AI’s chicken sandwich problem

I’ve written before about being between two communities with very different attitudes toward our current generative AI moment:

• Artists, writers, and other (non-tech) creators mostly loathe it with every fiber of their being, to the point where they’ll vociferously argue that so much as asking Claude for a list of good curry places in London is a betrayal of the Enlightenment.
• Tech people often embrace it even when they admit the downsides. There’s a lot of “the energy costs are horrible, stolen works are used in training, the economics are irrational, also here’s the five apps I vibe-coded yesterday afternoon” happening.

So, it’s not surprising that my own attitude remains conflicted. I’ve been chewing on this article for over two weeks now; I suspect I’ve written five or six thousand words, most of which ended up on the cutting room floor (the Emacs scratch buffer).

Full disclosure: I’ve used LLMs, lightly, in non-creative tasks. I’ve had Claude spit out a few Emacs Lisp functions for me, and write a Python script to convert the dates of a thousand or so Day One entries to the format org-journal wanted. I’ve had it build a simple data table or two. I’ve used ML-based image editing tools to judiciously remove background elements in photos like an overhead electric line.

What I haven’t ever done is ask Claude to be my writing partner, Codex to write a competitor to Scrivener, or ChatGPT to be my magical robot therapist-slash-girlfriend. I think there’s a fundamental difference between the first set of uses and the second set that boils down to enhancement vs. replacement. Asking Claude to write an Emacs function for me saves me a bit of time; asking it to write a research paper for me replaces work I, not the machine, should do, and more than likely perpetrates deception. Using Apple Photos’ Clean Up tool gets me a better result than I’d get after an hour or two futzing in Photoshop; using a generative image model to create a book cover for me literally replaces a human artist—by synthesizing the work of thousands of human artists who get neither credit nor compensation.

And I think that’s a perfectly reasonable way of looking at it: the problem isn’t the tool in and of itself, but how we use the tool. But I’m not sure that it’s enough. It’s not only how the tool is used, it’s how the tool is promoted and packaged, even how it’s designed and implemented. It’s about who benefits from the overblown hype and, at risk of sounding conspiratorial, how it advances an agenda you may not be on board with.

It’s about the chicken sandwich of it all.

Nerds would like to treat the technology as existing in a vacuum, as being separate from the companies who create them, from their agendas and their politics. But LLMs are not, as much as we wish they were, apolitical spherical cows. Recently I came across a link to writer Mike Brock’s newsletter; in one article, he makes the observation,

The AGI [artificial general intelligence] hype does more than extract capital from credulous VCs. It builds philosophical architecture justifying surrender of democratic judgment to those who control the servers.

You might think that’s overstated, but recall that one of generative AI’s biggest boosters is literally the same fucking guy that made many of us leave Twitter. A disturbing number of other ostensible luminaries in the field are openly antidemocratic, or are highly influenced by those who are. There’s no escaping the uncomfortable truth that the rush to AI all the things, to put it everywhere and use it for everything, is a rush to cede control of wide swaths of our business and personal lives to a small number of largely far-right multibillionaires. There are legitimate reasons why generative AI has become the poster child for everything wrong with the tech industry in 2026.

Okay, there’s an obvious rebuttal: come on, generative AI is not, in and of itself, pro-fascism. It can’t be. It’s just a tool. Yes. Absolutely true.

Likewise, a chicken sandwich from Chick-Fil-A isn’t, in and of itself, anti-queer. It’s just a chicken sandwich. But I’m still not going to give Chick-Fil-A my money. The issue isn’t the product; the issue is what they’re going to put that money toward. Driving away with a combo meal doesn’t mean I agree with their philosophy—yet I would be, in a literal sense, buying into it. This is why I don’t eat there. This is why I’m not on Twitter/X. This is why I’m not going to buy boxes from Uline or stickers from Sticker Mule.

You get where I’m going with this. Sam Altman and Dario Amodei are both hypesters, but they’re not just hypesters. Granted, Altman is in every effective way worse, from his personal philosophy to his company’s noxious market positioning. (It’s hard not to notice that nearly all stories about people developing “AI psychosis” are stories about ChatGPT.) Amodei, though, is positively soaked in TESCREAL, and Anthropic was the company sued—successfully—for massive copyright infringement. As for the next tier, Elon Musk is, you know, that fucking guy. Mark Zuckerberg is that other fucking guy. And Google has been that fucking company for years now.

I don’t think I have qualms about asking Siri AI the occasional complex query, or to draft a Shortcut. Yes, I know Siri AI is partially based on Google models, don’t well-actually me. But beyond that? If I had to choose between Claude and ChatGPT (let alone Grok, Gemini, and whatever Meta calls theirs, I don’t know, Lolita?), I’d choose Claude. But it’s hard to shake the suspicion that the most ethical answer is “none of the above.” We are at a curious moment where the least concerning companies on the politics and philosophy axes may be Chinese ones like Qwen and Kimi. Which is pretty concerning.

Where your personal lines lie are, well, personal. While staying to the enhance side of the enhance/replacement line increasingly strikes me as falling into the “necessary but not sufficient” bucket, I know not everyone will agree with me. But here’s what I do know: look, you just don’t understand how incredibly delicious the frontier chicken sandwiches have gotten in the last six months is not an adequate defense.