Controlled Chaos: Nuclear Samurai's Brainrot

By @JustinWetch

The name came out of a machine. In 2022, with a fresh Midjourney beta invite, he was feeding random word fragments into the prompt box to see what happened. Two words kept pulling him back, partly for the outputs, partly for the friction between them. Nuclear Samurai. He made it his name.

Only later did he learn the term already had an owner. The nuclear samurai were the elderly Japanese citizens who volunteered to clean up Fukushima's radiation, sacrificing their health so younger people wouldn't have to.

"I was like, oh, well, it's too late now because people know me as that."

Underneath the machine-generated name is the thing that actually explains him. "There's a bit of a gap for me in understanding other people," he says. He closes it by studying. Photographers, artists, internet culture, the machines themselves. Everything he makes is some version of the same move: study the thing until you understand it, then make something that passes the understanding along.

Nuke grew up in an economically depressed part of Melbourne and has never lived anywhere else. Art was "one of the big escapes" from how dreary life was. He drew, painted, studied murals under a mentor, and later helped that mentor teach substance-addicted kids to paint walls the way he did. After high school there was no university, just the workforce.

The other inheritance was machines. "We didn't have much growing up, but what we did have was computers, somehow." He has memories of using DOS at five, not understanding what he was typing but knowing what to type to run his games. MS Paint as a kid, Photoshop at eleven, and a pile of self-taught bad habits he'd spend years unlearning.

The engine underneath both is wiring. The ADHD is diagnosed; the autism, he says, is a suspicion, and some days he'll tell you yes. In practice it means hyperfocus. When something gives him the dopamine, he can lock in and pour "ungodly amounts of work" into it until it's done. When it doesn't, he waits until the pressure is high enough to run on adrenaline instead.

"You can achieve amazing things under that pressure," he says. "But it comes at a real cost, because it's not a really healthy way to live."

In his mid-20s, making digital art, he bought a camera for better references and assumed photography would be easy. It wasn't. So he tested himself weekly: flowers first, then macro shots of spiders and moths, then portraits, then fashion. He climbed until he hit a plateau he couldn't teach himself past. The photos were technically strong and "narratively weak," nothing pulling you back for a second look.

Then he got lucky by failing. An admin job he was bad at handed him a redundancy payout, and he spent it on a fine art degree as a mature-age student, working three jobs through school. He walked out of the course into a studio job and spent nearly a decade working up to running the place.

What he studied all along was narrative charge. The street photographers, Philip-Lorca diCorcia, Alex Webb, Cartier-Bresson, made images you'd read one way today and another way tomorrow, the eye led through the frame, the story fed by what's left outside it. He wanted to know why those photos rewarded a hundredth look.

The same compulsion produced X Figures, his biggest project before now. During a quiet stretch he studied XCOPY's entire output, start to finish, to understand the artist behind it, then rendered every single piece as an action figure. Study the thing until you understand it. Pass it along.

His bridge from photography to AI is street photography itself. You know what the light does on that corner at that hour, you set up, you wait, you see what comes through. "I really enjoy taking my hands off the wheel and seeing where the car goes," he says. You don't control all of it. You control some of it, and the controlling is where the authorship lives.

Nobody would mistake him for a credulous booster. Ask why AI output is so often bad and he has the answer: it pulls everything toward the average instead of the exceptional, because it's trained to give the most related answer rather than take a swing. "The one thing that is original in art is your perspective on it as a person," he says. "When you take the subjectivity or opinionation out of art, then it becomes really kind of not that interesting."

So the human work is taste. When AI hands him a bad idea, "that's an idea that's off the table now," and the clearing inspires the good one. He calls AI "the teacher who never gets sick of you asking questions," which is not the same thing as a teacher worth obeying.

There's an older knowledge underneath this. Commercial photography, his trade for a decade, exists to make you feel something so you'll buy something. He knows from behind the curtain how images manipulate. Now he turns that machinery around and asks questions with it instead.

Brainrot, the cultural form, is the churn of absurdist slop narratives flooding TikTok and Instagram. Italian Brainrot. Fruit Love Island. Fifteen-second sagas of ridiculous characters, made cheap, consumed by millions. A new post-AI aesthetic forming in real time, the way post-internet art once formed, mostly belonging to a generation younger than his.

He wanted to know why people were giving it their time. So he studied it the only way he knows, this time with help. He ran up a small fortune in Claude tokens having AI watch brainrot content for him, routing video through Gemini for visual reports, building a taxonomy of the recurring characters, narratives, and jokes. What made them funny. What made them spread.

Then came the discovery that shaped the whole project. "AI is actually very lazy," he says. Left to generate freely, it takes the path of least resistance, and at scale the same fifteen concepts repeat forever. Some expensive mistakes taught him that.

His answer was to force the machine to work against its own laziness. He had it build a combinatorial generator, closer to the engine behind CryptoPunks than to a prompt box, with a possibility space near a billion outputs. Then a network of AI agents to curate the results, judging which had memetic staying power. Reading their reasoning sessions, he found them so entertaining he built a third layer: an AI court, with a biased advocate, adversary, and judge arguing the merits of each piece, trials people can replay and watch.

The art is the system. And he left deliberate room for collectors to bring their own creativity to the outputs, so the network finishes what the machine started. He wanted "a synthesis rather than a derivation." Rigorous machinery, built precisely so he can take his hands off the wheel inside it.

There's a phrase on his Linktree: latent space voyeur. It comes from an unreleased project called Voyeur, born from the outputs that feel too intimate to look at. "It feels like you shouldn't be looking at this thing, and yet you are," he says. He describes the work like fishing in pure possibility. "Sometimes you pull something out from the depths that is so wonderful or horrible that you want other people to see it."

What does he wish people understood? That he's funny. "I'm not always being deathly serious," he says, and humor has been a lifeline through the darkest times. Brainrot, he figures, has probably underlined it by now.

Artists inspiring him lately: @deltasauce, a fellow latent space voyeur he finds endlessly inspiring. @serc1n, whose Normies opened up the idea of creating a space for other people to create, a direct influence on Brainrot. And @rocketgirlART, who paints with an expressiveness he says he can't reach, and who delights him every time.

For future Weekly Dose episodes, Nuke recommends @Bastienjpg, @pobedeen, and @YigitDuman.