Everyone's learning to prompt better. That's the wrong skill.

In a recent article on Dev.to, the author argues that the current obsession with mastering prompt engineering is misguided. While many developers spend hours refining prompts to get better AI outputs, the author contends that this is a low-value skill that is rapidly being commoditized by model improvements. The real challenge, according to the piece, is not how to ask the AI, but how to evaluate its output. Because AI models can generate highly convincing but incorrect code, the most critical skill for modern engineers is 'distrust'—the ability to rigorously verify, test, and challenge AI-generated results. The author suggests that instead of focusing on prompt optimization, developers should build systems that separate the 'author' (the AI) from the 'skeptic' (the verification process), ensuring that human oversight remains the final, essential filter for production-ready code.