AI is removing the middle class of software engineering

Modern software projects are becoming so convoluted with layers and services that teams no longer understand how their own systems work. Instead of building code from scratch, developers increasingly rely on generative artificial intelligence tools like Claude to both write features and troubleshoot bugs.

This reliance creates a dangerous cycle of cognitive debt. When users report recurring bugs, engineers repeatedly ask the AI to fix them rather than investigating the root cause. Team members often do not know where basic system data originates and must ask the artificial intelligence for answers, watching endless walls of text appear on the screen.

Neither the developers nor the artificial intelligence tools have any real way of verifying whether the generated text is true, yet the systems project extreme confidence. Because software engineering teams are outsourcing their understanding to automated tools, human expertise is eroding. This shift effectively removes the middle class of software engineering, leaving workers unable to comprehend or manage the complex codebases they are ostensibly responsible for maintaining.