Through the Interface

Like everyone reading this blog post, I've been thinking a lot about Artificial Intelligence, of late.

In the software development domain the AI-assisted coding tools have seen a radical step change since the beginning of 2026, which has made some people very excited and has left others reeling. In this post I wanted to opine a little on this, as there are many things being said and even a few things that seem to have remained unsaid.

For those not in the space, what has happened since January? AI-driven software tools have gone from "interesting - I can see this is great for vibe-coding throwaway prototypes" to "oh hang on, I think I can trust these tools to handle huge parts of the software development process, even when working with large, complex codebases".

This step change is impacting markets: in the recent "SaaSpocalypse" significant amounts of investment dollars shifted from SaaS companies to AI-centric stocks. I think this is a bit short-sighted, as most users of complex software tools are not just going to redevelop them in-house, but there you go. SaaS companies are adopting AI tools, too, and very effectively, but besides that software development is ultimately just one piece of the puzzle when it comes to successfully delivering enterprise software.

In terms of the broader implications these AI tools will have on "the future of work"… while I dislike making predictions, I will say a few things as food for thought: there's mounting evidence that the more extreme view that significant numbers of jobs - and even professions - will disappear because of AI is unlikely to play out in quite that way. What's happening right now in radiology is very interesting and is being held up as an example of Jevons paradox (thanks to Patrick Nadeau for pointing me to that!). Even if the number of software engineering positions were to reduce, I wouldn't be particularly unhappy to see only the truly passionate choosing it as a career. More realistically I expect fewer people to go into the career in the near-term but the overall need for their services increase. We'll see how things pan out.

What seems undeniable is that many of our jobs are going to change radically over the next few years. This is certainly true for software developers but for other professions, too.

Aside from my day job leading a research area, I also manage a team of software engineers and there's very much a spectrum of engagement level when it comes to AI workflows. There are people in the team who are all-in on AI, who have been exploring agentic development workflows for more than a year (a very long time in AI years), while others are more resistant. Frankly I'm happy to have this breadth of opinion in the team: the early adopters bring a sense of excitement and work through many of the kinks, the laggards bring much-needed skepticism and point out kinks that others might have missed. Over time we're all moving forward with the tools.

This is a hard shift for some folks, though. For some years AI assistance meant glorified code completion, leaving the human very much in the loop. That seems to be shifting quickly towards the human managing the process by creating and maintaining detailed specifications (from user requirements to more detailed engineering specifications) that the AI can now be trusted to implement. This is a very different activity than hacking away at code, and for some it's going to be really hard to make that mental shift: many of us have built long careers crafting code ourselves, but increasingly it's looking like the code that AI generates will be considered much like any other compilation target (intermediate or executable files), e.g. you wouldn't go and hand-modify a binary file, you would fix the source code. Now it seems likely we'll be fixing the specification, and in many cases never even looking at the generated code.

There's still creativity in the process, and the potential for a deep sense of satisfaction when you're firing on all cylinders and conducting the AI's work in a productive way. But it's not the same: I have no doubt it activates different parts of your brain than when you're writing code artisanally - it just feels different. It's also shifting the way people get meaning or purpose from their role, and this isn't a transition that everyone will find comfortable - or even be able - to make. A compounding issue is that there's often a culture of optimism when it comes to new tech, at least in software companies. The less enthusiastic can sometimes feel marginalized, which is definitely something for managers to keep an eye on.

One other aspect that seems to be bubbling up is the mental fatigue people feel from extended AI-assisted coding sessions. If I had to guess I would say that it's the fact that more routine aspects of the coding process are no longer there to let the brain relax a little (this is just a guess, mind). You're "always on" in creative mode, and it's often a lot of mental load to deal with.

One member of my team recently spent 3 days on an intense exploration of AI workflows on a side project he's had on the back burner for some time. He's going to write up his findings in a series of guest posts that I'll publish, as he went on quite the journey, both in terms of the project itself and his own mindset. I think it will be a really interesting journey to go on with him, and I'm looking forward to it.

I'll wrap up with this post with a fun anecdote on the topic of AI.

I was chatting to a colleague recently about having written a blog post that day, and this colleague said "oh, so did I!". They went on to describe how they'd used ChatGPT to expand a set of bullets and even create a supporting image for it. I think the look on my face must have been quite something, because we both ended up laughing quite hard. But it just goes to show that we all have "our thing" that we enjoy so much that we couldn't imagine farming it out to AI. The trouble is that it's not always true that "the market" will reward you (or a company pay you) for this, which is where the Venn diagram representing the concept of Ikigai gets a bit squished, as the "what you can be paid for" circle flattens a bit.

Image taken from this site.

It's going to be an interesting challenge for those of us who feel they've been (more or less) operating near that sweet spot in the centre to stay there, over the coming years.

As Alvin Toffler famously said:

The illiterate of the 21st century will not be those who can't read and write, but those who cannot learn, unlearn, and relearn.

This has never been more true than today, and will no doubt be even more true tomorrow. Onwards.