Links, June 13, 2025

§Stop asking me if I exist

CJ Chilvers writes a newsletter I have been continuously subscribed to for over 12 years. He echoes a sentiment I feel often:

Sometimes I wonder “what ever happened to…[creator who unsubbed me]?” That’s not a place you ever want a reader to be.

Like CJ, I too use mostly Apple products for my email, because they emphasize privacy, masking a lot of the intrusive tooling people use to measure email “engagement”. I’ve had online relationships turn sour over this: they consider any mail client which doesn’t give their tracking pixels everything they ask for to be immoral.

Look, it’s the same thing as with browser privacy protections: My willingness to give you attention is not an invitation to intrusion.

§Quarkdown

Have you ever wished Markdown were Turing-Complete?

I’m torn between horror, and thinking I should move the custom markdown-processing chain I have for this website to this.

§The Future of the Software Industry

Written two years ago, this feels prescient:

The software business has, despite dotcom busts and various other financial crises, only ever experienced good times: cheap energy, rising empire and - for more than thirty years - unipolar US hegemony. It’s hardly a high stakes gamble to suggest that businesses will do worse in times of economic decline, but the dynamics currently affecting our industry are completely new to it and us.

§good internet (a magazine)

It lives up to the name. I came to the site via My website is ugly because I made it, and enjoyed a number of other recent pieces such as Welcome to the web we lost, Webweaving and blogging can change your life and Can accessibility be whimsical?. They have a free subscription tier, which seems to be a mailing list, but they also have an RSS feed.

You can’t form a theory of mind for something that doesn’t have a mind

§The Gap Through Which We Praise the Machine

Fred Hebert on the form of LLMs and why chat in particular is a bad interface. There are so many great quotable excerpts but none really summarize the piece well. Nevertheless,

In a fundamental sense, LLMs can be assumed to be there to impress you. Their general focus on anthropomorphic interfaces—just have a chat!—makes them charming, misguides us into attributing more agency and intelligence than they have, which makes it even more challenging for people to control or use them predictably. Sycophancy is one of the many challenges here, for example.

This is a piece not so much about the ethics and drive to push the technology itself (and says so up-front), but rather the process of training oneself on using this technology to begin with that blinds people who say things like “It can’t be that stupid, you just need to prompt it better” to their own invisible scaffolding. The main thesis is that while the technology may offer some value, tools are, by design, poorly designed: chat is meant to be “perceived as the engine of a new industrial revolution”, to replace workers, not augment their own abilities.

§The librarian immediately attempts to sell you a vuvuzela

You pick out a book from the shelf, it appears to be named “History of the Piano.” You leaf through some pages. Oddly, there doesn’t appear to be any real content, just references to piano training classes you can buy.

A metaphor for the modern internet, which gets into a reason why LLMs have perhaps supplanted search, and why also they are not – as currently financed – a viable long-term replacement. The piece is however overall net-positive on the technology.

§I Think I’m Done Thinking About genAI For Now

This piece by @glyph touches on the aforementioned issues of chat and mental models, but also gets further into the aesthetics and affordances of the technology, and from there fundamental questions about the interplay of how technology & society shape each other, and it does get into ethics.

It gets most things right, but it consistently makes mistakes in the places that you are least likely to notice. In places where a person wouldn’t make a mistake. Your brain keeps trying to develop a theory of mind to predict its behavior but there’s no mind there, so it always behaves infuriatingly randomly.

§The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity

A great paper from Apple with a refreshingly readable tone for something so academic. It shows clearly that LLMs are no substitute for well-specified conventional algorithms, and because the chat interface, you may fool yourself into thinking it has created a generalizable solution to a specific problem when it hasn’t.

These insights challenge prevailing assumptions about LRM capabilities and suggest that current approaches may be encountering fundamental barriers to generalizable reasoning.

I suspect the timing is mean to coincide with WWDC, but I find it funny that this drops after the announcement that Jony Ive (whose work is I believe vastly overhyped) joins up with OpenAI.