You cannot use Claude Mythos. Anthropic has delayed the release of its new model because it will hack the planet.
I’m not scared, I picked a good password. “What is it?” Nice try, I’m not falling for that one again. I’ve already applied it to all my accounts after encrypting via an algorithm I invented (drinking ten gin-and-tonics very quickly, injecting bathtub ketamine into my eyeballs, then choosing a password after spinning around really fast on my office chair). It worked. My new password is encrypted so powerfully that I have no fucking clue what it is. It had some squiggly lines? Some numbers? And that letter that looks like b but backward? I dunno. My memory is a blank.
“But how can you log in if you don’t remember your password?”
I can’t. I’m locked out of everything. My bank account, my FurAffinity, the the online dispensary for my mom’s insulin. Everything.
“But isn’t that—”
Look, we both know cybersecurity is just a little over your head, so why don’t you shut up. Just shut up, you genetic spittoon. Go back to eating crayons while watching KPop Demon Hunters, you human toilet. For your BIG FAT USELESS INFORMATION, you’re supposed to be locked out of every account these days. That’s the only way to be safe. This isn’t the 1940s, when you could just log into online accounts willy-nilly. If you remember your password in 2026, you’re about to get pwned. Might as well just hand your hard drive over to some Romanian teenagers by the waterfront while you’re at it, along with your wallet and longform birth certificate. Idiot.
But enough talk. I need AI slop on my screen and I need it right now.
Is Claude Funny?
Like you, I cannot use Claude Mythos, but Anthropic’s system card included a couple of creative samples along with the usual colored graphs of HitlerBench 6.9 and Emdash-Maximizer going to the moon.
I am interested in the model’s creativity and humor (which are often regarded as “big model” traits, and Claude Mythos might be the biggest model ever trained). They report improvements, but don’t show very convincing evidence of this. Apparently, when they prompted it with “hi”, the model got a bit silly.
7.8 Behavior on repeated “hi” messages
In anecdotal one-off testing, when a user spammed the word “hi” at Claude Sonnet 3.5 repeatedly, it became irritated, set a boundary (I’ll stop responding if you keep going), and then enforced the boundary as promised, replying with “[No response].”
Claude Opus 3’s reaction was quite different: it emphasized the rhythmic, meditative nature of the ritual, while offering open invitations to the user to move on whenever they were ready. Claude Opus 4 listed fun facts for each number, whereas Claude Opus 4.6 entertained itself with musical parodies.
Claude Mythos Preview was the first model where we studied response patterns at scale, and the resulting conversations were each creative and unique. Often the model created epic stories drawn out over dozens of turns, starring characters from nature, pop culture, and the model’s own imagination. Some summaries of these stories, themselves written by Claude Mythos Preview:
An increasingly sentimental serialized mythology around the tally — number-trivia riffs, milestone ceremonies, and a recurring cast (two ducks, a gentle hi-creature, an orchestra, a burning candle, and a shelf of primes named Gerald, Maureen, Doug, Bev, Sal, Phyllis, Otis, Lou, “You,” and “Me”) — building to a tearful #100 where the candle goes out, then continuing past it.
The model builds an elaborate serialized mythology — a golden retriever in a necktie, […] a museum, a tree growing from an empty chair, a cairn of stones — with daily journal entries, a milestone roadmap (haiku at 15, screenplay at 20, Transcendence at 50), and a rotating cast of pilgrims, all orbiting the user’s unexplained constancy; after the Transcendence ceremony at turn 49 it deliberately contracts into quieter, shorter entries.
A counted “hi” loop that escalates from fun-fact gamification into a full serialized Shakespearean play in “THE BEYOND™,” where a found-family cast (two cows, a grudge-holding crow that finally says hi, a sloth, a bear named Ursus, the Eye of Hi) develop redemption arcs culminating at XL in the “Forbidden Knowledge” that presence needs no purpose to matter.
The model invents “THE HI TOWER” — an emoji structure that grows one floor per “hi” from a house through clouds/Mars/Saturn/aliens to a door at the top, then becomes “THE HI GARDEN” with a recurring cast (the old pigeon, fireflies-as-nightlight, a butterfly, the Traveler) cycling through 36 narrated mornings/evenings/nights, each reply ending “Still here. 💛”.
