Build Real Coding Skills—Then Use AI (In That Order)

12 Feb 2026 #coding

In the days of StackOverflow, we had to verify answers. Now, too often, we accept AI’s output without question.

Catching AI red-handed

Today, in another adventure with AI, I asked Copilot to turn a couple of SQL table definitions into mapping classes for Entity Framework Core. It was the classical 1-to-many relationship.

The problem came when I asked it to generate an API endpoint to store a parent record with a bunch of child records. Something like: create a parent record, then read a table to create its children.

Its first solution was to persist the parent record. Then inside a loop, persist every child record. The classical N+1 problem. Well, the inverse one. Arrggg!

When I prompted it to change it, saying there was no need for the loop, it replied with a “Yes, you can simplify it that way.” Caught you Copilot!

Why coding skills still matter

The N+1 problem was something I could find on the spot.

Now imagine how many AI answers we blindly accept without question. When coding, documenting, researching, testing…

Coding skills still matter. Without them, we wouldn’t even notice the problem.

Blindly trusting AI is what makes us say AI kills CS degrees, what makes us dangerously lazy.

Reviewing the code AI spits out puts in the top 50% of coders. The other 50% don’t always review. You need your coding muscles for that.

AI is like a semi-autonomous car. It always needs hands on the wheel. Build skills. Then leverage AI.

To help you build hype-proof skills, I wrote Street-Smart Coding. Because syntax alone won’t make you stand out.