As more people start using AI for research, they'll probably adopt AI'isms in their writing too, even if they don't use AI itself to do the writing.
Subconsciously we all pick up patterns that we regularly encounter.
As more people start using AI for research, they'll probably adopt AI'isms in their writing too, even if they don't use AI itself to do the writing.
Subconsciously we all pick up patterns that we regularly encounter.
I've had that happen to me. I actually recently wrote a draft blog and looked over it, and it felt AI generated. My writing style has definitely changed.
Hah absolutely, after of a very intense project I’ve undertaken with Claude agent teams for a few months now, I have caught myself using opus expressions, since they come to mind now naturally because that’s most of the text I’ve read since then … I’m catching myself so far, but in a year or two maybe I will not, fable doesn’t do as much “mannerisms” and speaks plainly, so maybe by then we will all be running fable level models on hour local machines and just not notice it anymore?
I've definitely picked up 'lever', I scrub it every time that I notice I write it but it's in there.
“blast radius” for me
Maybe we need our AI to only respond in bullet point lists.
With emoji for the bullets, natch.
Just put "Don't use emojis" into the prompt.
Pass it as a memory or a system prompt and watch it forget 3 agent messages later.
what frontier models are this bad these days?
I can't name one but I can tell you these models keep getting stupider as they keep existing and their system prompts get inflated with what not to do. I have experienced this happen with Qwen 3.8 Max Preview where for the early days it was blasting through Rust code fine and then before the release it silently got replaced and started to suck hard. Claude still has the "blue button" problem so it is probable that a casual chat with it could have this exact outcome of forgetting what to do. I saw some models also have this thing where when you tell them a text restriction they first ignore it in thinking, creating the full text and then apply them in their thinking rewriting it fully causing your output tokens to be inflated.
Well yes, there are only so many attention heads (well whatever magical variant frontier models are using these days) that can attend to the context and so as the context grows attention becomes spread thin.
But with reasoning enabled I find that even with a large context that induces mistakes most of the time a decent model realises and corrects itself before output.
And it's always been known that prompting what it should do is far better than what it shouldn't, since just introducing "DON'T do X" into the prompt means that the tokens for X are present and can be paid attention to in the wrong way.
But even then I've used plenty of "Do X, not Y" recently, especially for tools "Use this for x, don't use this for Y" and models perform like 90% of the time.
I would be interesting to experiment to see performance curves given a restriction on reasoning tokens allowed to n% of context tokens and see if there's some magic number of "reasoning should be at least n tokens for a context of length p" even ignoring the complexity of instructions in the prompt itself.
- The
- Distinction
- Is
- Important
Lucid only lets the AI give notes, not change your text: https://www.writelucid.cc/
But I still find that I have to consciously fight the urge to sound like an LLM. It's terrible.
I have the opposite, I can pick up AIsms much more now and they stick out like a sore thumb. A couple of years ago I had no such intuition, it just sounded like normal writing to me, but now if I try to use AI to write any prose it just sounds absolutely terrible.
This is the load-bearing result
You're right to call it out.
This distinction is important.