Thanks However No Thanks on the Claudeswarms, Kevin Roose

It’s beneficiant of Kevin Roose, New York Instances tech columnist and co-host of the Exhausting Fork podcast, to pity people who find themselves toiling away with out the good thing about claudeswarms. 

i comply with AI adoption fairly carefully, and i’ve by no means seen such a yawning inside/outdoors hole.

individuals in SF are placing multi-agent claudeswarms answerable for their lives, consulting chatbots earlier than each determination, wireheading to a level solely sci-fi writers dared to think about.…

— Kevin Roose (@kevinroose) January 25, 2026

In a January 25 X submit, Roose mentioned that he has “by no means seen such a yawning hole” between Silicon Valley insiders like him and outsiders. He says the individuals he lives close to are “placing multi-agent claudeswarms answerable for their lives, consulting chatbots earlier than each determination,” and “wireheading to a level solely sci-fi writers dared to think about.”

Exhausting Fork entails an excessive amount of guffawing from Roose—principally directed at his extra comedically nimble co-host Casey Newton—so it’s not misplaced on me that Roose is making an attempt to layer some irony and exaggeration on high of his condescension on this submit. He takes that mask right off, however, in his next one, wherein he says he desires “to imagine that everybody can study these things,” however frets that maybe, “restrictive IT insurance policies have created a technology of data staff who won’t ever absolutely catch up.” 

Current Exhausting Fork episodes have been unusually passionate about vibecoding—utilizing AI instruments to carry out speedy software program engineering. As soon as upon a time, Github Copilot and ChatGPT induced software program engineers’ eyes to bug out as a result of they might write code like an individual, and you might run the code, and the code would work. Since round 2021 AI’s knack for coding has been steadily bettering, and steering sure software program engineers towards prophesies of varied types of Armageddon.

As an illustration, Dario Amodei, the CEO of Claude guardian firm Anthropic, revealed one in all these earlier immediately within the type of a 38-page blog post. “Humanity is about to be handed virtually unimaginable energy, and it’s deeply unclear whether or not our social, political, and technological techniques possess the maturity to wield it,” Amodei wrote. 

Roose and Newton should not, initially, software program engineers, however Roose lately used Claude Code to make an app referred to as Stash, an expertise he talked about on Hard Fork. Stash is a read-later app just like the discontinued Pocket, or the still-extant Instapaper. Stash, in line with Roose, does “what I used to make use of Pocket for. Besides now I personal it and I could make adjustments to the app. And I made it, I’d say in about two hours.” Effectively finished. Sincerely. 

In one other episode of Exhausting Fork, listeners provided their own stories about what they’ve been vibecoding. Presumably these individuals didn’t used to code, and now they’re coding, which is admittedly sort of cool. One constructed a instrument for wallpaper shoppers to calculate how a lot wallpaper they should purchase. One other constructed a gamification system for his youngsters’ housekeeping. 

With all due respect to those individuals and the neat stuff they’re pulling off with vibecoding, that is simply individuals giving themselves busywork for enjoyable. There’s nothing flawed with that, however that’s what it’s. 

It’s true that most individuals don’t have the data to carry out software program engineering duties, and it’s intriguing to attempt vibecoding if, like me, you’ve by no means coded something. I’ve had LLMs make some rudimentary side-scrolling video games, construct ray-traced 3-D environments in javascript, and carry out another little experiments that glitched out. I discovered a little bit about LLMs, however it didn’t change my life.

Then once more, I, like many individuals, am bored by optimization and productiveness hacks, and it’s not in my nature to have software program concepts which might be purely software program. In uncommon circumstances the place I really feel a artistic spark that entails coding, the coding tends to be a small a part of the concept, and the remainder of the concept tends to contain much more participating with the world than an LLM can do. As an illustration, I reside in a type of neighborhoods the place individuals go nuts with their Halloween decorations, and I’ve daydreamed about organising festive garden animatronics, however vibecoding a management system would solely get me up to now within the technique of configuring my monsters. Many of the precise work could be me out in my yard with an influence drill, wires, and stakes, futzing with my werewolf dummy, and Claude Code isn’t on the verge of getting that factor to remain upright on my garden. 

Roose and different AI fanatics are speaking recently as if It’s. Lastly. Right here. They make it sound as if AI is absolutely about to take off, and the normies must strap in. 

Subsequent 6 months are going to be actually bizarre. https://t.co/TAtAomZQzb

— Alex Graveley (@alexgraveley) January 25, 2026

When Roose talks about these benighted “data staff” outdoors of San Francisco, if he solely means software program engineers struggling to perform duties that could possibly be carried out by claudeswarms (Claudeswarms, in case you’re questioning, appear to be little digital coder hives that perform complex coding tasks), I believe his pity is misplaced. If AI-inclined coders should not allowed to make use of the newest AI instruments whereas they’re on the clock, and so they’re additionally software program engineers of their spare time, it stands to purpose that they’re taking part in with AI toys at house in the event that they need to. 

And there will be little doubt that, half-joking or not, Roose’s expertise of individuals within the Bay Space “wireheading” and always asking chatbots for all times recommendation is actual. That’s to be anticipated. They’ve a variety of different issues too, like a horrifying new behavior of injecting themselves with peptide solutions they bought online. 

It’s under no circumstances shocking that individuals in San Francisco suppose AI is about to turn into the closest potential factor to a god, as a result of it feels prefer it’s near being the factor lots of people in San Francisco suppose is a god: a software program engineer. An comprehensible mistake. 

However the remainder of the pathetic data staff who aren’t blessed to be within the AI haven of San Francisco don’t essentially imagine software program engineers are all that highly effective, and a few of us are counting the months till subsequent Halloween, and AI isn’t going to be a lot assist getting our latex clowns to look scary by then. It in all probability by no means will, and that’s high-quality.