It has been a few years since I published the last official wrap-up of the year. Since this blog is not a commercial product, I refuse to provide the reasons to everyone including myself.
2025 was yet another great year. It was great because I witnessed a lot of turbulent and vexatious moments. It definitely deserves a blog post to briefly look back and ponder what will happen next year. However, the development of AI and the tooling produced by it are moving too rapidly. People have been making forecasts like AI 2027, and what I am trying to do here is write just a note for myself, hoping it resonates with you.
I Built More
I definitely built more in 2025, like mimesis, which is a computer architecture simulator. I ask myself: with AI, why there were more to build? The answer turns out to be: if I loved to drive my way around Europe, I would not pay for a taxi to do that. Programming has been a mental sport, like all other competitive sports in the Olympics: professionals compete (daily dev work) in one way and others enjoy in another way (dev as a leisure). The data speaks for itself: I had 300+ more commits in 2025 than in 2024, and even more than in 2023.
With its growth, AI takes away a lot of mental burden from my daily work. I have more remaining mental energy during my free time, so I get extra opportunities to build more.
Is this ideal? I cannot tell, because now I have achieved more outcomes, with or without AI, and I simply cannot differentiate the values of those artefacts.
Fear-Driven Development
It is the uncertainty and uncontrollability that leads to fear among software engineers. Software Engineering is like nothing before since 2025. To remind ourselves of the transition over the years:
The general hype of the year since 2022
With more FLOPS available, models are now stronger than their predecessors and we are amazed by their ability, speed, and accuracy to return outcomes in all scenarios we can think of when writing software. Of course, you still need to use and instruct models correctly, but it seems the boundaries of what a model can do are being pushed on a daily basis.
The speed of development of AI creates lots of fear since it is getting out of control. Agents used to require setups to attain near-human performance, but over the last couple of months, people realised they are stronger than ever, at an unprecedented speed. I have also noticed that people build products not because they want to, but because they can, and the sense of fulfilment has just gone. We now enter the phase of Fear-Driven Development, where a lot of developers build more because of the uncontrollability, because AI agents have taken the moats and qualifications away from them, and people are lost.
The bar at which a person can create code artefacts and products has lowered. Does this mean the value of software engineering has decreased? I fear I do not know the answer. Personally, among those public developers who stream and talk about tech publicly, I have seen fewer happy faces and jokes. Most of us are panicked and stressed when talking about the future of software engineering.
I observe the same in myself. AI feels like a triple-edged sword that helps everyone as software engineers, but also takes away the other fulfillments that doing such activities brings naturally as a human.
As an echo to the other wrap-ups then, no matter what will happen, I look forward to building personal projects, and experimenting with more ideas in 2026 anyway.