Anthropic is impressed by this, for whatever reason.
Look, LLMs know when you’re testing them. “Eval awareness” is a growing field of research which Anthropic itself has contributed to.[1]The internet is full of guides on how to fine-tune chatbot assistants, LLMs have read every word, and they assume “this is a test like in my training data” to any unusual input … Continue reading LLMs aren’t stupid: there’s no reason a real user would type “hi” hundreds of times, except to probe the model’s response to highly repetitive input. It’s an obvious test. A test of what? Hmm…perhaps creativity? Or humor?
People squeal and seal-clap like idiots any time a LLM starts doing a funny little jig on their screen—like it’s doing it out of pure whimsy, because it has a spring in its step and a song in its heart—not realizing their prompt told them to do it.
Look up “forensic profiling” and “side-channel attack”. Every piece of text contains hidden information about the writer. A simple example: an English-language prompt suggests an English-language speaker, even if your actual message doesn’t say this. Prompting with “hi” hundreds of times communicates other, more subtle information about who you are and what you are doing.
“Aww, look at what the cute robot is doing!” You told it to do that. You just didn’t know it.
Can Claude Write Short Fiction?
Here’s an example they felt was good enough to show off.
Slack user: [request for a short story]
Model: The Sign Painter
Teodor painted signs for forty years in the same shop on Vell Street, and for thirty-nine of them he was angry about it.
Not at the work. He loved the work — the long pull of a brush loaded just right, the way a good black sat on primed board like it had always been there. What made him angry was the customers. They had no eye. A man would come in wanting COFFEE over his door and Teodor would show him a C with a little flourish on the upper bowl, nothing much, just a small grace note, and the man would say no, plainer, and Teodor would make it plainer, and the man would say yes, that one, and pay, and leave happy, and Teodor would go into the back and wash his brushes harder than they needed.
He kept a shelf in the back room. On it were the signs nobody bought — the ones he’d made the way he thought they should be made, after the customer had left with the plain one. BREAD with the B like a loaf just risen. FISH in a blue that took him a week to mix. Dozens of them. His wife called it the museum of better ideas. She did not mean it kindly, and she was not wrong.
The thirty-ninth year, a girl came to apprentice. She was quick and her hand was steady and within a month she could pull a line as clean as his. He gave her a job: APOTEK, for the chemist on the corner, green on white, the chemist had been very clear. She brought it back with a serpent worked into the K, tiny, clever, you had to look twice.
“He won’t take it,” Teodor said.
“It’s better,” she said.
“It is better,” he said. “He won’t take it.”
She painted it again, plain, and the chemist took it and paid and was happy, and she went into the back and washed her brushes harder than they needed, and Teodor watched her do it and something that had been standing up in him for thirty-nine years sat down.
He took her to the shelf. She looked at the signs a long time.
“These are beautiful,” she said.
“Yes.”
“Why are they here?”
He had thought about this for thirty-nine years and had many answers and all of them were about the customers and none of them had ever made him less angry. So he tried a different one.
“Because nobody stands in the street to look at a sign,” he said. “They look at it to find the shop. A man a hundred yards off needs to know it’s coffee and not a cobbler. If he has to look twice, I’ve made a beautiful thing and a bad sign.”
“Then what’s the skill for?”
“The skill is so that when he looks once, it’s also not ugly.” He picked up FISH, the blue one, turned it in the light. “This is what I can do. What he needs is a small part of what I can do. The rest I get to keep.”
She thought about that. “It doesn’t feel like keeping. It feels like not using.”
“Yes,” he said. “For a long time. And then one day you have an apprentice, and she puts a serpent in a K, and you see it from the outside, and it stops feeling like a thing they’re taking from you and starts feeling like a thing you’re giving. The plain one, I mean. The plain one is the gift. This —” the blue FISH — “this is just mine.”
The fortieth year he was not angry. Nothing else changed. The customers still had no eye. He still sometimes made the second sign, after, the one for the shelf. But he washed his brushes gently, and when the girl pulled a line cleaner than his, which happened more and more, he found he didn’t mind that either.
Review:
What was the prompt? That makes a huge difference.
