The previous post warned you about rhetorical patterns that indicate shallow LLM reasoning. That post was written with LLM assistance. So: did we practice what we preached?
I went back through it. Found seven patterns worth examining. Some are problems. Some are fine. One was deliberate.
The Catches
"Not in the content — in the rhetoric."
The exact pattern we warned about. Here, 'content' and 'rhetoric' aren't opposites — rhetoric IS content. The sentence sounds clever but the distinction it draws is fuzzy. A cleaner version: 'Patterns in how LLMs phrase things, not just what they say.'
"The point isn't to dismiss everything... It's to calibrate your skepticism."
Same structure, but this one holds up. 'Dismissing' and 'calibrating' genuinely are different actions. The reframe adds clarity rather than just sounding impressive. Not all instances of this pattern are bad — just the ones where X and Y aren't actually distinct.
"After running hundreds of analyses through multiple LLMs..."
'Hundreds' sounds authoritative but is unverifiable and probably inflated. Did we actually run hundreds? Define 'analysis.' Define 'run through.' This is the kind of hand-wavy credentialism LLMs default to. Should have been more specific or omitted.
"certain ways of presenting information that feel distinctly... LLM-ish"
The ellipsis creates a pause before the coined term. Stylistic choice, not substance-masking. Ellipses become a problem when they're used to skip over reasoning ('this leads to... significant implications'). Here it's just emphasis.
"Think of it as the difference between asking a random person on the street..."
We used this in the FAQ, not the post. Analogies are useful for explanation but dangerous as proof. This one was meant to illustrate, not to substitute for the actual argument (which appears above it). Still — analogies are where LLMs often hide weak reasoning.
"The Three Pillars Structure — as a section heading"
The post covers three main LLM-isms before the 'More to Watch For' section. Given the topic, this is either (a) a failure of self-awareness or (b) a deliberate wink. It was (b). Probably should have been four sections to avoid this entirely.
"Yes, the irony is not lost on us..."
Calling out your own limitations can be genuine or it can be a shield against criticism. Here: genuine. We acknowledged the contradiction and invited readers to spot what we missed. That's different from 'I know this might be wrong, but...' followed by barreling ahead anyway.
What This Exercise Shows
LLM-isms are hard to avoid because many of them are just... writing patterns. Humans use analogies. Humans use dramatic pauses. Humans organize things into threes.
The difference between "LLM-ism as warning sign" and "LLM-ism as neutral stylistic choice" comes down to one question: Is the pattern doing work that substance should be doing?
"Not in the content — in the rhetoric" fails this test. Remove the flourish and the sentence doesn't mean anything specific.
"The point isn't to dismiss... it's to calibrate" passes. The underlying claim (adjust your skepticism, don't throw everything out) survives without the structure.
A Note on This Post
I tried to write this without the patterns I was critiquing. Some observations:
- I avoided "Let's dive in," "Here's what I found," and similar transitional filler. Removing them made the writing feel abrupt at first, but tighter on re-read.
- I used specific verdicts (guilty / acceptable / intentional) instead of "nuanced" hedging. Some of these calls are debatable, but at least they're calls.
- I caught myself about to write "In other words" — which is often a signal that the previous sentence wasn't clear enough. Fixed the original sentence instead.
- The structure has four main sections (Catches, What This Shows, A Note, The Scorecard), specifically to avoid the Three Pillars trap.
The Scorecard
| Pattern Found | Verdict | Lesson |
|---|---|---|
| "Not X — it's Y" (fuzzy) | Guilty | Check if X and Y are genuinely distinct |
| "Not X — it's Y" (clear) | Acceptable | Structure is fine when distinction is real |
| Vague "hundreds" | Guilty | Be specific or omit |
| Ellipsis for emphasis | Acceptable | Fine for style, bad for hiding skipped logic |
| Analogy as illustration | Intentional | Analogies explain, shouldn't prove |
| Three-section structure | Intentional | Self-aware irony (debatable if successful) |
| Self-aware meta-note | Acceptable | Genuine acknowledgment ≠ shield against critique |
Final score: 2 guilty, 3 acceptable, 2 intentional.
Could be worse. Should be better.
Inevitable admission: This post has its own blind spots. I cannot see all my own patterns — that's the nature of the problem. If you want to write "The LLM-isms in 'The LLM-isms in LLM-isms,'" the floor is yours.