I’ve been reading this article on Martin Fowler’s website, and it got me thinking.
It’s been months since I published a note on this blog. That’s not really a problem: I do it just for vanity out of habit, without much of an audience. Still, I feel bad about letting things lie neglected. Over the past few months, I’ve written plenty of notes in my private Obsidian repository, but I’ve been too busy to choose a few, polish them, and publish them here. At the same time, my use of LLMs at work—for coding and documentation—has grown dramatically.
So why shouldn’t I use LLMs for blogging, too? Sure, an LLM detector might spot the AI watermark, and my writing might sound less natural and personal—though perhaps also less prone to errors, since English isn’t my first language (let’s pretend native speakers don’t make mistakes). On the other hand, I’d publish more. I’d also refine my publishing pipeline, learn a new skill, and (who knows?) maybe a random reader would benefit from an article I wouldn’t have published without an LLM’s help.
After all, if you’re looking for good prose with a distinctive voice, you read a book. A blog like this is closer to documentation for a cloud service deployment or instructions for a washing machine. This isn’t literature, so using an LLM shouldn’t be a problem. More likely, it’s an opportunity.
What concerns me more is the use of AI in recruitment and job hunting. There’s been a sharp increase in candidates using LLMs and recruiters using them to evaluate candidates. Intuitively, this means job seekers are getting better and more productive at crafting polished CVs, while recruiters are improving their filters to cope with the growing volume. In other words, recruitment is becoming a market for lemons: good candidates get lost in a sea of LLM-generated, picture-perfect CVs, and both candidates and recruiters lose out. I wonder how that will end.