Time of writing: late October 2026
Change was looming since 2022, but in 2026, software engineering changed fundamentally? Or did it?
Code gets produced much quicker than it used to. That means that ideas previously left undeveloped, can now be made into production software. Jevons paradox applies here, I believe. More software will be shipped, with more features, and thus higher complexity.
AI amplifies output: if your team was shipping software correctly and timely, AI will accelerate them. If they struggled before, AI will just make it worse.
AI-tooling is changing rapidly so I am hesitant to go into it in detail. Everybody uses Claude Code, skills.md and is working on perfecting the harness for their agents, automating more parts of the SDLC as time passes. Whatever I write about it currently has a shelf-life of half a year at best. It’s easy to let following the bleeding edge distract you from getting actual work done. Trade-offs change continuously as model characteristics change, along with pricing. So as always, timing is everything so following the AI-news cycle is best done from a moderate distance.
Like any tool before it, I also put the agentic AI tooling to good use. I still read the code because I want it to respect the taste I have developed over the years. Because quality still matters for software that is meant to serve a business for multiple years, and I don’t work on anything else than that.
So I stay in control, and that’s how and I remain accountable. And ultimately, that’s what I’ve always been there for.