Syntax That's Sus From Writing That Sucks

I loathe LLM-generated writing, especially when someone launders it as their own work, thoughts, or insights. LinkedIn is a prime example of such lazy token-generated text.

Lots of people point to syntactic tells of artificial expression, from the glorious emdash to an abundance of bullet points. Others point to stylistic tells, such as overused formulations like, “It’s not X. It’s Y,” or phrases that serve as filler between equally mundane or generic assertions.

There's an understandable desire to identify and avoid such writing. Many readers don’t want to waste their time on something carelessly generated. The glut of generated text crowds out human voices, a further insult given how text and image generation are derivative of copyrighted material collected without consent or compensation.

Hence there are quick judgments based on the presence of an emdash or the poker-like tells assumed in a word like “delve”. A few services purport to identify LLM-generated text, albeit with questionable accuracy and a methodology that seems to vary between grep with a few regexes and guessing disguised with decimal points.

That desire for avoidance necessitates quick judgment. It’s much easier to scan for patterns before wading into an article, only to realize it bears the tics of tokenization. Unfortunately, just relying on a particular punctuation is insufficient. It misses the weaknesses in structure, sense, and style that make human writing more appealing.

I haven’t changed my writing style, despite having embraced the emdash for quite a while. I don’t and won’t use LLMs in my writing process. I enjoy the process of thinking — and struggling — my way through a sequence of sentences.

I also see it as respect for the audience.

Revise, Refine, and Read

Bad writing and poor conference presentations predate people spending money to trickle their thoughts through a random number generator. Humans on LinkedIn have generated vapid business-speak for ages. But I can accept many of those human efforts as people experimenting with their own first drafts. I’ve gone through and continue to go through that perpetual process of draft, revision, draft.

Learning a craft like writing requires exercise. Improving it takes patience and experience. There’s nothing wrong with scattered thoughts and early drafts. It’s the context of how they’re presented or consumed that makes a difference.

Cybersecurity relies on writing, from explaining vuln research to describing secure designs to sharing knowledge to the occasional blog post about taking security seriously.

Reading is a craft. Reading that exercises critical thinking is crucial to cybersecurity. Threat modeling is a loosely structured application of critical thinking. That’s another reason why reading text spewed from an LLM is so annoying – the boring patterns and extraneous phrases offer nothing valuable in terms of intellectual exercise.

Cybersecurity also deals with noise. It's understandable to reach for patterns to quickly identify and avoid the spam and lack of signal in LLM-based writing. Cybersecurity also likes to deal in lists of things not to do, like don’t use string concatenation to build SQL queries and command-line arguments.

But if you take away all the exhortations of what not to do, that leaves a void of recommendations about what to do.

Write Often, Read Often

Cybersecurity likes patterns. It talks about programming patterns and regex patterns. It draws on TTPs as patterns to vaguely define threat actors.

If we get stuck on syntax and simple patterns, we risk framing writing as an absence of annoying characters. There are absolutely situations where that's useful. I don't want to waste time on the output of a prompt. That's because I'd rather give attention to writing that makes me think, that makes a topic entertaining, or that captures my attention throughout. And from there to build my own ideas of what kind of writing I want to create.

Similarly, I'd never bother with creating a Top 10 list for LLM prompts, even one that starts with something like, “Don’t use an emdash”. (Oh, how cybersecurity loves lists.) I'd say don’t use an LLM to write in the first place. It weakens communication and critical thinking skills.

Write about the steps you took to find a flaw. Write about the steps that failed to find a flaw. Write something you would enjoy reading.