AI is the new rockstar developer: brilliant, novelty-obsessed, and it leaves you the cleanup
Jesse Skinner argues coding models behave like the worst kind of star engineer — generating tens of thousands of lines nobody else can maintain. An engineer on two hours' sleep after a SEV-0 had a blunter version.
The argument: a model will write an entire module where you needed a ten-line function, in a 'belt and suspenders' style that only an LLM can later decipher.
The thread split hard. One camp says this is just VB and PHP spaghetti all over again, nothing new; another runs paid cleanup gigs and calls it free money, 10–50k euros at a time.
The dissent that landed hardest came from someone mid-incident: people still have to review the code, and the reviewers still have to be good enough to write it themselves.
The metaphor: your most talented liability
Jesse Skinner's post reaches for an old archetype to describe a new tool. The "rockstar developer" is the gifted engineer who chases novelty, writes clever spaghetti, rejects everyone else's work, and leaves code only they understand. AI coding tools, he argues, are that archetype industrialized: they generate tens of thousands of lines in minutes with no regard for system cohesion or whether the team can follow along, defaulting to complex "belt and suspenders" solutions whether or not the problem warrants them.
The failure mode he names is dependency. Code can get complex enough that only an LLM can decipher it, and fragmented multi-chat sessions scatter the mess across the codebase. His prescription is to lead the tool deliberately — small focused snippets, simplified architecture, sometimes writing it yourself — and his closer is a line of craft pride: the one thing you can never outsource to a machine.
"People STILL need to review that fucking code"
The sharpest agreement came from an engineer who was, at that moment, living the post.
"people STILL need to review that fucking code, and the people who review the fucking code STILL need to be good enough to be able to write it themselves if they needed. Whoever says otherwise is either an AI investor, a linkedin influencer or a complete imbecile." — an engineer cleaning up a SEV-0 on two hours' sleep
The scale of the problem is what changed, one commenter noted: the daily ceiling on how much technical debt a team can create has gone up by one or two orders of magnitude.
"There was a practical limit on the amount of technical debt that could be created in a day before, now that limit is 10x or 100x higher. It can be generated far faster than it can be fixed." — a commenter on the new economics of slop
The reversal: this is just VB spaghetti, and it pays well
A large camp refused the novelty framing entirely.
"It's actually not much different than a decade ago cleaning up somebody else's giant Visual Basic or PHP spaghetti code." — a developer who runs a month-long class on exactly this cleanup
For some, the mess is the business model:
"some client tried to vibe some departmental tool; the result is some massive crap in nextjs that needs 10GB mem to compile, has 1000s of lint errors … its basically free 10k-50k euros over and over again for this type of work." — a developer who does paid AI-cleanup work
And a quieter rebuttal pushed back on the convenient "it's just like a compiler" defense some use to wave the problem away:
"Compilers are deterministic and they actually possess domain knowledge … AI models are non-deterministic, have no real domain knowledge due to lack of an underlying world model." — a commenter rejecting the compiler analogy
The throughline isn't "AI bad." It's that the bottleneck moved. Generation got cheap; comprehension and review didn't — and that's the part you still can't buy by the token.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
Apple won't ship Siri AI in the EU — and the EU says that was 'Apple's and Apple's only' decision
Brussels denied Apple an exemption from the DMA's interoperability rule, which would force Apple to open its phone-controlling AI agent to rival models. Apple decided not to launch rather than comply.