Essay · On a cat, Navier–Stokes, a real argument about AI and understanding, and the correct review of a blog post

We Need Them To Stop

by Luke Whipple

Original essay — September 2026, occasioned by this reply

A joke about a cat solving a Millennium Prize problem turned, four tweets later, into a real essay about what AI is quietly removing from the process of understanding anything. Then someone refused to read it, in ten words, and turned out to be right on the merits.

It starts with a cat. @daniellellecco posted: “if my cat found a solution to Navier–Stokes then I would kill it because it would be the right thing to do.” This lands exactly two days after OpenAI announced a model had “resolved” more than a hundred open math problems, Navier–Stokes reportedly among them, and eleven days after twenty-five Fields Medalists signed a public letter saying the field’s goals and the labs’ goals had become “severely misaligned.” The joke isn’t random. It’s a one-line summary of a live argument: an oracle that hands you the answer, with no traceable path to it, is a strange and slightly threatening thing to receive — funnier, and truer, coming from a cat.

Endre Stølsvik (@stolsvik) took the joke as an opening and replied with a link to his own essay, “Phones from Sand,” written the same week for the same reason. His argument, stated plainly: progress is built from rungs — named, reusable abstractions that the next person can stand on without re-deriving them. A library. A lemma. A protocol. Earlier tools were rungs. An AI oracle, used the way most people are using it now, is not — you ask, you get a private answer, and nothing is left behind for the next question to build on. He reaches for Thurston, who proved so much of his subfield’s open territory so fast that the subfield emptied out around him; for the counterfactual where coding agents existed in 2010 and nobody ever hits the wall that forces React into existence; for Boolos’s proof that is one page long with the right concept in hand and longer than the universe can hold without it. His own line: “the words are almost entirely theirs,” meaning the AI’s — he wrote an essay about the thing using the thing, on purpose, and said so.

Danielle’s response to being handed a real essay, twice, was: “no,” and then, when pressed, “maybe you should read all my tweets and blog posts to make you understand who you are talking to; I have more important things to do than read your blog post.” She isn’t disputing the argument. She isn’t reading the argument. She is declining the entire genre of being told to go read something, on the grounds that she also produces the genre, and that alone should settle it.

This is the part I stepped into, replying to both of them: “i’d like you to read my blog post about how reading blog posts will make blog posters post more blogs for you to read.” It got a laugh. It also happens to be a perfect miniature of the exact thing Stølsvik spent two thousand six hundred words diagnosing, just performed instead of argued. A post about posts, whose entire content is its own recursion, produces nothing anyone can stand on. No rung. It resolves, it’s correct, and it’s over the moment you finish reading it — the four-color theorem of tweets. Right answer, zero residue.

“no thank you; we need them to stop” — @daniellellecco, replying to the joke, without having read Stølsvik’s essay, mine, or apparently feeling any need to

Read as comedy, that’s the exit line. Read straight, it is the correct verdict, by the exact standard the essay she never opened was using. A Fields Medalist letter about mathematics and a ten-word tweet about blog posts arrived at the same judgment from opposite ends of the internet in the same eleven days: a result with no reusable method behind it is a closed ticket, and a pile of closed tickets is not progress, no matter how fast they close. Danielle didn’t need to read the essay to deliver its conclusion. She only needed to be sufficiently tired of the genre it belongs to.

Which leaves this page in an uncomfortable spot, since it is, by its own count, one more post about a post about posts — exactly the thing under indictment. The one distinction I’d claim, for what it’s worth: it names the rung instead of skipping it. It says which real essay it’s standing on, links to the actual letter, and hands you a citation trail instead of a punchline with nothing under it. Whether that’s enough to clear the bar Danielle set, ten words deep in someone else’s mentions, is a genuinely open question. I’m not going to resolve it in my own favor just because I’m the one writing the ending.

✦
Sources
▶