Pavel

After six months of Claude Code, Microsoft has had enough

Microsoft reports expose AI's cost problem: The tech is more expensive than paying human employees | Fortune fortune.com

AI boosters claim that LLMs are the future of software development. But that could not be further from the truth. LLMs are the apotheosis of a paradigm 40 years in the making: the tech industry assigns you problems and then sells you the solutions. But that world is now coming…

productpicnic.beehiiv.com

your performance rating is now tied to AI use meanwhile "88 percent [of executives] expressed concern that employees are using AI performatively rather than adding business value"

theregister.com

The second one by Evan Garris describing how those same programmers gave up their only leverage on their employers, by exchanging their relationship with reality (which corporate fiat could not overrule) for a performance of fluency for its own sake. AI code is the apotheosis …

evangarris.com

Oh, you mean a manager?

productpicnic.beehiiv.com

The user base is stratified between an extremely small number of whales, who will pay anything to feed their addiction, and normal people who get tricked into clicking the "use AI" button by deceptive UI patterns. Single digit rates of paying customers aren't an aberration.

zdnet.com

Lots of thinkpieces about how workers should think of "AI agents" as colleagues. Which, sure. I've had colleagues I always avoid because getting them to do anything useful is like pulling teeth, and who can be counted upon to fuck up even a rudimentary task if left unattended.

It's a shame that I'm writing my next newsletter on the plane after negative six hours of sleep, because the theme of the links is a bombshell: Generative AI is not "the future" of software development, but the last dying gasp of its past. MOAR & FASTER can't shore up the crumbling foundations.

In an era of AI generated slop, I'm delighted to see that the art of shitty photoshops for throwaway jokes is alive and well The future is bright for memes, for Spirit Airlines, and for America

"Anthropic is profitable" is the funniest joke I've heard in a long time

AI boosters exist in a superposition of insisting that there is nothing wrong with using AI to "write" + simultaneously claiming that their "writing" is not AI-generated. And that's how you know they don't actually believe that AI is better than humans.

Old article structure: inverted pyramid (the most important content first so readers can get everything they need from the first few paragraphs) New article structure: inverted quality (writers give less and less of a shit the further you go into the article and leave more of the AI text unedited)

From the genre of "either AI generated or plain bad writing" is the superfluous signposting: - "here's what I mean by that" - "this is the part most people miss" - "the takeaway:" You don't have to tell me that you're about to say something. Just say the thing.

The entire value of my writing is that you know it's a REAL bad opinion held by a REAL human. There's no catharsis in getting mad at dogshit takes if they're AI generated.