← Back to archive

AI Code Feels Like Extra Pipes, Not Refactoring

31 March 2025 · 70 views · 0 likes · 0 reposts · 98 blog views

AI coding learning product productivity

Just had a moment of clarity thinking about how AI writes code.

Working with tools like Cursor or Claude sometimes feels like hiring a plumber to fix a leaky pipe… and instead of sealing the leak, they just install two more pipes around it - one to “handle the flow” and another as a backup, just in case. Sure, the water moves. Technically, it works. But now your kitchen has three pipes where there should’ve been one.

That’s kind of what AI-generated code feels like right now. It builds features that function, but it doesn’t always build them the way a thoughtful engineer would. There's repetition, inconsistency, and a tendency to reinvent the wheel every time. It doesn’t refactor; it just adds.

But the more I think about it, the more I believe most of these issues are temporary. As LLMs keep improving - larger context windows, better memory, smarter reasoning - a lot of this messiness might start to fade. And let’s be honest: sometimes it’s not just the AI. It’s us. Prompting well is a skill, and the better I get at it, the better the output gets. When I give it more structure, more intention, the results shift from chaos to clarity.

So maybe it’s not about fighting the AI or forcing it to build like a human. Maybe it’s about learning how to work with it - how to guide it, shape it, prompt it properly. We’re still early. But I’m starting to think that plumber’s extra pipes won’t be around for long.

AI Code Feels Like Extra Pipes, Not Refactoring

Originally posted on X

PS: I'm pretty active on 𝕏 if you'd like to follow along

Related posts

Archive

2026

2025

2024

2018