Ethan MarcotteModify

2026-07-13 12:00:00

While I was working on my latest redesign, I added a statement explaining how “artificial intelligence” is used on my website. And in a fit of pique, I slapped this line onto the end:

I am very upset I had to write any of this, but here we both are.

Look, I don’t know if the exasperation coursing through that line is wise, as a working designer who would like to continue to, well, work. But I’ll own the sentiment. If nothing else, it lines up with how I feel about large language models (LLMs), and specifically how their explosive growth has made everyone suspicious of everything they see, read, and write. And understandably! Putting aside for the briefest of moments how these technologies have led to new forms of misinformation and disinformation, it’s fiendishly difficult to know just how human-made something is these days. A few short years ago we’d never ask the questions many of us are now asking. ¹ And many of those questions are simply some variant on, “did you actually make this?”

These were just a few of the things swirling around my brainpan when I originally wrote the statement. Yes, I firmly believe “generative artificial intelligence” is a failed technology; I also think it’s worth disclaiming how I did (or didn’t) use it in creating this website, or the words published on it. Putting that on the record feels important, even if I resent feeling I had to do it when launching a new design in 2026.

But when I shared my statement on Bluesky and Mastodon, I quickly learned there were many, many other “AI” statements out there. As someone who’s perenially late to the party, I should’ve known folks were thinking about this long before I did. But I can’t tell you how helpful it was to learn I was stepping into a little bit of a tradition around these kinds of statements. And what’s more, I appreciated seeing a variety of approaches to writing about “AI” — and just as many opinions about how and when it’s appropriate to use it.

If you’ll forgive me for riffing on something I said before, I realized that very few of these statements are really about “AI” itself. Instead, they’re statements about work: how the author thinks about their work, and about their relationship with the tools they use. So I thought I’d share a few favorites here, in case you find them as useful I did. Do I agree with every entry on this list? No, of course not. But I’ve learned something from every single entry on this list — something about the values I bring to my work, and about the values I want to bring to it.

This is all to say, just like last time:

Hey, you. I’ve made you a playlist.

All the words were written by me, a human — typos and all. All of the images were made by me or another human. All of the code was typed out by me.

Unless otherwise stated or owned by a third party, everything you see and read here (questionable punctuation and all) was created by me without “AI” involvement or assistance.

I prefer to avoid AI usage for ethical, practical, and financial reasons.

1. Ethically: Generative AI tools, powered by data centers which consume vast amounts of water and pollute our environment, are built on the collective theft of the works of millions of artists, developers, authors, and other creatives, supercharge the spread of disinformation and fascism, have repeatedly provoked psychosis and suicide, and concentrate wealth in fewer hands while providing cover for widespread layoffs.
2. Practically: I have found that AI tools overcomplicate implementations and that cleaner, simpler solutions can often be written by hand. AI doesn’t “know” anything, and is making life worse for open source maintainers. Overreliance on AI risks deskilling.
3. Financially: I dislike sending money to large corporations when alternatives include learning the skill for free and contributing to open source projects. Tokens are currently heavily subsidized and likely to become more expensive, so I would prefer not to make a habit of using them.

I use AI tools sparingly for assistance while refactoring code in languages I understand. I occasionally use it to help compose command line arguments for tools like ffmpeg. I never use AI to generate images, video, or prose.

In some cases I am sceptic that generative AI will help me achieve my goals:

• writing helps me think. Writing itself is the point, not the output. Output alone has no value and is unnecessary.
• I’m on the indieweb to connect, as a human, to other humans.
• language models normalise, which is often uninteresting. There is no need to make content that is like everyone else’s. My voice, beliefs, experience and background are my only chance at at least trying to create something novel. Even if I try manually, I’ll inevitably borrow from other people by accident. Why would I automate or accelerate that?
• LLM’s shouldn’t invent new information

This style you’re reading now is for stuff like my blogs, my posts, my SubStack. Key word: My. This is my voice. And I want to keep that work for my brain because thinking about the idea, structure, writing, and editing is what brings me joy.

So no, I won’t be training anything.

