Coding with AI: How to Trust the Storyteller

You may still write code by hand. Most of us did that for thousands of hours every year until recently. But things changed a year ago. AI models started writing better code than I can. And they can do it faster. Not just a bit faster; an AI can literally build a feature in hours, a feature that would take me a month or maybe two. However, LLMs are storytellers, not truth machines. Storytellers lie or simply invent things to make the story good. Can we trust AI with our code?

Can we trust AI? That’s the wrong question. Of course we can’t. But our job now is to verify and be confident that the code they produce isn’t just telling stories with code, but actually solving real-life problems our customers have. We’re still responsible for the code our AI produces.

Never trust the storyteller. Only trust the story.