The rise of AI-powered coding tools has given birth to a new reality for experienced developers — acting as AI babysitters. While these tools promise faster development, senior engineers say the extra hours spent reviewing and fixing AI-generated code have become unavoidable.
Carla Rover, a 15-year veteran in web development, experienced this firsthand when she had to restart an entire project due to AI errors. “I treated the AI like a co-pilot, but it wasn’t. The code was full of issues,” she said. “Working with AI can be worse than babysitting.”
Rover’s story mirrors findings from a recent Fastly report, which revealed that 95% of developers spend significant time correcting AI-generated code. Senior engineers, in particular, bear the brunt, acting as AI babysitters rather than just coders.
Why Developers Are Becoming AI Babysitters
Developers told TechCrunch that AI code often “hallucinates” package names, introduces security vulnerabilities, or duplicates work unnecessarily. Instead of eliminating tedious tasks, it creates new ones.
Feridoon Malekzadeh, a software veteran, compared vibe coding — the term for working with AI-generated code — to “hiring your stubborn teenager to help you.” He said:
“They do some things right, some things wrong, and break stuff along the way. You can’t just trust it blindly.”
Malekzadeh estimates he spends 40% of his time fixing AI mistakes, a sentiment echoed by many senior developers surveyed.
The Rise of ‘Vibe Code Cleanup Specialists’
The problem has grown enough to create a new corporate role — vibe code cleanup specialists — whose main job is to fix AI’s coding errors.
Austin Spires, senior director of developer enablement at Fastly, explained:
“AI prioritizes speed over correctness. It’s like teaching a kid to make coffee — possible, but messy.”
Security experts warn that this lack of rigor can lead to vulnerabilities, especially for startups relying heavily on AI-generated code.
Balancing Speed and Accuracy
Despite the challenges, most developers admit that AI accelerates work. Senior developers in the Fastly survey were twice as likely as juniors to put AI-generated code into production after review, citing faster prototyping and reduced manual labor.
Malekzadeh calls it “an innovation tax.” Developers spend more time checking the code, but the overall productivity gains make it worthwhile.
Rover agrees, saying AI tools help her build interfaces faster:
“It’s not a business model on its own, but it removes the boring parts so we can focus on creativity.”
The New Normal for Developers
Experts predict that AI babysitters will soon become the standard in software development. Developers will act less like coders and more like consultants — guiding AI systems, reviewing their work, and ensuring accountability.
As young engineer Elvis Kimara put it:
“We won’t just be writing code; we’ll be mentoring AI systems. It’s the future of software development.”