1. None of the proper content on this site is written by AI unless I explicitly say so. The articles, notes and pages were written by me, an actual human. If you’re taking the time to read something here, you should know that I took the time to write it.
2. Many of my thoughts begin as rambling voice notes, dictated while driving or walking. LLMs are amazing for transcribing these, and for turning them into bullet-list outlines for me to riff off later. These outlines sometimes turn into articles on this site.
3. I often use LLMs for the following on this site:
   1. Pre-publishing proofreading, grammar, spelling & syntax checks.
   2. Generating dummy/example content in places where I’d previously have used some version of Lorem Ipsum.
   3. Generating meta descriptions and the like from my actual content. This is not automated and I inevitably edit by hand before publishing, but LLMs are unquestionably better at this than me.
   4. Drafting and updating functional content like the privacy policy.

While generative AI can assist with drafting, summarising, proofreading, or refining academic content, it cannot be considered capable of producing an original piece of research or submission without substantial human intellectual contribution and oversight. In accordance with [Committee on Publication Ethics] guidelines, generative AI systems cannot be listed as authors, as they cannot assume accountability for errors, ensure ethical standards have been adhered to, or respond to critiques of the work.

BERA expects all researchers to uphold the highest standards of academic integrity and ethical conduct in their work.

I’ll be plain about this, because I would rather you hear it from me than decide it for yourself. I use AI to help write these entries. On my own, as one person with a day job, I could not have built a catalogue this size by hand. The AI helps me draft. It does not get the final word.

Instead, I’d offer an alternative mental model: generative AIs are tools, devised and deployed by corporations to operate at scale, laundering content the corporations generally do not own with the active intention of reducing the need for, and the value of, human labor. Whether machine learning is like human learning is irrelevant to the real-world impact of its use.

Our recent experience has shown that the use of AI-generated make-up visuals not only contradicts our brand philosophy but is also rejected by our community. The response was clear and immediate: our audience values authenticity, real artistry, and the work of human creators. They expect and deserve nothing less from us.

Computer science is a complex discipline, and those who excel at it are rightly lauded, but so is understanding and critiquing power and holding it to account. Understanding technologies requires also understanding power; it requires social, cultural, political, and media literacy as well as technical literacy; incisive questioning and sober critique as well as shock and awe.

Our AI policy exists to do two things: protect the intellectual property of our speakers, and make sure AI tools don’t become intrusive or distracting during events. To support both goals, we ask that attendees not use AI tools to record, transcribe, or repurpose content from our sessions, including sending AI transcribers in your place.

We provide live closed captions during all our events, and we also offer video recordings afterward. If you need to use any additional AI tools for personal accessibility reasons, just reach out. We’re happy to work with you to make sure you can fully participate.

My approach to “AI” technologies, as with other tools, is to situate them within what we know about the world, how people learn and/or make meaning, and how we can build a future that improves outcomes for everyone.

Promising alternatives are already being created and used by people like the FOSS Academic, as well as ecologically-inclined or permacomputing-inspired groups like the Hundred Rabbits collective or Low←Tech Magazine.

While those approaches vary, they share a common theme of attention to the material costs of computing and to liberating tech users from influences that seek to undermine their own intentions.

Our bread and butter is information design: the pervasive challenge, in every domain, of taking a complex idea or story and distilling it into a clear, intuitive, beautiful format. If we’re making tools for experts, we want to respect their expertise and time, and craft a product that helps them with the difficult parts of their job, not another kludge that they’re burdened to spend time wrestling with and debugging. We want to make thoughtful products, where it’s clear that someone has spent time thinking about the hard parts of a task and has invested the time and effort into making that a smooth process. If we’ve found ourselves in a hole where the black box model of solving a complex information problem is the only solution, we’re not doing our job as information designers, and we need to go back to the drawing board.

To develop that body of knowledge, one must interact directly with the literature, use it to actively learn about connections between one catalog or book or article and the next, and between collectors, authors, and institutions. With it, one can often anticipate where to find a coin (and where else), often in ways or places inaccessible to current search techonologies. [sic]

And, of course, by developing those skills, one better appreciates the historical and intellectual significance of all that “object biography.” How those threads tie together. How they form a coherent narrative about the place of ancient coins (and this coin, specifcially [sic]), their scholarship, and the reception of antiquity across historical and cultural contexts.

