The honest guide to using AI coding assistants as a beginner without losing your ability to actually code

The Frontend Mentor guide https://www.frontendmentor.io/articles/ai-coding-assistants-for-beginners covers something that most AI coding tutorials skip: how to use AI assistance while actually learning rather than just getting working code you cannot explain.

The core advice is practical and uncomfortable for people who use AI to get answers rather than to get better: ask questions rather than accepting code, treat the AI as a tutor who explains rather than a ghostwriter who writes, verify suggestions against documentation, and maintain the ability to explain every line before shipping it.

The dependency risk is real and worth naming directly. A developer who cannot write a basic function without AI assistance has a knowledge gap that will compound over time in debugging, architecture, code review and any context where the AI tool is unavailable or unhelpful.

The counterargument is also worth making: AI assistance used well is not cheating, it is a productivity tool that professionals will be expected to use competently. The question is whether your use of it is building your capabilities or replacing them.

What rule have you set for yourself around when to use AI coding assistance and when to work through something manually, and has that rule changed since you started using it?