AI Won't Check Your Code for You

I've been a programmer for a good while now, and I use AI in my work. I won't deny it helps a lot, but I still have to fix code, rethink decisions, and dig into problems in the solutions it generates.

Sometimes I go with a suggestion that seems to make sense, and only later, when I test it, do I realize it wasn't the best call and start over.

Experience helps me question the code. Even so, I still have to study, read the docs, and walk back some of my choices.

That's why it worries me to see someone click through an app's buttons, check that the data got saved, and decide they're ready to sell the greatest SaaS in the universe.

A permissions bug can expose customer data without a single error showing up on screen. And whoever thinks they're the next big tech guru has no idea it's happening, or how to fix it.

When you put a product out there, the chance to make money comes with responsibility for your users' data and their work.

You can start out without knowing much. But you have to set aside time to learn programming, understand how data is handled, and test the scenarios where things go wrong. When you can't judge a decision on your own, get a technical review.

And put people who've never used your system in front of it. They'll take paths you never imagined and find problems that slipped right past your own testing.

I'm still learning and fixing my own mistakes, including the ones I make when I trust an answer from Claude or GPT a little too much.

These tools cut out a lot of the distance between an idea and a working app. Checking what got built is still part of my job.