I use AI to help write code every day in real work, so this comes from practice rather than headlines: AI has changed programming a great deal, and it has made people who think like programmers more valuable, not less.
What AI genuinely does well
- Writes repetitive, patterned code many times faster than a person
- Explains code you cannot follow
- Spots small mistakes people skim past
What it still cannot do — which is most of the job
Here is what people outside the field rarely see: writing code is a small part of a programmer's work. The real work is:
- Deciding what to build. A client asks for A when their actual problem is B.
- Knowing whether the answer is right. AI is wrong in a confident tone; without foundations you cannot tell.
- Designing how the pieces fit so it can be changed later without collapsing.
AI is an assistant who types extremely fast and knows nothing about your business. The more clearly you know what you want, the more useful they become.
What has really shifted in the job market
Junior work that is purely instruction-following is thinner — "turn this design into this page exactly" is easy to replace.
People who understand the problem and can direct AI are worth more — and doing that requires programming foundations.
Should we teach children differently?
Yes, and I have changed three things:
- More emphasis on reading code, because they will review AI output more often than they write from scratch
- More emphasis on asking and designing before typing
- Still writing it by hand early on — like doing arithmetic in your head before using a calculator, so you can tell when an answer is nonsense
The shortest answer
If the goal is to type code to order, that job is shrinking.
If the goal is to think in systems, solve problems and direct powerful tools, this is the most valuable moment in history to learn.
We cover the durable skills in If my child will never be a programmer, why learn to code?
