Why AI hasn't replaced software engineers, and won't
Points and comments are a snapshot, not live.
AI has not replaced software engineers despite rapid adoption, and structural factors suggest it won't.
The article argues that despite rapid AI adoption in software engineering, layoffs attributed to AI are mostly "AI washing" masking financial troubles. A Federal Reserve study found employment still growing, just 3 percentage points slower post-ChatGPT. The real reason: software development is a "decide-execute-deliver sandwich." AI compresses the execution layer (code writing), but decision-making and delivery/verification require deep understanding of systems, business context, and regulatory constraints that resist automation. A GitHub study of 100,000 developers found AI agents increased code lines 8-fold but releases only 30%, confirming humans remain bottlenecks at both ends. The article cites Block, Snap, and Intuit layoffs as instances where executives attributed cuts to AI capability when underlying causes were financial pressure or unrelated restructuring.
What commenters are saying
Commenters split on whether AI will replace engineers. Top voices emphasize that specification, testing, and accountability remain human responsibilities, with one noting AI productivity euphoria is a temporary phase and that poorly-supervised agents generate "slop." Others counter that while engineers may survive, "developers" writing CRUD apps will be displaced, especially for greenfield projects where scaffolding and platform engineering handle deployment. A third camp argues the analogy to Rails generators shows this has happened before without eliminating the profession. A concrete observation: LLMs still produce buggy code (string-sorted dates, missing context), suggesting current tools lack the reasoning depth the article credits to them.