That feeling of working in software in 2026

I remember the first time I used an AI coding tool. It must have been around 2022 as I had just received access to Github Copilot. The first feature I tried were auto completions.

The completions were wildly off sometimes, but in most cases I would type and it would auto-fill the following lines with an implementation close to what I had in mind. It was fun. It felt a little magic, but was clearly an optional helper to the day-to-day tasks of writing software.

It’s 2026 now and things have changed. A lot. Everyone in my team and everyone I know of who works in software is using an AI agent.

AI-based code completions are already a thing of the past - most tools don’t ship with that feature anymore. Instead you give an agent instructions and it writes the code for you. From features that touch dozens of files to entire codebases: AI coding tools have become crazy good.

In the community, opinions vary.

There are some who have their identity connected to the task of “writing code”. This new way of creating software threatens that identity.

Then there are valid concerns about security, maintainability and responsibility. Not to mention sustainability, considering the energy consumption of large models running 24/7.

Many of the people I know (myself included) currently see it as an opportunity.

If you have always valued creating software that solves a problem, then code is really only the artefact that is produced along the way.

Don’t get me wrong: I still like writing code. But there’s a time for casual programming and there’s a time to get things done. Code can be poetry. But a good software product doesn’t need to be built on poetry.

I personally appreciate having more time to think about the bigger questions: What is my product? Which features are worth building at all? What does my architecture look like?

Even though I have worked with AI for close to 10 years now, I have been hesitant to write about this particular topic. There are so many voices out there already. So much noise. A thousand sloppy LinkedIn posts about how generative AI is disrupting software engineering.

I decided to write this post nonetheless. Simply to document the current moment for myself. Because that’s what I believe it is: A fleeting moment we are experiencing right now: You can feel a shift happening in the industry.

There’s no going back, no matter if the financial part of the AI bubble bursts tomorrow or two years from now. The technology is here to stay.

It feels odd working in software in 2026.

I started my career at a time when producing software was 100% human labor. Now? At least 50% is done by a machine already. More in some companies, less in others.

People are worried about LLMs writing unmaintainable code. But how long will humans still maintain that code? It’s already cheaper to rewrite entire components from scratch than to spend a day debugging legacy code.

Change is happening at such a pace that it’s hard to tell where we’ll be 5 years from now.

It’s both scary and exciting, all at the same time.

(Note: The illustrations in this post are based off AI-generated images. I drew and colored them myself, using Procreate on my iPad.)