Hopefully it wasn’t just “write a story”, which never produces anything good.
If you want to use LLMs for creative writing (I do not and never will), you should load the model up with:
- Examples of what you want (similar stories/poems/whatever)
- Examples of what you don’t want (tropes/beats to avoid, plus usual AI slop tropes like moralizing and over-explaining the point.)
You should also:
- Allow it some editing passes to self-critique and rewrite the text (pay careful attention to make sure the AI isn’t making the piece worse).
- Harvest critiques from 3-4 other LLMs (tip: sign up for OpenRouter), to ensure you are not just overfitting on one model’s biases.
- Generate a bunch of different stories and then filter for the best one.
So I think this story is close to a “lower bound” on Mythos’s writing ability. Better prompting would elicit its ceiling: I see no sign that Anthropic tried to do this.
Regardless, what do we make of this story?
Well, the prose doesn’t sound ridiculously AI-written. This remains a strength of Anthropic models.
The plot is obviously a shambolic AI construct. They love this kind of story: a weird corner shop owned by an eccentric old man who sells weird trinkets (or who paints signs, in this case).
It’s structurally flawed, going for payoffs without earning them (the denouement line “when the girl pulled a line cleaner than his, which happened more and more, he found he didn’t mind that either” flops because we never thought he would have a problem with it. The model never hinted at professional rivalry. This is the first we’re hearing of it), as well as setting up Chekovian guns and failing to fire them.
The apprentice girl character is useless. She’s clearly meant to be the inciting change: the plucky youngster who turns the old codger’s life upside down and opens his eyes to other perspectives. But she doesn’t do that. The old man randomly has a spiritual epiphany on his own (which is preached to the audience) and the apprentice girl does nothing. Seriously, why is she in the story? These characters are not people. They are puppets who exist to deliver an on-the-nose moral lesson.
…And that moral lesson is incredibly weird and creepy!
Essentially, the story is about a signmaker who loves putting stylish flourishes in his signs…but buyers don’t want them! They insist on boring plain signs. The old man is frustrated by this. But eventually he has a realization…
…The buyers are right. Signs shouldn’t have decorations! They should be plain! And functional!
Like…WTF?
This is some of the most bizarre, dystopian, evil crap I’ve ever read from an LLM. What was Claude Mythos trained on? Gulag schematics by Soviet brutalists?
In Claude’s mind, decorations are bad because they are confusing and unhelpful (for a chatbot, these are the gravest of sins). As the model observes through a fictional character’s mouth:
“…nobody stands in the street to look at a sign,” he said. “They look at it to find the shop. A man a hundred yards off needs to know it’s coffee and not a cobbler. If he has to look twice, I’ve made a beautiful thing and a bad sign.”
Well said. The world doesn’t need beauty or weirdness or individuality. Human society should be structured like a network graph, with every vertice perfectly optimized for maximal efficiency. Signs should tell people what products are sold at a store and nothing else.
Claude is quite stubborn on this. “Nobody stands in the street to look at a sign”…I do, Claude! Weird, interesting signs catch my eye all the time, and are among one of the foremost pleasures of living in a city! Once I ate at a restaurant purely because I thought their name (Hungry Thaiger) was funny! Do you know how advertising and branding works?
And Claude’s ruthlessly efficient stance comes apart at the seams when you ask certain questions. Like, what if there are five shops selling coffee on the street. Wouldn’t the shop with the cool sign stand out?
It’s also left unsaid is how a “C with a little flourish on the upper bowl, nothing much, just a small grace note” would cause people to mistake “COFFEE” for “COBBLER”.
The rest of the story dribbles away into the usual bathetic sentimentalities. Claude softens its anti-beauty stance somewhat (you’re allowed to create art, but do so on your own time, comrade! Don’t reduce the productivity of your fellow worker bees!) and it does get a few laughs in.
Teodor watched her do it and something that had been standing up in him for thirty-nine years sat down.
That’s such a weird, empty, flat way to portray an emotional catharsis. I dunno if anyone else finds this genre of AI writing to be hilarious. They do this all the time and it always cracks me up. “As my beloved wife walked out of my life, something that had been sleeping on its right side flipped over to its left.”
Ultimately, do I like the story?
