Kevin Roose, a technology columnist for the New York Times and host of the Hard Fork podcast, sympathizes with people who work hard without earning cloudswarms.
I follow AI discovery very closely, and I’ve never seen so much yawning in/out of the gap.
people in SF put multi-agent cloudswarms in charge of their lives, consulting with chatbots before every decision, teleporting to a degree only sci-fi writers dared to imagine.…
– Kevin Roose (@kevinroose) January 25, 2026
In a January 25 X post, Roose said he had “never seen such a yawning gap” between Silicon Valley insiders like himself and outsiders. He says the people around him are “putting cloudswarms of multiple agents to control their lives, communicating with chatbots before every decision,” and “wireheading on a level only sci-fi writers dare to imagine.”
Hard Fork involves a lot of cheating from Roose—mostly directed at his very funny co-star Casey Newton—so it’s not lost on me that Roose is trying to spread irony and exaggeration over his submission in this post. He removes that mask, however, in his next one, where he says he wants to “believe that everyone can learn this stuff,” but worries that perhaps, “restrictive IT policies have created a generation of knowledge workers who will never fully experience it.”
Recent episodes of Hard Fork have been unusually enthusiastic about vibecoding—using AI tools to perform rapid software engineering. Long ago, Github Copilot and ChatGPT caused the eyes of software developers to pop out because they could write code like a human, and you could use code, and the code would work. Since around 2021 AI’s ability to write codes has been steadily improving, and is guiding some software developers to prophecies of various kinds of Armageddon.
For example, Dario Amodei, CEO of Claude’s parent company Anthropic, published one of these earlier today in the form of a 38-page blog post. Amodei wrote:
Roose and Newton are not, first and foremost, software developers, but Roose recently used Claude Code to create an app called Stash, an experience he talks about on Hard Fork. Stash is an app that reads later like the discontinued Pocket, or the still-extant Instapaper. Stash, according to Roose, does “what I used to do with Pocket. Except now I own it and I can make changes in the app. And I did, I’d say in two hours.” Well done. Submissive.
In another episode of Hard Fork, listeners gave their own stories about how they’ve been singing with vibecoding. These people probably weren’t used to coding, and now they’re coding, which is admittedly cool. Another built a tool for wallpaper clients to calculate how much wallpaper they need to buy. Another created a system for measuring household chores for her children.
With all due respect to these people and the neat stuff they put out with vibecoding, these are just people doing it for fun. There’s nothing wrong with that, but it is what it is.
It’s true that most people don’t have experience doing software engineering jobs, and it’s amazing to try vibecoding if, like me, you’ve never coded anything. I’ve had LLMs make weird side-scrolling games, build 3-D ray-traced environments in javascript, and do other small experiments that fail. I learned a little about LLMs, but it didn’t change my life.
Then again, I, like most people, am bored with efficiency and productivity hacks, and it’s not in my nature to have software-only software ideas. In the rare cases where I do feel a creative spark that involves coding, the coding is often only a small part of the idea, and the whole idea often involves a lot more interaction with the world than an LLM can. For example, I live in one of those places where people go with their Halloween decorations, and I have dreamed of setting up animatronics for the festive grass, but the vibecoding control system will only get me so far in the process of setting up my monsters. Most of the real work would be me out in my yard with a power drill, ropes, and poles, meeting my werewolf dummy, and Claude Code is not on the verge of making that thing sit upright on my lawn.
Roose and other AI enthusiasts have been talking lately as if That’s right. Finally. here. They make it sound like AI is really about to take off, and regulations need to be tightened.
The next 6 months are going to be really weird. https://t.co/TAtAomZQzb
– Alex Graveley (@alexgraveley) January 25, 2026
When Roose talks about these “knowledge workers” sleeping outside of San Francisco, when he specifically means that they are software engineers struggling to accomplish tasks that cloudswarms can do (Claudeswarms, in case you’re wondering, appear to be nests of tiny coders performing complex coding tasks), I suspect his pity is misplaced. If AI-oriented coders are not allowed to use the latest AI tools while on the clock, and are software developers in their spare time, it makes sense that they play with AI toys at home. if they want.
And there’s no doubt that, slightly joking or not, Roose’s experience of people in the Bay Area “wireheading” and constantly asking chatbots for life advice rings true. That is to be expected. They have a lot of other problems too, like the scary new habit of injecting themselves with peptide solutions they bought online.
It’s not at all surprising that the people of San Francisco think AI is about to become the closest thing to god, because it feels like it’s about to become what most San Franciscos think is god: a software engineer. An understandable mistake.
But some sad information workers who aren’t blessed to be at the AI terminal in San Francisco don’t really believe that software engineers are that powerful, and some of us are counting down the months until next Halloween, and AI won’t be much help in making our latex clowns look scary then. You probably won’t, and that’s okay.
