Categorising software jobs

AI is going to disrupt software engineering.

I’ve heard this so many times, and it’s too darn vague to be useful. What if I said “AI is going to disrupt electrical jobs”? What kind of jobs involving electriciy? Electricians? Electrical engineers? Factory workers who solder electronics? The type of discussion can vary greatly depending on what you’re talking about, and I doubt there’s much meaningful conversation to be had at such a high level of abstraction where all those careers are grouped together.

It’s helpful then, to have categories to guide these discussions. But we can’t lean on accreditation from governing bodies that we could do with other industries, because software is a new industry that has changed enormously since its inception.

Coming from an academic background, I’ve learned that you must be careful and deliberate when categorising. It is impossible to be unbiased when categorising. By its very nature, you are taking continuous data and drawing lines to put them into groups. There will be similar data points that you will be presenting as distinct.

This 3.95% sugar snack is very healthy, and this 4.0% sugar snack is terrible for you.

Says the analyst who grouped data by >= 4.0% sugar.

As such, I am hesitant to categorise without proper consideration of the limitations. But others have no such qualm, happily pushing whatever the latest trend innovation is on LinkedIn.

There are two types of engineers, those who understand agentic loops and those who don’t. Follow my advice for guaranteed success.

It’s so frustrating! What are these people trying to say? What are they hoping to teach? Under what conditions is your advice applicable, and when is it not? It seems like every day more people are talking like experts, without saying anything useful.

And it’s such a shame, because good categorisation is really powerful. It guides and focusses discussions, and can change how people think about the world. Unfortunately, it’s rarely done well amongst AI discussion.