I work hard at this project because I find joy, pleasure and fulfillment in the process of this labor. I can say I am proud of the results without needing to qualify it. Work is the medium through which we negotiate what we owe to each other. If you have paid me your attention, then I owe you my best effort, and mine alone.

The writing we use, the very substrate of our culture, is the result of a process older than we can comprehend. Who are we to decide that process is finished? That the time of invention and human input is done, and every letter created from now on should be a remix of what came before?

AI tools perceive reality as a finite set of data, up-to-date as of about 2021: the shadows on the wall of Plato’s Cave. Right now they are very good at making shadow puppets, but they cannot see the world outside.

My thanks to Lisa Maria Marquis and Sean Tubridy for contributing a few resources to this post.

This has been “Folded between the pages.” a post from Ethan’s journal.

Reply via email

2026-07-06 12:00:00

It’s been, like, hot-hot for the last week, with Boston seized by the same heat wave affecting…well, everywhere else. Personally I’ve never cared for summer, even before it entered its villain era. And after a few days stuck inside avoiding the heat, the cabin fever had properly set in. So when I saw the temperature was in the high sixties ¹ this morning, it felt like the best birthday present. Like some misaligned cog in the planet’s inner workings had suddenly snapped back into alignment, and things started humming along happily.

Naturally, I woke up too damned early, as is my curse idiom. I drank some coffee on the couch, and watched some Critical Role with the kittens until She woke up. Then I opened some lovely gifts, and fielded some calls and texts from friends and family.

After that, I went out for a run.

Pretty sure I’ve said this before, but I’m a slow runner. And I’m fine with that: I don’t run competitively, and I really just love getting outside for a couple hours. And whenever possible, I try to get down to the river.

(Fun fact: when you start running along the Charles River, nobody prepares you for the frankly unhinged levels of gooseshit you’ll be dancing around.)

I struck out, running down through Cambridge and across the water. On the way I stopped at the emergency room where I got my diagnosis a few years back, took a quick selfie, and then kept on running.

I’m entering the last year of my forties, and it’s hard not to feel a bit reflective about the decade I’m in the process of closing. Did some good design work; started and left some jobs; started talking about labor in tech, and wrote a little book about it. There are old friends who’ve moved away, and I’ve been lucky enough to make a few new friends; having dinner with two of them tonight, in fact.

There’s been loss and laughter aplenty. Throughout it all, She’s been there.

I cross the bridge, snapping a few photos as I do, and start heading toward home. Always and happily toward home.

This has been “49.” a post from Ethan’s journal.

Reply via email

2026-06-25 12:00:00

Two weeks ago, I quietly pushed up a new look for this website. Some redesigns happen over a year or two; others happen over a few months. Neither one’s easy, mind you, but this one’s the latter.

A month or so back, I told an old friend I was in the middle of redesigning this little site. She laughed and said, “Again?!” Which, you know, extremely and completely fair! The last design went up in March of last year, giving it a little more than 15 months in the sun. I loved it when it went up, and I still do: there were some little flourishes that felt right at launch, and I’m proud of myself for figuring out how to build a few design elements that I honestly didn’t think I could build.

But in last few months, I couldn’t look at the site without seeing every rough edge: the uneven gaps that had crept into the design, the too-clever layouts that felt fragile, the long passages of text that never felt like they nailed the rhythm. And the boxes! My god, the boxes: for related articles, for newsletters, for books, for little autobiographical blurbs. Boxes for just about everything but thee and me.

This is all to say that I started this redesign to pare things back — way, way back — and try to find my way to something a little calmer. A little more open, maybe.

This is where I’ve landed. At least for now.

In addition to generally simplifying the design, I knew I wanted warmth. A lot more. The blues of the last design were feeling too cold, too distant. I started reaching for reds, for yellows and creams, for an anchoring earth tone or two. I’m sure the work I’d done redesigning You Deserve A Tech Union was still swirling around in my brain, but the warmer hues felt like a nice change after a year awash in deep, dark blues.

