It’s been quite a while since my last post here. Honestly, 2025 was an incredibly overwhelming year for me professionally. I finally found the time to sit down and write this slightly delayed wrap-up, reflecting on what happened and where it’s all heading.
The biggest shift for me was not just that AI tools became better, but that my role started changing around them: from writing more code myself to organizing work, context, quality, and small agent-driven workflows. This post is my attempt to make sense of that shift: the productivity gains, the stress, the documentation paradox, and why I think 2026 will be less about hype and more about control.
I was quite an early adopter of GitHub Copilot back in 2022. Even with those modest capabilities, it felt like a big thing immediately. It could be annoying and distracting, but it pushed my performance in certain tasks significantly, so I never really thought about giving up using it. In 2025, I’m not sure if there was a single working day when I didn’t use an AI agent. I’ve had a very small development team for a while, and I was getting used to looking at the roadmap and thinking: this is just too big for us now. The team didn’t grow, but I started asking myself instead: “Okay, I just need to figure out how to organize it properly, and we can do it”. And this newfound power is something overwhelming.
AI models are becoming smarter every day, but it feels like the main focus has shifted to the tooling around them. I’m personally very glad about this. No matter how curious I was about machine learning, I’ve always been more of a product developer (I still remember that “Machine Learning course by Andrew Ng” on Coursera which I was too lazy to finish). Now you can radically improve results by programming agents and orchestrating them, leaving the nuances of model optimization and fine-tuning to the professionals in that field. This is exactly what most companies and startups are doing - copying features from each other, sometimes improving them along the way.
In the pursuit of incredible productivity, it’s best not to forget about such boring stuff as quality and security. It seems to me - or perhaps I’m just hoping - that in 2026 we will see a more active development of tooling in these directions. After all, any euphoria eventually fades, and the time comes to look at the situation as a whole, including the less stunning aspects. There is a temptation to write all the code in the world, to implement all the features. But what to do with them next, how to find the time and mind space for maintenance - we are not quite ready for this yet. As for code maintainability, the opinion is gaining popularity that source code will become a black box, like bytecode for high-level programmers. But I don’t think anything will change radically in this regard in the near future. After all, the entire history of programming evolution is a transition to increasingly higher-level abstractions, yet the principle itself hasn’t disappeared.
QA in the context of agentic development is a big and very interesting topic. Although I’ve had to deal with this part of software engineering a lot, including designing automated testing frameworks (more than one, actually!), I’m still watching from the sidelines without diving in too deep and waiting for my dear colleagues in this field to show something impressive.
The situation with documentation, in my view, is quite paradoxical. On one hand, internal documentation is more important than ever. It is the foundation for agentic work. With the growing pace of development and the volume of code, few can keep everything in their heads. And writing documentation has become much easier, especially for those who experience “fear of the blank page” or have difficulties with the language. But there is a trap here: it’s never been easier to mimic good documentation. AI can generate pages of text that look helpful but lack actual insight. In 2026, being able to maintain high-quality, truthful documentation will be a superpower.
Unfortunately, the human psyche is much less adaptable and struggles to cope with such a pace of technological development. The “flow state”, when you are deeply focused on a task, is becoming increasingly mythical and unobtainable. It doesn’t matter if you’re writing code, drafting an API spec, or planning a new feature. Due to the high volume of information and constant context switching, it’s easy to feel stress and anxiety. And the result doesn’t always bring satisfaction because it doesn’t feel entirely yours in the usual sense.
What am I doing about it? First of all, it’s acceptance. I realized the importance of accepting the thought that development has changed forever. I’m not wearing rose-colored glasses, I see a huge amount of excess, wishful thinking, and deafening hype. Some of this is caused by human nature itself, some is the cold calculation of financiers and marketers, and some is the plain imperfection of technology in its current form. But none of this will turn things back (we’ll leave any events on the scale of Dune’s The Great Revolt for another time 😄 ). There’s a lot to get used to and adapt to, but our field has always been dynamically evolving.
What will happen to the industry as a whole? It’s a question troubling many, including myself. Recent years have already been full of turbulence. All that’s left is to wait and see. But I am sure of one thing and I advise all my mates: try to be on the cutting edge. Now is not the time for skepticism, you risk getting left behind.
Everyone has long known about the importance of soft skills in engineering roles and how damaging ignoring their development can be. It seems that today their importance, as well as related management skills, is reaching a new level. After all, even the most hard-core engineer is now required to build their own processes, data pipelines, and manage their own small team of agent-developers. In a sense, the concept of the “individual contributor” role is becoming blurred and is gradually being taken over by AI agents. Perhaps this will be a very painful blow to many top professionals if they refuse to step up to the next level.
There’s a lot of talk about the bubble right now. It’s important to distinguish between a financial bubble and a tech bubble (at least, that’s how I define it for myself). Although I am interested in the topic of financial markets and investments, the former has little to do with the main theme of this post. The technology is undoubtedly revolutionary, regardless of how many companies get rich or go bankrupt running their schemes.
Agents have been useful to me outside of work as well. I finally tidied up my HomeLab infrastructure, reworked the Ansible scripts for my home and cloud servers, which host my Home Assistant and other self-hosted or pet projects. I’ve known for a long time what needed to be fixed and how, but I never got around to it, because late on a Sunday night I would have had to answer the question: was this really the best way to spend the weekend? (lately, it’s become harder for me to justify 😅). Now, the whole thing was settled in a few hours.
I built myself several agentic n8n workflows to simplify some routine tasks. It wasn’t always strictly necessary, but curiosity took over. Generating a custom app for a short trip to Paris, with a custom plan, an interactive map, and a built-in AI assistant - in 2025, that doesn’t even surprise anyone.
Well, let’s see what the year 2026 has in store for us.
This article was originally published on LinkedIn on 2026-01-10.