“AI Code = Dumpster Fire?” asks an advertisement wrapped around a set of pillars in New York’s Penn Station. “We can help,” it answers. The ad is for CodeRabbit, a company whose software identifies and fixes bugs in code written by artificial-intelligence software. Besides the obvious irony of such a product — if AI can replace software engineers, why does AI-generated code require another AI to clean up its mistakes? — the ad conceals a more important reality: much of the work created to fix AI slop is done not by AI but by a new class of gig workers.

As Devvin, a tech consultancy owner, told me in a recent interview, the boom in AI products available to the public meant that start-up founders without a computer science background could now “vibe code,” using AI chatbots to create unpolished but functional app prototypes. Yet when it came to scaling the app to meet the needs of customers and investors, they needed help, usually quickly and cheaply. “When they have a problem,” Devvin explains of new founders, “they’re not going to go to Google and search for a legitimate software company in my area; they’re gonna go to Upwork [an online platform that connects freelance engineers with customers for one-off gigs] and search for software developers.”

Contracting and freelance work in the computing industry are not new. Subcontracting in the tech industry led to the unionization of contingent tech workers at Microsoft in the late 1990s. Freelancers (or gig workers) on platforms like Upwork and Fiverr live precarious lives, waiting for gigs that they don’t know will come. Whole subreddits are dedicated to Upwork contractors recounting months without work and their struggle with the lack of stable income that comes with contracting. AI is not eliminating jobs because it performs so well. Instead, the hype surrounding “AI” conceals the continued growth of businesses reliant on precarious workers.