AInews - 2026-06-30

AI can make developers faster. It can also make them less sharp.

A founder-led take on the tradeoff between using AI for speed and keeping real coding skill, systems thinking, and judgment intact.

Does using AI make developers weaker? I think, unfortunately, yes, at least in some cases. I saw a simple example today that made the point better than a long theory discussion ever could.

A good friend of mine, a former C++ developer, has not been writing much code by hand for a while. He writes with AI now. I asked him to write bubble sort from memory, in iPhone Notes, with one condition: no looking things up, no asking anyone, and the code had to compile on the first try. He said, sure, of course. He had learned it at university, used it at work, and he knew it.

Then he actually tried to write it. The logic was mostly right, but he missed a small conditional in the `for` loop. That sounds like a tiny thing, and maybe it is. But it is also exactly the kind of detail that falls out of your head when you stop doing the work yourself. I would not call that proof that he became a bad programmer. I would call it a sign that constant AI use can soften memory and attention in ways people do not notice right away.

That is the part people should be honest about. AI makes it very easy to relax. It also makes it very easy to hand off work that is still simpler, faster, and more reliable when a person just does it. Once that becomes a habit, the knowledge gap grows quietly. Newer developers already tend to have less hands-on depth in some areas than people who had to search, fail, compare options, and fix their own mistakes for years.

That does not mean AI is bad. It is useful. It speeds up routine work and gives you room to focus on the parts that actually need judgment. But architecture is still built on your knowledge, not the model’s. System design, tradeoffs, and the shape of the project are still yours. The AI does not own the outcome. You do.

So my own position has become more cautious. I still use AI, and I do not think going back to pure hand-written code is some noble answer. That would be silly. But I do think it is a mistake to delegate everything. If you can write the code yourself, sometimes you should. If you are learning a new language or a new system, you still need the base layer in your own head.

That base is where expertise comes from. Not from pretending the model knows the system for you, and not from turning every problem into a prompt. AI is useful. But it should not replace the part of you that stays sharp, makes the calls, and takes responsibility when the code is wrong.