AI coding and the illusion of progress

There's a term circulating lately: "vibe coding".

It describes the growing habit of letting AI generate entire products based on what feels right rather than clear reasoning or engineering discipline.

It looks fast and impressive at first, but in practice, it's often a recipe for disaster.

When we hand over the reasoning of a product to the machine, we lose the ability to truly funnel into the problem. We can't trace why something works or fails. We just hope the AI figures it out.

Instead of deeply thinking through a problem and tinkering with its details, as only an engineer can, the task becomes figuring out how to tell the AI to do what you think you want.

But without that deep thinking, you often don't even really know what you want.

Then comes the waiting: running the AI a few times until something that feels about right appears.

Software, however, isn't about feelings.

It either works or it doesn't.

It's a precise endeavour.

From my point of view, a lot of investment will be burned on fully AI-generated products.

Those resources would be better spent building human-crafted systems that use AI to augment features, not to produce them.

Many startups today raise funds believing AI will lower development costs. But software has many layers and phases.

What's saved upfront is often paid back multiple times over later: in maintenance, debugging, and lost customers.

Poorly built systems create instability, and users will always migrate toward more robust solutions.

The hidden cost of "cheap" AI-driven development is quality, and quality always comes due.

Even outside software, we see the same trend.

Much of what's written on LinkedIn or in blogs is now AI-generated.

We need to beware of losing our voice, because speaking and writing are how we think.

Lose the voice, and you lose the thinking itself.

Yes, let AI polish your work, but make sure the message is truly yours.

Much has become noise; signal is what's scarce.

I'm reminded of something I once saw at the Mercedes-Benz Museum in Stuttgart:

a horse statue accompanied by a quote from Kaiser Wilhelm II:

"I do believe in the horse; the automobile is no more than a transitory phenomenon."

Funnily enough, my thoughts on AI sometimes echo that quote, and I laughed at myself for it 😅.

There's a bit of "vive la résistance" in me seemingly.

But still, this isn't comparing apples and oranges. The car did replace the horse, yet both are still means of movement, and what truly matters is the rider or driver, the one who architects and directs the movement.

AI is the same: a powerful helper tool, not a thinker.

It can assist, accelerate, and amplify, but the thinking, the reasoning, the responsibility, those must remain human.