One financial services company adopted the AI coding tool Cursor and watched its code output jump tenfold. The result? A backlog of one million lines of unreviewed code. According to Futurism AI, which highlighted a behind-the-scenes look from The New York Times, the corporate rush to AI-powered coding is creating a productivity paradox that’s equal parts alarming and absurd.
The promise was simple: AI writes code faster, companies ship faster, everyone wins. The reality is messier. Mountains of AI-generated code are piling up faster than human engineers can review them, and skipping that review isn’t an option. Bad code, whether written by humans or machines, causes security flaws and software failures. Amazon and Meta both experienced disruptions after AI tools took unauthorized actions. Those are just the incidents that went public.
The Security Gap Nobody Planned For
“The sheer amount of code being delivered, and the increase in vulnerabilities, is something they can’t keep up with,” Joni Klippert, CEO of security startup StackHawk, told the NYT. The stress isn’t contained to engineering teams either. Sales, marketing support, and other departments are feeling the ripple effects of accelerated, half-checked output.
The talent shortage makes this worse. “There are not enough application security engineers on the planet to satisfy what just American companies need,” said Joe Sullivan, an adviser to Costanoa Ventures. Companies eliminated the humans who would review code, then discovered they need more humans to review code. The irony is hard to miss.
The Burnout Nobody Talks About
What stands out here is the human cost hiding behind the productivity metrics. Software engineers are reporting that supervising AI tools while being expected to produce more is pushing them toward burnout. Researchers studying this phenomenon have dubbed it AI “brain fry.” Engineers aren’t coding less. They’re coding more, prompting more, reviewing more, and understanding less of what ships.
AI was cited in layoff announcements affecting over 54,000 workers last year, Futurism AI reports. Block and Atlassian cut thousands this year while publicly pivoting to AI. The math doesn’t add up: companies are cutting headcount while simultaneously generating more work that requires human judgment.
The Fix? More AI, Of Course
Corporations are responding to the AI code glut in predictably different ways:
- The hardline approach: Sachin Kamdar of AI startup Elvix requires human review of all AI code. His logic is straightforward: “It’s just going to break something, and they’re not going to know why it broke.”
- The AI-on-AI approach: Anthropic and OpenAI have released AI agents designed to review code. Cursor acquired code review platform Graphite in December.
- The democratization bet: “The blessing and the curse is that now everyone inside your company becomes a coder,” said Replit’s Michele Catasta.
Using AI to review AI-generated code is the kind of recursive solution that sounds elegant until you ask who reviews the reviewer.
What This Means for the Industry
This is a pattern worth watching. The first wave of AI adoption optimized for speed and volume. The second wave will have to optimize for quality, security, and human sanity. Companies that rushed to cut engineering teams may find themselves hiring specialized reviewers, security engineers, and AI supervisors to manage what their tools produce.
For AI practitioners and business leaders, the practical takeaway is clear: measuring AI’s value by lines of code generated is like measuring a writer’s value by word count. Output without quality control isn’t productivity. It’s technical debt with a marketing budget.
The full story is worth reading over at Futurism AI for anyone managing engineering teams through this transition.