These days it feels like AI can handle almost any coding task. Need a React component? GPT-4 will spit one out before you even finish typing “npm install”. Forgot how to write a SQL join? GitHub Copilot will autocomplete it while you’re still Googling “SQL joins for dummies.” We’ve truly entered the “I don’t need to code, I just need to describe coding really well” era.

But here’s the catch. AI-generated code is like a room full of monkeys with typewriters. It can be enthusiastic and even occasionally creative, but it often leaves behind a mess. This is exactly why code reviews are more important than ever. We need them to catch the errors, inconsistencies, and potential bugs that AI might overlook.

So, if you’re thinking you can just unleash the AI and walk away, think again. To really leverage these tools, especially as a junior developer, code reviews are the most important skill to learn. Forget the old way of learning where you write code first and maybe get involved in reviews later. Now, it’s all about understanding and reading code to actually use AI effectively.

Why? Because as amazing as these LLMs are, they are also infamous for generating nonsense or just plain wrong information (hallucination). Junior devs are in a tough spot. With fewer entry-level openings, the opportunities to contribute to real-world projects and learn are limited. That is why, if I were starting to code today, I would focus on code review as the most important skillset to develop right from the start.

If the production server crashes, it’s fine - we can blame the AI :)