Brian Jenney, a senior engineer and founder of coding school Parsity, makes a straightforward argument in IEEE Spectrum: if you let AI write your code without understanding what it wrote, you're training yourself to be replaceable.
The piece, published in Spectrum's Careers section, pushes back against the "vibe coding" trend where developers accept AI output at face value. Jenney's core advice is to use AI as an accelerant for learning rather than a substitute for thinking. In practice, that means tracing through AI-generated logic line by line, questioning design choices the model made, and verifying correctness before committing.
None of this is radical, but it's the kind of discipline that erodes gradually. When Cursor or Claude Code produces a working function in seconds, the temptation to ship it without fully understanding the implementation is real. Multiply that across dozens of AI-assisted changes per day and you end up with engineers who can build features fast but struggle to debug them when something breaks at 2 AM.
The practical takeaway: next time an AI tool generates a solution, spend two minutes asking it why it chose that approach. Ask what alternatives it considered. Ask where the edge cases are. You'll learn more from interrogating one AI response than from accepting ten of them.