I am a bit embarrassed, all I have written about on this blog is AI related. But so is the way of the world and my thoughts these days. I really want to try and explore a bunch of different ideas around the future in an AI world.
We started with (and continue to have) lots of fake promises and incorrect predictions. Leaders in the space since 2022 seem to have been promising the death of certain industries in 9 months so many time, and yes of course they have all been wrong. By 2026 the latest predictions feel less so “We will take your job” and more like “AI will wipe us out in 10 years”. I’m really not here to debate any of these statements. What is true so far from a software engineering perspective is that it has massively changed the job.
I feel like 2026 has really been a turning point for software engineering. Pre-2026 we had the line completion which was okay, but not a massive change in how work is completed. Then by 2025 we could get AI to start writing functions, but I felt like they rarely got it fully correct. The code they provided was always a starting point for an engineer to develop from. Speeding up the engineer’s job yes, but still requiring an engineer who knew the details. But now this year, 2026, the engineer doesn’t just have a tool which is speeding up development. The engineer is entirely replaced for the code generation part. Code is cheap.
There’s an incredible amount of nuance to this opinion and the statements such as “code is cheap”, “engineers don’t have to write code anymore” and so on. These are definitely generalisations, but still generally I believe these statements have become true. I feel a portion of the people who disagree are rightfully upset that a portion of their unique skills have quickly been replaced. At the end of the day, we work at companies to make the companies more money. If AI can create the outcome a business wants for cheaper, then businesses will eventually follow that path.
The goal is profit optimisation.
The software engineer’s job is now one based on systems experience, knowing how best to architect a solution. Maybe in future even that will start to be replaced? Maybe it will be normalised to have agents handle all the infrastructure work? It could become normal to just say AI go create a 5G core, and it will break down the work correctly and make all the reasonable decisions quickly on the design, then go out and implement and deploy it?
We certainly live in uncertain times.