And the new design has a number of textures I quite like. Some of them are new, like the splashes of watercolor that bookend each page; the navigation has a little blending applied to it, which I think gives it a nice gold leaf-under-oil effect. But it’s not all new: for one, the map relief is still with us, in part because I can’t imagine this site without it. (It’s adapted from a relief map of a mountain near where I grew up.)

And I suppose the site’s typography continues that theme of blending old and new, mixing new faces in with familiar ones. Klim Type’s Untitled Sans is still tucked in around the edges, but I’ve pulled in General Type’s Cambon for headings and titles. For body text, I’ve brought back Black[Foundry]’s Vesterbro, which I used (and loved) in an earlier redesign.

When it came to figuring out what needed to live on my homepage, I started by looking around a bit. I love Naz’s homepage, as well as mandy’s brand new one; you won’t have to squint too hard to see the influence of their homepages’ structure on mine. I love how both Naz and mandy treat their homepages like the start of a branching forest path: there’s a brief introduction to the author, then a few links to recent entries, and then a few hand-picked favorites. And that felt right for my homepage. With that in place, I took all the biographical fooferaw I used to have there, and shunted it to a proper about page. And that division feels right: after all, that content’s about me, not the website.

And finally, and maybe most importantly, journal pages are much more open and calm than they used to be.

The site’s layouts have been dramatically simplified throughout, and journal entries are no exception. I’ve ended my dalliance with sidebars and three-up grids, instead reverting back to one single column of text. And at least right now, I quite like the more streamlined reading experience.

I will say I’m glad I finally got a chance to play with view transitions — shout out to Durgesh Rajubhai Pawar’s excellent tutorials — which I think add a lovely little effect in browsers that support them. They won’t show up if you, like me, use Firefox as your main browser, but browsers like Chrome, Edge, or Safari should work. (In my testing, they seem to work best in Chrome.)

There are a bunch other little details throughout the redesign that I’m proud of, but I’m aware that this post is already long enough.

If I’m honest, I struggled with this redesign. I’d spent the weeks prior to launch doing nothing but unglamorous finishing work: building new features, tidying up ancient templates in Eleventy, and fixing bugs. It’s about at this point that the self doubt settled in. I wondered if it was too spare, or maybe not spare enough. And as always seems to happen around this time, I wondered if the design was even any good. Why didn’t things feel easier? Shouldn’t they feel easy if this redesign was a step in the right direction?

(I swear, a personal site redesign never used to feel this challenging. But then, I used to be a lot younger; that probably has something to do with it.)

I’ll also say that a new kind of worry crept in this time, one I’d never dealt with before. Just this week The New Yorker wrote a piece on the emerging aesthetic of LLM-generated websites, which all seem to be settling into a smear of same-y markers: of “beige- and cream-colored backgrounds, rusty orange-hued accents, and large serif typefaces.” Naz even mentions he tackled his redesign in part to avoid that aesthetic, and here I was worrying I’d stepped directly in it. I love what I do, but I feel like I have to work hard — hard — to produce something I can be proud of. And I couldn’t stand the idea of a visitor looking at my work, and just assuming I’d pressed a button to make it happen. ¹

My response to all this hand-wringing was an overwhelming desire to yeet. I launched the redesign, at least in part, to get it off my desk. And in what I’m sure future design historians will surely call a stunning display of timeliness and marketing savvy, I didn’t write about it anywhere until now. In that time, a little space has been helpful. I’ve been able to reconnect with the parts of the design I first loved, and just, like, enjoy looking at my website. The rough edges have even fallen away into the background. They’re still there, of course, but they feel less prominent, less sharp. They even feel like they have a little bit of promise, if that makes sense? They feel like future questions to explore — opportunities to refine, or to redesign.

In short, it feels good to have my worry stone back. I hope you like it, of course; I’m just happy to have it at a place where I like it. It feels a bit like coming home.

Thank you, as always, for reading.

This has been “The house is valuable because it is the house.” a post from Ethan’s journal.

Reply via email

2026-06-15 12:00:00

Let’s review.

