Reuel's blog

About AI

DISCLAIMERS

Introduction

2026 has been a hell of a ride so far in the software world, and we haven't even reached the halfway mark. All because of those two letters that have (at least for now) changed our relationship with software development and engineering.

The sole purpose of this article is to get all the thoughts out of my head that I've been having over the past few months. It's a look at this journey while everyone is trying to learn how to use these tools, figure out the best ways to apply them, avoid the hype, and, above all, read between the lines of those who have conflicts of interest and are promoting them. In short, I'm just trying to stay sane.

Initial reaction

You could say I'm in the "anti-AI" group. The group that says, "Oh, AI only generates slop," or "I'm better than it at doing my job," or "It takes me so long to use it that it's easier to do it myself." I have said all of these things before.

At the same time, I know that one of the things that keeps a professional relevant in their career and in the market is not fighting against revolutions in their field.

At first, I didn't care much about the hype, as anything hyped triggers a deep feeling of rejection in me.

Until I heard someone in leadership say something like: "I had a meeting with other CTOs and tech leaders. Everybody is using AI. That is no longer questioned. We are behind."

At that moment, I knew things were serious.

Fear of Missing Out

Most engineers are probably in this stage right now. New models are emerging faster than npm packages and becoming obsolete on the exact same day; new buzzwords are emerging (and re-emerging), and so-called "specialists" are trying to push the "new best way to work" while trying to grow their follower count on social media.

All of that is generating a massive wave of anxiety and fatigue among engineers fearing for their careers and having to spend hours upon hours outside of work trying to keep up with the latest releases. All of this is driven by big AI companies trying to shove it down our throats to win the race and turn a profit.

"We're doing it because everybody else is doing it" has become the norm. But it seems nobody is trying to assess whether we should be doing it in the first place.

Management-imposed AI usage

Because "everybody is doing it," a stronger wave of something that has always happened has emerged, but it is much more intense in the AI era: people who know nothing about a subject trying to tell you how to do your job. This irritates me profoundly. It feels like pure disrespect for all the effort someone put in to become good at their craft. No amount of AI will ever make me believe that, without proper training, I would be more qualified than a doctor, HR professional, lawyer, or psychologist to do their job. Why should we accept this approach for engineers?

It is not uncommon to see posts and testimonials from employees who were forced by their leadership to use AI in their day-to-day work without any preparation or understanding of where it fits as a tool, what it is good for, and what it isn't.

This has led to absurd situations, like measuring productivity by token usage or counting the number of lines written by the AI as if those were good metrics (which they aren't).

Most of this productivity is just an illusion. For now (I don't know about the future), the majority of companies don't have a significantly positive ROI from using AI.

The lie of time-saving and toxic productivity

One thing is real: AI and LLMs allow you to produce things faster. One of the most common marketing points big companies used was that you would save a lot of time. It was a lie. What happened instead was that everybody started doing more things (tasks, projects, documents). Some did it because they liked it, some because company pressure for output increased, and many because humans are greedy.

This led to a scenario of toxic productivity: everybody is trying to become more productive, and at the same time, everyone is falling behind because the rate of production is so high that it is impossible to keep up.

Tokenomics

This is something I'm interested in. Until now, none of these companies have been profitable, and people are starting to see what many engineers saw a long time ago: AI is not cheap. There is a lot of cash being burned in the race to win the prize of being the AI company that runs the market and holds the highest market share.

And it is already happening (but somehow silently): pricing plans are shifting, business models are changing, and access to models is becoming more restricted. I'm just waiting to say, "I told you so."

Profit over Value

That was one of the many disappointments I had when I was still naive. The majority of companies care solely about profit, not about delivering value.

Of course, I knew that profit was essential. That's not the point.

The disappointment comes from learning that delivering value is just a means to an end (profit), and not also an end goal in itself. That is the part I was naive about.

Now I think that all those things about mission, vision, and values are just lies.

The vision is: Be rich.
The mission is: Make a profit.
The values are: Do whatever it takes to complete the mission and achieve the vision, even if it means delivering some value along the way.

Identity crisis: Am I an engineer, a programmer, or neither?

I think this is the first time in my career that the distinction between coders, programmers, and engineers has made sense to me. Because coding was such a massive part of developing software, I didn't care much about nomenclature. With the current state of AI writing so much code, I don't think "coder" or "programmer" are terms that suit people who don't do it as much anymore, so we need something else. I still see myself as a programmer, even though I work across all parts of the software development life cycle: planning, architecture, cloud, and talking to people. Writing code is now a smaller portion of that. I understand that "engineer" encompasses many other activities beyond writing code, but I still just prefer "programmer."

On the other hand, this has sparked a much deeper reflection: I didn't notice until now how much my job is tied up in my sense of self. Quite a few times when someone asked me "who are you?", my first thought was, "I am a programmer." Notice that the question was "who are you?" and not "what do you do?"

That hit hard when I finally realized it. Your identity is something far beyond your occupation and your career, and now I have to recalibrate that. Let's say for now that the answer to "who are you?" is something along the lines of: "I'm a son of God who [insert many things here to figure out later] and develops software for a living."

