AI is removing the middle class of software engineering

In 2020, a senior software engineer found their team's codebase in disarray when they returned from vacation. Upon their return, they discovered that their colleagues had merged numerous merged PRs without adequate review, leading to a chaotic and problematic codebase. By 2026, this same engineer was faced with reviewing 13 PRs each morning, far more than they had completed during their vacation.

The excessive use of AI in software development has accelerated the failure of projects with weak engineering practices. Previously, engineers would discuss and deliberate on their designs before coding, but now they can generate code through prompts and quickly add it to the project. This rapid generation of code has caused the codebase to become so convoluted that no one can comprehend its workings, making it nearly impossible to fix issues.

The author argues that while AI can generate code quickly, it is not a solution to complex problems. Technical debt, while not always bad, becomes a significant issue when decisions are made without knowledge of the system's intricacies. The rapid pace of development has led to a growing disparity in salaries, as companies seek to attract top talent capable of navigating the complexities brought about by AI.

As a result, good engineers are becoming more valuable, while the quality of work produced by less skilled individuals is increasing, leading to a decline in overall software quality.