Token costs are rising. In response, Meta and Amazon have shut down their internal “artificial intelligence” leaderboards; Microsoft has pushed its employees toward Copilot, and away from Claude Code’s rising costs. Uber has already capped its usage of LLMs internally, setting a monthly cap for every employee’s token spending. All of this is happening against the backdrop of the biggest “AI” platforms still borrowing money against hypothetical future returns, and markets are becoming increasingly skittish over America’s big, all-encompassing bet on LLMs. And according to reports from NOTUS and Reuters, major “AI” providers are already having discussions with the United States government, in which they’re laying the groundwork for “too big to fail”-style bailouts.

This is all fine, of course. Totally fine, and great.

I’ve noted before that the so-called “AI” industry is not built on a sound financial foundation. ¹ But in the last few weeks, it feels like new, even worse stress fractures have been racing up the walls. And it’s been hard not to wonder what will happen next.

What will happen next?

I must admit I feel cursed by this stupid question. Why do we even ask it? I mean, look: my endocrine system is evolutionarily hardwired to worry about unseen threats. Something buried deep in my subconscious still thinks alpha predators lurk around every street corner, waiting to snack on my kidneys. “Something terrible is about to happen,” my hindbrain insists. “Look at the evidence before you. Anticipate. Prepare for the unseen danger,” it whines. “Shut up,” I hiss back, “I am basically a fancy typist with bad eyesight, there are no cheetahs lurking in the tall grass, I do not need this much cortisol.”

But professionally? Working in tech, we’re expected to care about this question. We’re supposed to have educated opinions about the future, to have some idea of what will happen next, all so that we can better plan for it. To design and build something that’ll last.

And I don’t know about you, but whenever I’ve been asked that question — “what’s going to happen next?” — it has changed its shape. When I first started talking about responsive design, I’d get asked to weigh in on the future of mobile, or what the then-explosive levels of device diversity might look like in a few years’ time. A few years after that, I started getting questions about automated website builders, and how they might change design.

I can’t see the future, but I can state a blindingly obvious thing. None of these are questions about technology. These are all questions about work: what it might look like in the future and, most importantly, whether or not the person asking will even have a place in that future. That’s the real question behind the question, and it’s always tinged with worry or fear. And now, with hundreds of thousands of tech workers having lost their jobs since ChatGPT was released, there are a lot more of us asking the question. All of us would very, very, very much like to know what happens next.

I have no answers here. But two things come to mind.

Here’s the first one.

What will happen next?

When I’m asked about where “AI” might be going, I don’t think about the terrifyingly large bubble that’s (possibly? probably? definitely?) about to pop; I don’t think of the careers, lives, and futures that will be upended by that “market correction”; I don’t think about the potential after a crash for these big, lumbering, literally-world-devouring centralized platforms to be replaced by smaller local models. ² Instead, when I hear that question these days — what will happen next? — I think about William Gibson and plastics.

“And these guys were very common,” [Gibson] went on, taking down a small plastic spaceman: red, wearing an elaborately earmuffed helmet with an antenna on top. “These spacemen were dime-store toys at a time — which I can actually remember! — when cheap plastics were still weirdly novel. Like Gore-Tex or something. You’d ask, ‘What is it made of?’ ” He looked wistful, then thoughtful. “I’ve decided that one of the most significant things I ever saw in my life was the arrival of completely ubiquitous injection-molded plastics. I was certainly aware of them as the onset of something new. They cost practically nothing. But no one had any idea what a disaster we were all witnessing. Now the oceans are full of it.” He handed the spaceman to me. I hefted it, weightless, in my palm — an antique bit of misread future.

I’ve seen Gibson use this line in a few interviews. And when he does, he’s occasionally mentioned how he can browse through old photographs, and see all the toys he was raised on: crafted little objects made of cloth, wood, tin, even paper. But after a certain point, it’s like a switch gets flipped — and then he’s leafing through picture after picture of nothing but synthetic, injection-molded playthings.

I’ve been haunted by this story since the first time I read it. And it’s useful for me when I think about where we are now, and where we might go from here.

