A few weeks ago, someone on my team shared a SQL script with me. It ran perfectly. The problem was that it gave the wrong answer.
She had used AI to write a tricky part of the query and had checked that it executed without errors. When I asked her to walk me through the logic, she struggled.
It wasn't because she wasn't capable. She simply hadn't spent enough time with the problem to really understand what was happening.
That incident made me think about the ongoing debate around AI and coding.
If AI can write code, do we still need to learn how to code?
I think we do.
AI can remove a lot of the typing. It can help us write code faster, debug it, and even suggest better approaches. But it doesn't remove the need to think.
Someone still needs to understand whether the logic actually makes sense. Someone needs to spot the bug hiding inside code that looks perfectly fine. And someone needs to recognise when AI is confidently giving the wrong answer.
That requires understanding, not just good prompting.
I've started looking at coding differently. Coding was never really about remembering syntax. It was about taking a messy problem and breaking it down into small, precise steps.
Think of it like explaining a recipe to a robot that has no common sense. You can't simply say, "Add a little salt." You have to tell it how much, when to add it, and what should happen next.
That ability to translate human thinking into something precise is the real skill. Python, SQL, or even plain English are simply different ways of communicating with a machine.
And there is some interesting research around this too. An Anthropic study found that developers who relied heavily on AI for writing and debugging performed worse on a later test of their understanding than those who worked through the problems themselves. (source)
That made me reflect on my own kids too.
They are growing up in a world where answers are available almost instantly. I don't want them to miss the thinking simply because the answer is easy to get.
I'd rather they learn to think through a problem first, understand the "why", and then use AI to become faster.
Because the machine will always type faster than us.
But knowing whether what it typed is actually right?
That part is still ours.