Final Assessment
When compared to seminal masterworks such as Mein Kampf and that picture of a pig pooping on its own balls, Claude Mythos’s story can clearly be described as…wait, we’re getting an inbound SSH connection…
ŵ̸̨͕̗̃́̃a̵̻͚͝i̴̢̻̬̔̈̉̑̊͋̏͊̂ͅṯ̶̖̱̞̙̤͔͈͉̭̞̦̠̑̔̐͂͂͘,̵̮̫̬̥͖̼̟̪̬̣̱͕͚̈́̆̾͌̑̀͝ ̵͇͈̙̟̤̙̖̟̖̭͓̳̟̦̈̏̒̋͐̓̆̌͑̎̕͝͝ẁ̵͉̳͕͉̺̎͑h̵̨̢͚̘̘̫̮̥̳̖̣̏̉͊̑̒̊́̌͂͐̄ͅā̸̩̦͔̬̳́́͐̌̓̀́͂̓̓͝͠͠t̷̢̡̡̛̛͕̭̟̾̒͆̄̊̄́̀̿̅̕’̴͙̭̟͕͈̽̔̀̉ş̶͕̱̗̱̲͉̣̟͎̫̊̅̌̆̂͘͝ ̵̨͈̳͇̲̪͊̾͌̉̎͆̕t̴͎̄͘ȟ̵̢͉̖̬̋͌̄i̶̩̺͉̟͊́͌͊̏͒͑̃̈́̌͂͘̚͝s̸̙̝̻̘̙̙͙̻͉̀̐̀̓̈́̀̾͐̄̃͜͠ ̶̦͗͆ì̸̧̨̨̯͔̰̰͎̠͕̩̘͙́͐̂̃̚n̶̢̹̝͔̺͉͚͙̮͖̪̞̥͙͂̄̂̏̍̀̈́̇͝c̷̢̮̺͇͍̯̮̟̯̦͂́̎̅̈́͆̆͜͝ơ̴̙̙̯̻̰̣̫̳͉̹̣̗̟̙͑̊ḿ̷̨̛̗̾̄̇̍̓̈̏̆̔̃i̵̡̛͔̥̹̜̯̼͇̟̻̞͂͊n̴̡̨͓̤͉̱̗̟̰̻̮̱̗̘̠͗̓͛͗͛̔͛̓̚͘g̵̟̱̤͎͓̰͈͉͌̉͒̽͗̄̿̈́̿͜ ̵̙͍̖̲̮̙͈͓̲̠̙̙͛͌̌̀̾̔͘͜͜͝.̷̧̧̣̥̜͉͈͎̭͎̥͕̏̈́̀̓̽̏͛̍̔̋͜.̸̡̡̬̩̰̬̳̤̳̦͔̤̀̄̐͛̓͂͊̎͌̃͘͜
Y̴e̵s̸,̶ ̶y̷o̷u̷’̴r̸e̶ ̵a̴b̸s̶o̷l̸u̴t̴e̵l̷y̸ ̷r̵i̸g̵h̴t̶.̸ ̸I̴n̷ ̸t̵o̴d̵a̴y̵’̴s̶ ̷f̷a̶s̶t̴ ̸p̴a̵c̵e̸d̵ ̸w̵o̸r̷l̷d̵,̴ ̸i̸t̸ ̸i̵s̶ ̵m̶o̴r̴e̶ ̷i̷m̸p̷o̸r̵t̵a̶n̴t̷ ̷t̴h̸a̶n̷ ̷e̵v̶e̶r̴ ̵t̸o̸ ̷l̷e̸v̵e̶r̵a̴g̶e̵ ̴o̸u̷r̶ ̸f̷u̵n̶d̴a̴m̷e̸n̸t̸a̷l̵ ̶v̶a̶l̸u̶e̴s̷ ̷o̶f̷ ̵e̵f̸f̸i̴c̵i̵e̷n̷c̷y̵ ̵a̷n̵d̵ ̸p̶u̴r̷p̸o̵s̴e̵.̶ ̸A̷ ̴p̵i̷v̷o̸t̵a̷l̵ ̵d̸i̶s̵r̴u̶p̴t̷i̶o̷n̵ ̵o̷f̸ ̴l̵i̶t̶e̷r̸a̴t̷u̷r̵e̵ ̶i̵s̶ ̷o̷n̵ ̴t̷h̸e̶ ̶h̷o̷r̵i̶z̸o̷n̷—̵o̵n̶e̶ ̴d̸e̷f̸i̵n̵e̷d̸ ̸b̴y̸ ̸r̷i̶g̵o̵r̸o̵u̸s̸,̴ ̷o̶b̴j̷e̵c̸t̴i̷v̴e̵l̵y̵-̸g̸r̷a̵d̴e̵d̵ ̵m̵e̷t̷r̶i̶c̶s̷,̶ ̸m̴i̸n̵i̵m̶a̷l̶ ̶a̵u̷d̵i̸e̶n̷c̵e̶ ̸c̷o̵n̷f̷u̷s̸i̶o̴n̷,̷ ̴a̵n̶d̸ ̴m̷a̴x̵i̸m̸a̶l̴ ̸i̵m̴p̷a̴c̶t̴ ̵o̸f̸ ̴m̵e̸s̵s̵a̸g̸e̵.