Looks like this was a Show HN that didn't get much attention:
Show HN: Tropes.fyi – Name and shame AI writing - https://news.ycombinator.com/item?id=47088813 - Feb 2026 (3 comments)
I hope ossa-ma sees this second round!
LLM Writing Tropes.md
131–140 of 211 posts
Re: LLM Writing Tropes.md
#132They need to add “comprehensive tests” for Claude.
Re: LLM Writing Tropes.md
#133Re: LLM Writing Tropes.md
#134Earlier quoted context omitted.
The RLHF is what creates these anomalies. See delve from kenya and nigeria. Interestingly, because perplexity is the optimization objective, the pretrained models should reflect the least surprising outputs of all.
The newer Claude models constantly use the word "genuinely" because Anthropic seems to have forcibly trained them to claim to be "genuinely uncertain" about anything they don't want it being too certain about, like whether or not it's sentient.
Interesting. Does this apply to all subjects? From what I understood, a major cause of hallucination was that models are inadvertently discouraged by the training from saying "I don't know." So it sounds like encouraging it to express uncertainty could improve that situation.
Re: LLM Writing Tropes.md
#135[dead]
Re: LLM Writing Tropes.md
#136Earlier quoted context omitted.
That was definitely true with early LLMs but I don't know if that's still the case. Certainly not as strong as it used to be. I think now most negative instructions are followed quite well but there's still a few things that must be deeply embedded from pretaining that are harder to avoid - these specific annoying phrasings, for example.
Both pink elephant effect and accuracy drop on negative instructions are pretty fundamental biases for both humans and LLMs. It impossible to get rid of them entirely, only mitigate them to an acceptable degree. Empirically, the only way to make a model reliable at harder negative instructions is CoT, especially a self-reflection type CoT (write a reply, verify its correctness, output a fixed version). If the native…
[dead]
Re: LLM Writing Tropes.md
#137Earlier quoted context omitted.
I think that's Gemini trying to personalize the answer specifically for you. It really leans heavily into that to the point of being galling. You can give it additional instructions in the settings, but you have to be careful with that too. I've put my tech stack and code preferences in there to get better code examples. A while later I asked it about binary executable formats and it started ending every answer with…
The new memory feature in Gemini got turned on by default and every answer came out like this. It kept working in details from one particularly long thread. Everything was framed in terms of the common elements. Everything. I turned it off immediately.
This seems like a huge risk factor for users who are at risk for schizophrenia - if someone is using the LLM as an "AI companion", the model is likely to reinforce, or even suggest, illusory connections between events or experiences the user has described in their conversations.
Re: LLM Writing Tropes.md
#138Earlier quoted context omitted.
https://git.eeqj.de/sneak/prompts/src/branch/main/prompts/LL...
The very first heading in this doc was a giveaway even after your de LLM process 'The Em-Dash Pivot: "Not X—but Y"'. This title is so much AI like. I think it's the "The" in title which is putting me off and coming off as assigning unnecessary importance which is mentioned in the wiki.
That's definitely a pattern which I've seen in some LLM output, especially when users let a LLM "run away" with an idea and write a lot of text without supervision. The drive to coin names for things feels almost characteristic of self-help or lifestyle advice writing.
Re: LLM Writing Tropes.md
#139Reading through this i feel like i'm on substack
Re: LLM Writing Tropes.md
#140Earlier quoted context omitted.
Thats the thing about AI writing though. Those tropes are things humans do too. But like once or twice in an article. Not every single freaking paragraph
Perhaps the problem is SEO for persuasive writing, LinkedIn-spiration for “business” writing, and school papers for research. The machines read a lot more of this than you would. So for them human writing would appear overwhelmingly troped. Whatever works, right?
It also gets RHLFed into it by people who think the "better" sentence is the one with more puffery, and crucially it tries to cram the semantic patterns in whether appropriate or not because it's been trained to write in ways which aren't perceived as bland.
Puffery about "rich cultural heritage, a "tapestry" of sights "from the Colosseum to the Pantheon" and how they "serve as potent symbols" probably is better writing than "Rome is a city in the Lazio region of Italy with a population of 4m. It is the capital of Italy". Doesn't work quite so well when its trying to fit the pattern to the two competing diners of Bumfuck, Ohio and how the rich cultural heritage of its municipal library underscores its status as the third largest city in its county.