AI Challenges

Ugh... There are so many AI-related topics, buzzwords, and content every day that I feel a bit uncomfortable writing one more.

I just want to keep it here so I can come back to this text later and check how much of it is still valid, and whether my predictions (I promise not to fantasize too much) actually happened.

So let’s get to some points I believe are [almost] facts.

Profiles are becoming irrelevant

The border between backend and frontend engineers was already quite thin even before the rise of LLMs, since we started executing JavaScript not exclusively in the browser. However, that happened a while ago, and those profiles still existed, and arguably got even more solid as frontend infrastructure grew exponentially over the last 10 years.

It was pretty annoying for people from other platforms and languages to learn it, and vice versa: frontend engineers, I believe, had just enough mental capacity to keep up with industry changes. Of course, I am generalizing, and there are plenty of exceptions in my orbit.

So what changed?

Scaffolding is now basically free, in terms of mental capacity.

To set up a simple frontend app, following not only general rules, but also some specific internal company ones is easy. Adding that app to the CI pipeline and having it deployed is not a matter of day(s) anymore, but minutes.

The same goes for the server side; having a simple CRUD service running inside your internal infrastructure is nothing (honestly, as it should have been from the start, LOL).

But it’s not just that. If you understand how to do engineering, and you can explain what you want to get eventually, as if you explained it to another engineer of a different profile, or to a junior, you can get a working result: tested, linted, and deployed. The code is nice; it can be split into the modules you want. Abstractions are suggested and implemented.

So in many cases there is no point in having several engineers with different profiles building one feature.

Obviously there are plenty of exceptions (yet?), and this border has not completely dissolved. But the trend is obvious, and I believe it will happen. Like 15 years ago, when we had CSS-HTML-only engineers.

Tech debt might not be a thing anymore

Yes, we are yet responsible for what gets merged to the main branch. But let’s be honest: LLMs are becoming better reviewers than we are, and what’s left for us is to look for obvious crap. If your architecture is quite strict, then adding services and features is more or less boilerplate.

And it is not a problem anymore to rewrite things completely (hello to people from Cloudflare). So what is the problem with what we used to call tech debt? It seems this term is becoming irrelevant.

But what are the challenges?

I was in a meeting recently where I heard a phrase:

It is a really bad time to be [engineering] manager

In many aspects I agree with the saying. We are stepping into a place where one person can 'move a mountain.' And let’s be honest, colleagues: not many of us are used to keeping our hands 'dirty' and producing things. And it is an exciting time for those who do.

The challenge is that more and more organizations will flatten the structure to have direct access to the person who builds stuff. There is less and less need for orchestrating and syncing several profiles of experts (engineers) to do something, and we kinda risk becoming an extra wheel.

We still have to synchronize departments: API producers and consumers, though. I hear many voices (I am, or I was, that voice as well) saying that engineering is rarely (or at least is not the main) bottleneck in the industry, and that decision-making, alignment, and communication are.

I have a theory in that regard. It might be that the AI revolution in high-tech is so powerful with its pace and overwhelming nature, that it might just demolish old-school corporate decision-making bureaucracy. Many people might lose control over production, and as a result, many roles might become extinct.

On the other hand, I believe it is a wonderful time to be an engineering manager.

In a way we already have great experience controlling and producing things using prompts. I used to use this famous joke a lot: that my IDE is an email client (or Slack, or a GitHub issues project, or whatever). And suddenly it is not such a joke anymore.

And especially if you have already lived through one or two paradigm changes / technical revolutions / call it what you want. This isn’t me solely referencing my age, LOL; it’s that we already had our share of FOMOs, and we developed patterns that allowed us to adapt (or better lead) to real change and skip what is bullshit.

So whatever is happening in the industry, I say bring it on; it can hardly make me shit my pants; I live in Ukraine in 2026.