LmCast :: Stay tuned in

Various LLM Smells

Recorded: May 28, 2026, 9:03 p.m.

Original Summarized

Various LLM smells | Shiv After Dark

Shiv After Dark

Home
Blog
Essays
Notes
About

Various LLM smells

28 May, 2026

Late last year I started writing a math blog and decided to use LLMs to polish/enhance my writing. The LLM generated writing obviously felt significantly better than my own writing. It had better vocabulary, interesting sentence structures etc etc. I swear it did not seem like AI-slop to me at the time. Then about 3 months later, I see the exact sentence structures appearing ACROSS THE ENTIRE F***** INTERNET. And what is fascinating to me is that ai-smell seems like an artifact that emerges across various AI assisted tasks that you can now easily recognize. A few examples that I've collected so far to show the "ai-smells" across two domains:
1. LLM writing (beyond the obvious em-dashes):Some picks from my math blog (now deleted) and the drafts that accompanied it
Way too many punchlines
"Humans trust symmetry because it feels like intelligence made visible."
"The Tiger fit the story. Jin-yong fit the physics."
"Symmetry becomes a trap."

Consecutive short sentences
"Yet the tilt is not an accident. It is the shape of the optimum."
"Then AlphaEvolve arrived. It had no preference for symmetry. No aesthetic prior. No instinct to preserve harmony."
"These examples are not decorative. They form a distributed argument."

"X is the Y of Z"
"Cringe is the visible signature of moving along a gradient you chose."

"ist not just X, its Y"
"solutions that do not merely satisfy the constraint but satisfy the aesthetic instincts"

2. AI generated websitesThe "JetBrains Mono" font
The "step" and bullets on every webpage with this exact font:
Exactly these button
These cards
This blinking-dot in a badge component

Footnotes
I'm not against LLM/AI usage for creative tasks. This is just me noticing things.

 

Powered by Bear ʕ•ᴥ•ʔ

The author, Shiv After Dark, explored the phenomenon of "ai-smells" observed when using large language models for various creative and content generation tasks. Initially, the author used LLMs to enhance writing, finding the output significantly improved in vocabulary and sentence structure, but soon noticed that specific linguistic artifacts began appearing across the broader internet, suggesting that the "ai-smell" is an emergent property of AI-assisted processes that can be recognized in different domains.

In the context of LLM writing, the author collected several specific patterns that characterize this effect. These include the overabundance of punchlines, particular sentence constructions, and specific relational phrasing. Examples from the writing domain included statements such as "Humans trust symmetry because it feels like intelligence made visible," "Symmetry becomes a trap," and the deployment of consecutive short sentences like "Yet the tilt is not an accident. It is the shape of the optimum." Furthermore, patterns involving relationship framing, such as "X is the Y of Z," "Cringe is the visible signature of moving along a gradient you chose," and the phrasing "ist not just X, its Y," were identified as indicative of machine-generated style. The author also noted a tendency toward solutions that satisfy aesthetic instincts rather than merely satisfying constraints.

Beyond textual output, the author identified similar artifacts in AI-generated website design. These visual smells manifest through the consistent use of specific aesthetic choices, such as the persistent use of the "JetBrains Mono" font across pages, the standardized implementation of "step" and bullet notations utilizing this font, and the specific styling of components like buttons, cards, and badge elements featuring a blinking-dot. The observation suggests that consistent stylistic patterns are also emerging in the domain of visual design when AI tools are involved.

Ultimately, the author states that this observation is not a rejection of using LLMs or AI for creative work but rather a personal notice of these emerging artifacts. The exploration serves to highlight the recognizable patterns that arise from current AI-assisted generation processes across both written and visual output.