LLM Disruption

30 years of commercial development later, and LLM code generation is the biggest disruption I've encountered.

Put aside the moral, environmental, bubble worries or aversion to the hype. I think the bubble is accounting malpractice and Sam Altman makes me sick.

But if you're hiring developers, running development teams, are a developer, or just starting out - sticking your head in the sand is going to get you nowhere. It's not going away, whether you wish it would or not. Not using AI tools is soon going to be like not using a mouse.

I tried them all in 2025. Wrote web and mobile apps from scratch. Used them to help development on existing projects. To explain things to me. To fix bugs. To generate bugs. I've shipped production apps and production features. Some things worked, some things didn't. From Claude to Codex, from Deepseek to Qwen.

My conclusion? It's like having a superpower. Features that wouldn't have got written because they would take too long. Fun projects I wouldn't even start because there's only so many hours. Optimisations I hadn't seen. Trips I could go on, because I wasn't spending as long working.

Can it do everything and solve every problem? Of course not. But I'm comparing against not having it at all, not comparing against the hype.

I don't know where it's going next. And I wouldn't want to be a junior developer right now. But you have to be thinking now about what happens when AI is writing 10% of your code. Then 20%. 50%. More. How do you project manage it? Do you still hire as many developers, or hire more QA to test their output? Does programming language matter anymore? What happens when the tooling improves and customers can generate their own solutions?

2026 is going to be fun.