Just to be clear: I don’t think LLMs will be with us for long, at least not in their current form. But I do think that we’re living through a period of generational damage, one that will take years to repair, if not decades. For the last few years we have been watching a failed technology being deployed at scale: a technology built upon theft; upon inaccessible code; upon a flagrant disregard for work and for workers; upon ruinous environmental harms. After more than three years, “AI” has produced very little financial benefit outside of the largest companies hawking it. And in that time, after years of near-constant “AI” layoffs, I have heard and read so many accounts of people wondering if there’s still a place for them in the industry. ³ I personally have several friends who have already left the industry.

When I say the phrase “generational damage,” I mean it.

So, no: I don’t know precisely what will happen next. But I’ve come to realize the outcomes matter a lot less than the harms we’re watching unfold right now. Whether the bubble pops or doesn’t, I feel confident we’ll look back at the first few years of “artificial intelligence” much as Gibson does a toy astronaut: as a herald we ignored, as an antique bit of misread future.

Second thing. And it’s about the question itself.

What will happen next?

I’ve been thinking about this question, like, structurally. Isn’t the phrasing a bit…passive? No judgment intended at all here, mind. I’m just struck by the fact that it’s posed from a place of observation. The person asking the question is an onlooker, a spectator; they’re trying to anticipate how events in front of us might be unfolding, and guess where the river might flow next. And when we ask ourselves what might come next — for “AI,” for the industry, for the future of our work within both — I bet we could start asking another question.

Maybe it’s this one:

What do you want to happen next?

Halfway through my unions book, I sat down with Jacky about some of the organizing conversations he’d participated in.

We were talking about some brainstorming sessions he’d been in at both companies, in which he and his coworkers were dreaming up ideas for what they might include in [their union contract] — ideas people had thrown around for what their contracts could look like. Things like, “What if our parental leave policy lasted five years?” And: “Could our contract have mandated that 20 percent of every sprint was spent on tech debt?”

Alciné hastened to note that none of these proposals ever appeared at a bargaining table; they were just ideas that came up in brainstorming sessions. But that’s what I think is so valuable about the collective bargaining process, and the discussions around them: that these discussions are meant to be generative.

This was the common thread that ran through every single one of the interviews I conducted for the book. Amid layoffs, intimidation, the sheer difficulty of organizing, every person I spoke with touched on just how the process opened up new possibilities for them. A new future — and it was a future they wanted to fight for.

Amid all this talk about the inevitability of “AI”, I think it’s okay for us to ask what kind of future we want, and then move toward it together. And it’s already happening across the industry. ProPublica’s guild conducted a strike earlier this year, in part to win contract language that would prohibit layoffs resulting from “AI” adoption. UK workers at DeepMind, Google’s AI Research Lab, voted to unionize, in part to block usage of their employer’s models in military contracts. Thousands of tech workers in the University of California system voted to unionize, and gained the right to bargain over the use of “AI” tools in the workplace. DAIR has released a hub filled with resources for people looking to push back against “AI” and automation at work.

I think we can figure out our future together, right now. And as Mandy reminds us, it all starts with conversation. We have to talk with our friends, colleagues, and coworkers. We have to talk about our concerns, and what we wish were different. We have to map out how we’ll collectively instrument change in our workplaces, and in our industries.

I want to fix this industry. I want you to have a place in it. I want us to have a place in it. Maybe you do, too.

So, really: what do you want to happen next? I’ve got some ideas, but I’d love to hear yours.

This has been “All tomorrow’s parties.” a post from Ethan’s journal.

Reply via email

2026-04-21 12:00:00

I’m giving a talk at next week’s Boston WordPress meetup! If you’re around on Monday night — that’s 27 April 2026, for you Julian calendar sickos — I hope you’ll be able to join us. I’ll be giving a short talk on design systems: how they’ve changed the way digital teams work, and how we might need to rethink them. It’s a fun talk, and I’m excited to be there.