Making software is actually hard

People don't understand this. In the same way, I don't understand how hard it is to pilot a plane, be a surgeon, or be an elite athlete. The difference is that I acknowledge that and don't try to teach those people how to do their jobs.

One of the most humbling things anyone can do is try to do something they think is easy. Our world runs on (mostly crappy) software. I would like to see most people trying to create (even with AI) a system that solves problems, is cost-efficient, and doesn't wake you up at 3 AM because everything is burning down. They would appreciate and respect the craft a lot more if they did.

Because engineers know this (we know that we will be the ones awake at 3 AM to fix things) we are generally the ones who say "no" and "it depends" so much. This brings us to another reality: (almost) nobody likes engineers.

(Almost) Nobody likes engineers

I think this is one of the biggest disappointments I've had with the market in general in this AI era: how happy everyone seemed to be when they believed they wouldn't need engineers anymore (without becoming engineers themselves).

I always knew that in the corporate world, technology was just a means to an end. I always interpreted this as: "Beyond software engineering, I need to understand more about business." The real message I get from the market now is: "As soon as I can completely replace you, I will."

Although I don't believe that this is possible (at least not replacing developers completely), the intention is there, and it's real for any position beyond technology.

With that in mind, another question emerged for me: why work for someone else?

Why work for others?

Part of the nature of software engineering is continuous learning. Just as an example, a software engineer in a high-level position has to know about:

If you already have to know all of these, why not learn a few more things (taxes, sales, marketing) and work for yourself?

Of course, everyone's motivations are different, but it became clear to me that learning and responsibilities are not the limitations. We already have to deal with those.

And speaking of motivations, this AI revolution brought about another disappointment: software engineers themselves.

My disappointment with engineers

For most of my career, I naively thought that the software field was something only people who truly liked writing software would enter. Oh boy, was I wrong.

The first signal came years ago, even before AI. In a call with a former colleague, I shared some courses I was taking to understand computers at a lower level (OS, assembly, instructions, memory, all that nerdy stuff). The response I got was: "I don't know if I ever want to reach that level."

It was shocking. "What do you mean you don't want to learn? How do you not want to know about the main tool you use for work?" I thought to myself.

This year just confirmed it: the high volume of posts by engineers saying "code was the easy part" or "I now spend less time coding and more time building" was just... disappointing.

I agree that code is faster to write now, and its quality is often better than that of the majority of professionals out there. However, I highly disagree that code is the easy part, and that coding is not building.

But the worst part is the disregard for the activity itself. To me, it sounds like hearing a doctor say they don't want to know about anatomy, a musician who doesn't want to know their instrument, or a cook who doesn't care to learn how to mix ingredients. It makes no sense.

Again, in the end, it is about personal motivations. Most of these people probably entered the field motivated by the money. I think that also explains why there is so much crappy software out there.

I'm still learning to accept this reality. I just hope to never work with that kind of professional again.

Joy of working

It is completely gone. I always knew I was privileged to do what I liked (writing software and using it to solve problems) and get paid for it. All the rest was just part of the job. AI took away exactly the part I liked the most.

If things don't go back to the way they were, the challenge now is to find joy in another part of the job, switch careers, or learn to do what I don't like (which makes no sense to me).

It feels too strange to ask a robot to perform an activity I enjoy, even if it is faster and of good quality.

The best comparison I have in mind is artisans or luthiers. Machines can sometimes do it faster, and maybe even better, but where is the fulfillment in that?

I don't take any pride in the work I do using AI. I don't feel fulfilled by it.

Experience

Another thing I've been thinking about is the new generation of engineers. Many engineers today say that "code is the easy part" and that we should focus on the higher levels of development. I agree that we should focus on the higher levels, but not only on them.

What I don't get is: how does someone reach that level of expertise without learning to code and making mistakes themselves? Learning is a process that demands time and effort.

Junior developers are almost completely out of the game nowadays since companies are cutting costs. The gap between the juniors who are in the field and the mid and senior developers is rapidly increasing, and there doesn't seem to be much effort to develop new talent for the future.

AI (LLM) Limitations

I think everybody is finally hitting a wall now. And somehow it is both funny and curious.

Most of the initial marketing appeal was that LLMs (and later agents) could do anything. The reality now is that any person that wants to build something have to learn a massive amount of new concepts, techniques, and strategies just to create an environment where an agent's results are actually good. To name a few:

You get the point, right?

If AI is so good, why do we need all of that?

SaaSpocalypse

Another term coined as a result of stock market speculation, as I understand it. It made a lot of people say that "software is dead" and "SaaS is dead."

To be fair, I don't really believe that. I just think they will be reshaped and their business models will change. At the end of the day, agents and LLMs are still software. And in a sense, they're a "worse" kind because they are non-deterministic.

At first, companies believe that building software in-house is much cheaper and faster. Fair point. But they still have to deal with everything else (cloud, hosting, security). Can AI also do that? Maybe. Will it be worth it? I have no idea. But someone needs to be responsible for it. Who is going to pay for the mess?

Did it bring anything good to the table?

Although most of this article leans pessimistic, it is not about the technology itself, but about how it is being used. I think there are indeed some good things about LLMs and agents in general:

My predictions for the future

I believe nobody really knows what is going to happen. But there are a few things I believe can happen: