but Ed Zitron has seemingly stumbled upon something true and worth investigating now

[keeda]

The SPV thing has been known for months though; here's Matt Levine talking about it in his inimitable way last October: https://www.bloomberg.com/opinion/newsletters/2025-10-29/put... (https://archive.md/A40Fv)

Maybe what's become more public is the scale of these SPVs, but Matt Levine's writing makes it seem like a sensible thing to do given the $$$ amounts involved.

[dragontamer]

Matt Levine doesn't talk about what's good in finance. Matt Levine talks about what's bad in finance. Very consistently.

If you think that Matt Levine is making something sound sensible, you probably are falling for his dry sense of ironic writing.

[keeda]

That's totally fair, I was saying that Zitron has not discovered anything new, he just seems to have latched on to the next thing that lets him harvest anti-AI sentiment.

What I meant by "sensible" is, if you were in the hyperscalers' shoes and wanted to spend a trillion dollars that you didn't have while minimizing your risks and it's all legal, why would you NOT engage in this kind of financial engineering?

(I know the answer is something like "to avoid a potential economic crash," but hey, we're playing Capitalism here!)

[tptacek]

I don't have a direct opinion about Zitron's economic takes. I don't have a stake in that argument, I don't do financial planning for inference providers or training, I'm open to a variety of arguments, in either direction, about viability.

But I have direct subject-matter expertise with software engineering, and too much of what Zitron has to say about AI-mediated software development is clownish. Not wrong (everyone's wrong all the time), but literally risible. I can't get past that.

My thing is: there are smart people doing analysis with skeptical takes. Zitron gets cited because he has a business engine going revving people up and arming AI skeptics with (what I believe to be) Gish Gallup takes. It's profitable for him, and I don't begrudge him, but when people I otherwise respect cite Zitron it makes me sad: Zitron is going to make fools out of them. Cite better skeptics!

[dragontamer]

For all of Zitrons faults, he has workable numbers that a can be used and cited in arguments.

[tptacek]

I believe that he might, but anything valuable he says will be said better and more credibly by someone else.

[dragontamer]

> someone else.

> Cite better skeptics!

You are welcome to share any that you know.

[tptacek]

Gary Marcus is the cite I usually give.