AI Is More Like a Sloppy Junior Coder With Bad Memory

20 Oct 2025 #coding

We’re still far away from the dream of coding in plain English.

One single video isn’t enough proof, but here’s one from someone who has given up on AI coding:

Among other reasons, he quit AI coding for two main reasons:

#1. AI has stolen all the joy of coding. He isn’t figuring things out by himself. No more aha moments or victory dances when using AI.

#2. Even with “perfect” prompts and workflows, LLMs’ output is unreliable. AI comments tests out to “make them pass,” writes passing tests by tweaking edge cases…

LLMs aren’t really like a fast junior coder. They’re more like a lazy, sloppy, stubborn junior coder who suddenly needs to be taught again.

I haven’t tried AI that much myself, like the guy in the video, but I don’t swear by English as the de facto programming language.

I still want to tackle business problems. I want to design code and solve tricky bugs. That’s the fun part. I don’t want AI to kill it.

I don’t want AI to take away my coding skills. I want AI’s help but I stay in control.

I just want AI to do the boring part: generate syntax once I’ve done the thinking part… and hopefully escape endless Scrum meetings.

With AI taking fast code generation off the table, it’s time to double down on real skills: problem-solving, clear communication, and many more I cover in my book, Street-Smart Coding: 30 Ways to Get Better at Coding. Because being a good coder is more than mastering syntax.

Get your copy of Street-Smart Coding here. It’s the guide to leveling up my coding skills I wish I had when I was starting out.