A post titled "Using AI to write better code more slowly" is sitting at 923 points on Hacker News, making it by far the most engaged story on the platform right now. The title alone explains why it resonated. In a world where the dominant narrative around AI coding tools is speed, productivity, and automation, someone making the case for deliberate slowness is going to stop a lot of developers mid-scroll.

The core argument, which has been circulating in developer communities with increasing urgency, is that AI code generation tools like GitHub Copilot, Cursor, and Claude Code are most valuable when used as thinking partners rather than output machines. The failure mode most experienced engineers are observing is not that the AI writes bad code. It is that it writes plausible code so quickly that developers stop understanding what they are accepting into their codebase. The result is systems that work until they do not, at which point nobody knows why.

Writing code slowly with AI assistance means doing something counterintuitive: using the tool to force yourself to understand each decision rather than to skip past it. It means asking the AI to explain its reasoning, pushing back on suggestions, exploring alternatives, and treating the interaction as a Socratic dialogue rather than a dictation session. This produces less code per hour and better code per project.

The deeper issue this conversation is surfacing is about expertise and how it develops. Programming ability has historically been built through the friction of struggling with problems, making mistakes, and debugging the consequences of your own decisions. When AI removes that friction entirely for beginners, it may also remove the conditions under which genuine understanding develops. You can build without learning, which feels like progress until you need to understand what you built.

For experienced developers the calculus is different. They already have the mental models and can use AI to handle tedious implementation details while retaining architectural judgment. The concern is about what happens to the next generation of developers who start their careers with AI assistance as the baseline and never develop the underlying intuitions that make that assistance meaningful.

The 923 upvotes on a post arguing for slowness in an industry that fetishizes speed reflects genuine anxiety in the developer community about what these tools are doing to craft, learning, and the long term health of the profession. It is a conversation worth following closely as AI coding tools become standard rather than exceptional.