I’ve been rehearsing the talk this week, and was struck by…well, how weird it felt to rehearse a talk. That’s when I realized that next week’s meetup will be the first in-person talk I’ve given since 2019. Twenty-nineteen! Seven entire years! An eternity ago. Public speaking used to be, like, a whole entire thing I used to do. And now it’s just…not. I’ve given plenty of remote talks since then, mind you. (And they’re still plenty of work!) But as much fun as a virtual event is, I’ve missed getting a chance to meet attendees, hear what they’re working on, and what they’re excited about.

A lot of this is due to a changing world. The amount of travel I did fell off sharply back in 2020, and my house acquired some new health concerns back in 2021 that’ve made it harder to attend public events amid an ongoing pandemic. And the industry hasn’t escaped unscathed, either: many of the conferences I used to speak at simply don’t exist anymore.

This is all to say that amid all that change — and my nerves — I’m so glad to see that local, community-focused events are still going strong. And I’m honored to be in the room at the next Boston WordPress meetup. I’ll be masked up, sure, and probably a little more rusty than I used to be. But if you’re able to come, I hope you’ll say hi. It’d be great to meet you.

This has been “Retour.” a post from Ethan’s journal.

Reply via email

2026-03-04 12:00:00

Amid everything happening right now, I read with great interest about the technologies used in planning our latest horrid war. According to The Wall Street Journal, Anthropic’s Claude was used in “intelligence assessments, target identification and simulating battle scenarios.” This was all happening in the midst of a lot of sturm and drang between Anthropic and the federal government, during which OpenAI inked a new deal with the United States military.

A few hours before reading all of this, I’d learned of the airstrike on a girls’ school in southern Iran, a bombing that killed nearly two hundred people. Many of those killed were students attending classes.

I don’t have the words to describe how I feel about this.

But I will say one thing.

I am on the record as saying “artificial intelligence” is a failed technology, and I stand by that assessment. These platforms do not work as advertised; features that use them are loathed by the public; they exact terrible mortgages against our future; and they likely wouldn’t even exist were it not for the staggering amounts of capital being poured into them.

It’s that last point I want to focus on for a moment. To put it mildly, the “AI” industry is not built on a sound foundation. Investors are spooked by the staggering amount of debt underpinning the industry’s obscene growth, especially since none of the flagship names care about profitability. And in their desperate search for capital to shore up their ballooning valuations, these companies have turned toward governments, and the nigh-infinite supply of tax dollars they possess. The largest “AI” providers are positively sprinting to enter into procurement agreements with governments — and, more specifically, with the parts of government responsible for violence and war.

Time and again, these organizations have contorted their corporate values to support that sprint. Sarah Shoker, a former lead of OpenAI’s geopolitics team, has written more eloquently than I could about the incoherent policies at these companies around “acceptable” military usage. She notes that OpenAI quietly deleted its outright ban on military usage back in 2024; she also notes that, according to some reporting by Bloomberg, OpenAI has recently agreed to contribute its models for a drone trial.

As it turns out, that “growth at any cost” mindset does have an incredibly high cost. Because what all this corporate flexibility means is that — in addition to all their other many failings — these “AI” platforms are machines of war and death. Anyone infuriated by the military-industrial complex should stop supporting them.

And I don’t just mean monetarily. It’s not lost on me that these commercial “AI” platforms improve by learning. Well, “learning” is too anthropomorphic a verb: these large language models analyze user input, and may be trained on the data we feed them. And as I’ve read about the opening days of this pointless, stupid, awful, horrifying war, I couldn’t help but think that the everyday use of these machines is quite literally enhancing them.

None of this is new, of course. Microsoft’s “AI” products were used by Israel’s military to surveil Palestinians; the 2018 Google Walkouts were driven in part by employees protesting Google’s machine learning contributions to drone strikes; during the height of the Vietnam War, tech workers were mobilizing against the industry’s embrace of defense contracts. But today, we cannot meaningfully separate the everyday use of “AI” platforms from their application in death and war. Imagine asking Northrop Grumman to summarize an email for you, or refactoring code with help from Boeing and Raytheon.

These tools, and the companies that manufacture them, have tremendous costs — to our labor, to our environment, to our futures. And as we’ve been seeing, those costs also include actual human lives.

This has been “Propellant.” a post from Ethan’s journal.

Reply via email