AI is removing the middle class of software engineering?
Points and comments are a snapshot, not live.
AI makes projects with weak engineering culture fail much faster by removing the speed limit on bad decisions.
The author argues that AI tools let engineers generate thousands of lines of code per day, overwhelming review processes and compounding bad architectural decisions before anyone can intervene. In a typical scenario, a team member opens a 25,000-line PR, adds Kafka without justification, and cannot explain where data comes from without consulting a Claude conversation. The author says fixing such messes requires migration plans and rollback strategies far harder than generating the code was. Bad engineers become unemployable because their output is indistinguishable from functional code but lacks the judgment to evaluate AI recommendations. Good engineers become more valuable because AI lets them move faster, while the rest are replaced or cheaply hired.
What commenters are saying
Commenters largely agree that AI accelerates the creation of unmaintainable, brittle codebases. Several note that this mirrors past cycles of bad engineering practices, just faster. Some point out that intentional technical debt is fine when understood, but AI-generated debt often goes unrecognized until too late. A few argue that for startups racing to market, the trade-off may still be worthwhile. Others warn of real harm: corrupted data, security vulnerabilities, regulatory breaches, and production outages. One commenter observes that by delegating thinking to Claude, engineers become managers of the process, not builders. The thread also highlights Oracle banning AI code from its JDK project, partly due to reviewer burden and unsettled copyright law.