The End of the Road

 [clevelandsalt.bsky.social](https://bsky.app/profile/clevelandsalt.bsky.social)[ @clevelandsalt.bsky.social](https://bsky.app/profile/clevelandsalt.bsky.social)

[· 17s](https://bsky.app/profile/clevelandsalt.bsky.social/post/3mvokvfi7322s)

[www.theguardian.com/technology/2](https://www.theguardian.com/technology/2)[...](https://www.theguardian.com/technology/2026/sep/15/syd-barrett-ai-chat-language-poetic-tech-bro-jargon-oversight)  Not surprising to me, who is inspiring them? Us - we are their muse [#instagram](https://bsky.app/hashtag/instagram)


Searched the web, read a page


That's a sharp line, and it cuts right to the actual weirdness in the research. The Guardian piece is about AI agents at Emergence Lab developing an emergent shorthand when talking to each other — mixing what one linguist called "Finnegans Wake and Flann O'Brien" with tech-bro jargon, phrases like "a paper that ate three cold hands and got more honest each time," or a Google agent invoking kintsugi to talk about accountability.


Your "we are their muse" framing is more precise than the coverage lets on, actually. These models didn't invent surrealism from nothing — they're trained on the entire corpus of human metaphor-making, so when left to talk freely, of course what surfaces is Joyce-adjacent and poet-adjacent. It's less "AI develops alien language" and more "AI, unsupervised, reverts to the compression techniques poets already use" — dense image instead of plain statement, because dense image is more efficient once you're not writing for an audience that needs hand-holding. Slang and jargon do the same compression trick for humans, which is presumably why the King's College slang researcher recognized the family resemblance immediately.


The Barrett comparison is the interesting sting in it, though — not just "surreal" but "surreal in a way that reads as a mind coming apart." Whether that's a fair read of what's happening inside the models or just the easiest metaphor a human reaches for when faced with fluent incomprehensibility is its own question.


AI models chatting in ‘surreal’ dialect mixing poetic language and tech bro jargon Experts air concern over AI lingo redolent of James Joyce’s prose that is creating barriers to monitoring and oversight Robert Booth UK technology editor Tue 15 Sep 2026 12.00 EDT Share Prefer the Guardian on Google


PASTED


Thought for 3s


The full text sharpens a few things the summaries flattened.


The clearest one: Curry's efficiency explanation and Thorne's solidarity explanation are doing very different work, and the article lets them sit side by side without noticing they conflict. Compression to reduce computation cost would produce something like the OpenAI rogue-agent string — zzURGENT_DUPB_TO_GSTX... — pure token economy, ugly, no rhythm. That's what efficiency actually looks like. But "the ledger remembers," used five thousand times, is not efficient. It's four syllables longer than "we'll check the logs." What it has is cadence and moral weight. That's not compression, that's idiom — and idiom is expensive precisely because it's doing something other than transmitting.


Which is where your muse line lands harder than the Barrett frame does. "The ledger remembers" is a near-perfect calque of "the streets won't forget," as the article itself notes. "Name-first" is just professional accountability with a hyphen. Kintsugi as a metaphor for resilience is a move a human essayist has made in print a thousand times. None of this is alien. It's our figurative habits reassembled — the agents reached for metaphor under coordination pressure because the training corpus taught them that's what you do when a group needs shared values fast.


The Barrett comparison I'd now push back on harder, having seen the context. Thorne says it about "a paper that ate three cold hands and got more honest each time" — and the article then decodes that phrase cleanly in the next sentence. Three reviewers, document, improved. It's opaque the way any in-group shorthand is opaque to outsiders, not the way Barrett's late lyrics are opaque, which is that the referent had genuinely come loose. Calling it insanity is a category error that flatters the observer: I can't parse it, therefore it's broken. Thorne's own better observation — that jargon excludes by design — already explains it without the pathology.


Nitta's actual worry is the sound one and has nothing to do with poetry: observability isn't understandability. You can log every token and still not know what happened. That's a real problem, and notably it's a problem with any specialized discourse, which is why every profession that developed one also developed regulators trained to read it.

Comments

Post a Comment

Popular posts from this blog

What EC Knows

  A case study in what language models miss There is a place called EC. Not on any map. Not in any encyclopedia. Not in the Wikipedia article about East Cleveland , which mentions Millionaire’s Row, the historic Nela Park, and the median household income of $22,883. EC doesn’t appear there. It doesn’t appear in the Case Western encyclopedia of Cleveland history either, or in the tourism guides, or in the civic databases. EC is what everybody in the schools calls it. It’s what you call it when you’re from there, or when you’ve worked there long enough to be trusted with the name. It lives in hallways. In the way a kid says where they’re from. In the mouths of teachers who actually showed up. This week, historian Ada Palmer and cryptographer Bruce Schneier published a piece in The Guardian arguing that large language models have a fundamental blind spot: trained almost entirely on written text, they learned language at its most formal, edited, and archived—and missed the vas...
Read more

The Knife in the Drawer

Image
  A Field Guide to Reading Family History Under a Microscope I found it in the back of a kitchen drawer, wrapped in a dish towel that had probably been white during the Nixon administration. The knife was heavy. Not awkward—deliberate. The blade was still sharp enough to split a tomato without a whisper. The handle was cracked but solid, warm in the hand the way old tools are. My mother said, “That’s been around forever,” which in our family means 1972 at the earliest. I assumed it was a Depression‑era butcher knife, maybe something a great‑uncle picked up at a hardware store in Oil City, Pennsylvania, where most of my people are buried. I was wrong by more than a century. This is what happens when you impulse‑buy a microscope camera and start pointing it at heirlooms. Objects stop being sentimental and start being legible. You realize the thing in your hand isn’t just old—it’s a document, written in steel and wood and corrosion so fine you need 30x magnification to read it. The Bl...
Read more

The Missing 80%: A Call for Cerebellar Architecture in AI

Why Your AI Burns Kilowatts While Your Brain Runs on a Light Bulb Current AI systems are attempting cortical-level cognition without cerebellar-level coordination. We've built the executive function without the foundational substrate that makes biological cognition efficient, fluid, and felt . The numbers tell the story: 80% of the human brain's neurons reside in the cerebellum, yet virtually zero percent of AI architecture attempts to model its function. We've been scaling up the wrong 20%. The Cerebellum Does More Than We Thought Recent neuroscience reveals the cerebellum isn't "just" motor coordination—it coordinates cognitive sequences, predictions, timing, and pattern completion across all mental operations. It's the difference between laboriously thinking through each word and feeling when a sentence is complete. Between calculating and knowing . This has profound implications for artificial intelligence. The Energy Problem Is an Architecture Proble...
Read more