It's almost as though the LLMs were trained on all the writing conventions which are used by humans and are parroting those, instead of generating novel outputs themselves.
They haven’t picked up any one human writing style, they’ve converged on a weird amalgamation of expressions and styles that taken together don’t resemble any real humans writing and begin to feel quite unnatural.
As someone who uses em-dashes a lot, I’m getting pretty tired of hearing something “screams AI” about extremely simple (and common) human constructs. Yeah, the author does use that convention a number of times. But that makes sense, if that’s a tool in your writing toolbox, you’ll pull it out pretty frequently. It’s not signal by itself, it’s noise. (does that make me an AI!?) We really need to be considering a lot more than that.
Reading through the first article, it appears to be compelling writing and a pretty high quality presentation. That’s all that matters, tbh. People get upset about AI slop because it’s utterly worthless and exceptionally low quality.
The repetitiveness of the shell commands (and using zig build-exe instead of zig run when the samples consist of short snippets), the filler bullet points and section organization that fail to convey any actual conceptual structure.
And ultimately throughout the book the general style of thought processes lacks any of the zig community’s cultural anachronisms.
If you take a look at the repository you’ll also notice baffling tech choices not justified by the author that runs counter against the zig ethos.
(Edit: the build system chapter is an even worse offender in meaningless cognitively-cluttering headings and flowcharts, it’s almost certainly entirely hallucinated, there is just an absurd degree of unziglikeness everywhere: https://www.zigbook.net/chapters/26__build-system-advanced-t... -- What’s with the completely irrelevant flowchart of building the zig compliler? What even is the point of module-graph.txt? And icing on the cake in the “Vendoring vs Registry Dependencies” section.)
Yeah and then why would they explicitly deny it? Maybe the AI was instructed not to reveal its origin. It's painful to enjoy this book if I know it's likely made by an LLM.
If you find it useful no harm in enjoying it! The main problem with AI content is it's just not good enough...yet. It'll get there. The LLMs just need more real-world feedback incorporated, rather than being the ultimate has-read-everything,-actually-knows-nothing dweeb (a lot of humans are like this too). (You can see the first signs of overcoming this w/ latest models coding skills, which are stronger via RL, I believe.) (Not first hand knowledge tho -- pot kettle black situation there.)
