It’s sad to have to start like this, but this blog is born today out of a moment of frustration. Out of a moment of sadness and confusion.

2026 is definitely being a year. Don’t get me wrong, it’s far from a bad year—a lot of good things have been happening in my life, but in parallel, many bad things have happened as well.

We are at the peak of LLMs, with their imminent decline visible on the horizon, and it’s in this scenario that I find myself in sadness, despair, and frustration.

Earlier this year, you could see a massive shift in the dev bubble. The new models and harnesses released by the leading LLM players caused an uproar by proving to be highly efficient at writing code. Many of the staunch defenders of not using LLMs for development work suddenly switched sides and began supporting their full-scale adoption.

I want to make it clear that, while I’m not a huge supporter of LLMs—because I understand the social and environmental impact they cause—I’m also not a complete conservative when it comes to using AI. I didn’t adopt them when they first appeared because I didn’t think they were as efficient as just Googling things, but as the models improved, they practically replaced my use of Google during development. I saw Copilot as a great tool to help me autocomplete code.

Over the last few years, even with tools and other things being built with LLMs, I didn’t see them as a threat to the activity I love so much: writing code. But that changed. At the beginning of the year, a switch flipped, and now the narrative is that the future is AI writing the code while the developer merely orchestrates and validates it.

I notice a mixed feeling among devs: many share the same sentiment as I do, while others not so much. Phrases like “a developer’s job was never to write code” are shouted to the wind, while I’m left questioning myself: am I a fraud? Was I never a real developer? And I understand the phrase. The job isn’t about creating software; it’s about solving problems through technology. We’ve always spent more time thinking about how to solve a problem before actually writing the code to solve it.

But, despite being a manual and labor-intensive process, writing code was a fun part of the job. It was something artisanal, personal, and meditative. So many times, I lost track of time while writing code to the sound of good music.

In other words, the process using AI isn’t inherently different from the current process; it’s just faster because it skips one of the steps. The problem, for me, is that it has become an exhausting and tiring job. At first, it was because I was constantly reviewing the generated code, but even after I gave up on reviewing and just started accepting whatever came out, the exhaustion didn’t go away. And that happens because an essential part of the process was lost.

Oddly enough, writing code by hand was less exhausting.

I noticed that I entered a state of cognitive atrophy as I kept using LLMs in my development workflow. And worst of all, I’ve been getting more and more frustrated because it really feels like the tool could do everything without my interference, and imposter syndrome strikes harder and harder. It’s definitely me overthinking things, but it’s how I’ve been feeling over these last few months.

Anyway, after that vent, let’s talk about the purpose of this blog. This right here is a logbook. A place where I can post how I’m dealing with this new reality, how I’m trying to find myself again in this field, looking for a way to find fun and comfort while showcasing the processes I’m using to avoid just running on autopilot. I also want to post the personal projects I’m working on, study notes, and maybe another rant here or there.

I’m sorry the beginning is so depressing, but this is the current state of things. A few months ago, I wrote an article about LLMs and art. Maybe I’ll bring it over here, who knows.