Jeremy Keith

I have little understanding for people using large language models to generate slop; words and images that nobody asked for.

I have more understanding for people using large language models to generate code. Code isn’t the thing in the same way that words or images are; code is the thing that gets you to the thing.

And if a large language model hallucinates some code, you’ll find out soon enough:

With code you get a powerful form of fact checking for free. Run the code, see if it works.

But I want to push back on one justification I see repeatedly about using large language models to write code. Here’s Craig:

There are many moral and ethical issues with using LLMs, but building software feels like one of the few truly ethically “clean”(er) uses (trained on open source code, etc.)

That’s not how this works. Yes, the large language models are trained on lots of code (most of it open source), but they’re not only trained on that. That’s on top of everything else; all the stolen books, all the unpaid creative work of others.

Even Robin Sloan, who first says:

I think the case of code is especially clear, and, for me, basically settled.

…goes on to acknowledge:

But, again, it’s important to say: the code only works because of Everything. Take that data away, train a model using GitHub alone, and you’ll get a far less useful tool.

When large language models are trained on domain-specific data, it’s always in addition to the mahoosive amount of content they’ve already stolen. It’s that mohoosive amount of content—not the domain-specific data—that enables them to parse your instructions.

(Note that I’m being very delibarate in saying “parse”, not “understand.” Though make no mistake, I’m astonished at how good these tools are at parsing instructions. I say that as someone who tried to write natural language parsers for text-only adventure games back in the 1980s.)

So, sure, go ahead and use large language models to write code. But don’t fool yourself into thinking that it’s somehow ethical.

What I said here applies to code too:

If you’re going to use generative tools powered by large language models, don’t pretend you don’t know how your sausage is made.

An Entirely Other Day: The Triumph of Triumphalism

Scratch the skin of wild-eyed AI proponents, and a thick syrup oozes out, made up of the blendered remains of Roko’s Basilisk, barely sublimated Christian end-times thinking, and the mis-remembered plot of that one cool science-fiction story they read when they were twelve. This is the basis for the new order, just like the blockchain was a couple of years ago, and a dead-eyed, low-poly, pantsless rendering of Mark Zuckerberg was a couple of years before that.

“You’re going to be left behind” is only the latest version of “Have fun staying poor.” It’s got every ounce of the smug self-satisfaction that it shouldn’t need if the inevitability it promises were actually inevitable.

“AI-first” is the new Return To Office - Anil Dash

AI is really good for helping you if you’re bad at something, or at least below average. But it’s probably not the right tool if you’re great at something. So why would these CEOs be saying, almost all using the exact same phrasing, that everyone at their companies should be using these tools? Do the think their employees are all bad at their jobs?

Pluralistic: The enshittification of tech jobs (27 Apr 2025) – Pluralistic: Daily links from Cory Doctorow

The point of AI isn’t to make workers more productive, it’s to make them weaker when they bargain with their bosses.

Bias in Design Systems - bencallahan.com

Thoughtful analysis from Ben (as always).

What we talk about when we talk about AI — Careful Industries

Technically, AI is a field of computer science that uses advanced methods of computing.

Socially, AI is a set of extractive tools used to concentrate power and wealth.

In exactly six weeks time, UX London is happening!

I am ridiculously excited about this year’s line-up—I can’t wait to see the talks and get hands-on in the workshops.

If you haven’t yet got your ticket, now is the time. There’s a flash sale this week: use the discount code FLASH20 to get a whopping 20% of any ticket. Do it before the end of Friday!

Whether you’re coming for all three days or choosing one focused day, you’re in for a treat.

• Day one on Tuesday, 10 June is discovery day.
• Day two on Wednesday, 11 June is design day.
• Day three on Thursday, 12 June is deliver day.

Head on over to the website to get all the details and then get your discounted ticket.

See you there!

Are you looking for a book that explains why you should care about web accessibility?

This is not that book.

Manuel Matuzović has too much respect for your intelligence to waste time trying to convince you of something you already know. You already know that web accessibility is important.

What you really need is a guidebook, a handy companion to show you the way through a tangled landscape.

This is that book.

If you want, you can read it cover to cover. That’s what I did, and I enjoyed every moment of the journey.

But you might not have time for that. That’s okay. The way that this book is subdivided means you can deep into any chapter and it will make perfect sense by itself. It really is like a cookbook. Every chapter is like a standalone recipe.

Whether you read this book linearly or dip in and out of it is up to you. Either way, Manuel is going to massage your brain until something new takes shape in there. An understanding. Not just an understanding of web accessibility, but of the very building blocks of the web itself.

See, that’s the sneaky trick that Manuel has managed with this book. It’s supposed to be an accessibility cookbook but it’s also one of the best HTML tutorials you’ll ever find. Come for the accessibility recipe; stay for the deep understanding of markup.

Best of all, Manuel manages to do all this without wasting a word. Again, he has too much respect for you to waste your time. The only unnecessary words in your Accessibility Cookbook are the ones you’re reading now. So I’m going to follow Manuel’s example, respect your time, and let you explore this magnificent book for yourself.

Enjoy the journey!

Polishing your typography with line height units | WebKit

I should be using the lh and rlh units more enough—they’re supported across the board!

The Hidden Cost of AI Coding – Terrible Software

Feels like an emerging trend:

Instead of that deep immersion where I’d craft each function, I’m now more like a curator? I describe what I want, evaluate what the AI gives me, tweak the prompts, and iterate. It’s efficient, yes. Revolutionary, even. But something essential feels missing — that state of flow where time vanishes and you’re completely absorbed in creation. If this becomes the dominant workflow across teams, do we risk an industry full of highly productive yet strangely detached developers?

But what if I really want a faster horse? | exotext

Overall, consistency, user control, and actual UX innovation are in decline. Everything is converging on TikTok—which is basically TV with infinite channels. You don’t control anything except the channel switch. It’s like Carcinisation, a form of convergent evolution where unrelated crustaceans all evolve into something vaguely crab-shaped.

A Web Component UI library for people who love HTML | Go Make Things

I’m obviously biased, but I like the sound of what Chris is doing to create a library of HTML web components.