AI code: 30 days of full AI coding? Here's the disaster.

Well, the code works, but only at first glance. One developer decided to completely delegate code writing to artificial intelligence for 30 days. Not just suggestions, but everything. The result? A mountain of code that, yes, ran, but was often a convoluted and inelegant solution.

The author, infoinlet1, detailed his 30-day coding experiment on dev.to. He wanted to see if AI could genuinely replace a human developer. He found that AI excels at producing boilerplate, those repetitive parts nobody enjoys writing. However, when a touch of human logic was needed, things got quite complicated.

The main problems were twofold: efficiency and logic. The AI-generated code was often a lumbering beast, slow and heavy, taking unnecessary detours to reach the same destination. It's like asking a robot to peel a banana with an axe.

Furthermore, AI struggled with a project's overall "cohesion." It didn't grasp the broader context or the rationale behind past decisions. This led to duplicate functions or solutions that ignored existing code. Quite a mess, really. Ultimately, the time saved writing code was spent fixing the AI's mistakes. Not exactly the promised paradise, was it?