A frivolous prediction

This is a bit far-fetched, and probably wrong, but here are my observations and a frivolous prediction: the demand for software engineers will increase, not decline, as a result of Generative AI.

With the rise of no-code, low-code, and vibe-code tools, the number of people who can churn out code, and the cost of producing said code, is rapidly dropping. That does mean, for some tasks, you don’t need to hire a software developer. But it is also too early to call the end of software. The utopia is a world where we write code in English (or whatever human language we feel comfortable with).

The reality today is somewhat different.

We know a few things about LLMs: because they blend code (instructions) and data (context), they are inherently and incurably insecure. You can always prompt-inject, and you can only scaffold and filter to mitigate it; there is no way to re-engineer the LLM to completely prevent it. They tell you what they think that you want to hear - not the truth or the best way of doing something. Their usefulness is the inverse of their predictability: the creativeness that makes them valuable is reduced the more you try to control the outcome.

The result is machines that output large volumes of code that sort of works - great for prototyping and testing out new ideas, not so much for production workloads. Dangerous for anything mission-critical. Potentially deadly for anything that impacts human lives.

So we have more code than ever before, generated at an unprecedented pace, and somebody needs to make sure it runs securely, does the right thing, and does the same thing every time. Maybe tomorrow that can be done by a machine, or next year, or in five years - but until then, that’s a job for a human. And those jobs need doing, well, now.