Hey there,

Happy Monday!

I highly recommend watching Felienne Hermans' AI made me doubt everything about programming talk at DDD Europe 2026.

She raises some good points about what programming actually is and the impact of (overly) using AI.

It's got me rethinking how I use AI tools, and how technical I need to stay.

I was happy to see ThePrimeagen and ArjenCodes landing on the same thing: fundamentals matter.

I agree! We won't be better engineers by pumping more prompts into agents.

I also think software design questions are more interesting than learning yet another AI tool.

Especially if those tools are increasingly going to tempt me to outsource my thinking.

I do enjoy working with agents, I use them every day, but it's a constant challenge to use them tactically and in a well scoped manner.

A practical example of this is CommitGraph, a simple habit tracker I built last week with Claude.

The first version was done in ~2 hours but it was also only a ~75% prototype.

When using it I noticed that I was only half done:

• It occurred to me that email addresses were not verified nor rate limited.

• Having a friend try it, I noticed that the streak calendar (which measures the number of consecutive days you successfully complete a habit), didn't work for US people because I had defaulted it to CET without timezone support.

• There were countless feature tweaks, one important one was to go from a weekday/all day toggle to daily toggles to measure habits in a more granular way.

You only find these nuances when your app meets real users, and also when you distance yourself from the "initial prototype" phase (that's why I think "spec-to-MVP" is a myth).

All this takes time and space, which goes counter to the direction AI vendors seem to take to hand over more control to the agents.

No longer owning most of the code, and relying on a vendor subscription to be able to write it, is a scary thought.

I like what Corey Schafer recently concluded in a video about his AI use: detailed prompts, review everything, learn anything unfamiliar, and stay critical:

Which is my summarization in my Guardrails Protect Your Codebase. What Protects Your Judgment? article.

That, and keep studying the fundamental concepts. And in my case teaching of course.

What fundamental skills do you think are especially important these days?

Let me know replying to this email, I am curious to hear.

Have a good week!

Bob

P.S. If AI writes the code, own it.

Struggling to stay in control while working with agents?

Book me in for a call and I'll help you find the balance.