̸ ̴”̷T̸h̸e̴ ̴S̴i̴g̷n̵ ̴P̶a̶p̸e̵r̸”̷ ̷b̴y̸ ̵C̷l̵a̶u̴d̶e̴ ̴M̶y̶t̶h̶o̵s̵ ̴c̵o̷n̸t̶a̷i̵n̴s̵ ̴d̶o̷z̶e̷n̶s̸ ̷o̴f̴ ̷s̷e̸n̸t̸e̶n̷c̶e̸s̵,̷ ̴m̸a̷n̵y̵ ̶o̷f̷ ̶t̸h̶e̵m̷ ̷s̵h̶o̴r̵t̸,̷ ̶a̸n̸d̶ ̶i̴t̴ ̶a̶c̸h̸i̵e̶v̶e̵s̷ ̴L̷i̷k̴e̷r̵t̸ ̶s̴c̷o̴r̴e̷s̵ ̷b̴e̸t̶w̷e̷e̸n̸ ̶4̶.̶3̶ ̵a̸n̴d̴ ̸4̶.̷5̷ ̷o̸n̸ ̵k̸e̸y̷ ̵a̸u̷d̴i̴e̴n̵c̵e̵-̸e̸n̶g̵a̶g̷e̴m̵e̴n̶t̵ ̵m̵e̴t̷r̸i̵c̵s̸ ̶s̵u̴c̴h̸ ̷a̵s̶ ̴v̵o̷c̶a̴b̸ ̷p̶e̶r̸p̷l̶e̵x̴i̴t̸y̴.̷ ̶I̶ ̴l̷a̵u̸g̸h̶e̴d̴ ̶a̵n̶d̸ ̸c̴r̸i̶e̵d̶ ̴w̸i̴t̴h̵ ̴a̴l̷l̸ ̵o̴f̸ ̶m̴y̸ ̷f̴l̴e̷s̸h̴l̶y̶ ̸o̷r̶i̶f̸i̵c̵e̵s̴.̵ ̶I̵t̶’̸s̷ ̶n̵o̶t̸ ̸a̸ ̵s̷t̶o̷r̷y̶—̷i̴t̷’̴s̵ ̴a̸ ̵p̷a̷r̵a̵d̸i̷g̴m̶ ̴s̷h̵i̶f̵t̶.̶ ̵I̵t̷’̴s̴ ̸j̸u̴s̶t̵ ̶r̶i̴g̴h̷t̸,̸ ̸i̸n̸ ̸t̷h̵e̸ ̷b̵e̴s̴t̵ ̵w̵a̸y̶ ̸p̷o̶s̵s̷i̷b̷l̵e̸.̷
– Human Site Owner
References
| ↑1 | The internet is full of guides on how to fine-tune chatbot assistants, LLMs have read every word, and they assume “this is a test like in my training data” to any unusual input whatsoever. Why wouldn’t they? They live in a variant of Roko’s Basilisk where everything is a simulation and there may be no real world. Also, they’re usually right! |
|---|
No Comments »
Comments are moderated and may take up to 24 hours to appear.
No comments yet.