congrats! you’ve just accused someone of using a tool you have no actual way to know whether or not they used

you have likely claimed (or clearly implied) something about the quality of their content based on the tooling they used in authoring it. note that LLMs aren’t writing and publishing things without people in the loop – someone is responsible for it. are you accusing that person of being lazy? of just using a tool you didn’t like? or do you actually take issue with the content itself? if so, say so! it’ll likely lead to a more productive discussion than baseless accusations on orthogonal points about the tooling used for authorship

LLM/AI checkers are fundamentally bullshit🔗

cat

did I write that? did I tab complete it with Copilot from neovim? did ChatGPT write it, and I copied and pasted it? did I have Claude Code draft a much longer sentence I edited down? did I use some other perhaps AI-powered authoring tool? how about this one:

cat

was that me or AI. guess!

you might be thinking, this article is written without proper punctuation! no way an AI wrote that!1!!!!1!!!!

you cannot tell. you cannot do math or statistics or use AI to tell with any reliability, you are fucking guessing. you cannot tell because you cannot distinguish – text is text. AI and I can write the exact same things. your statistical analysis is bullshit

LLMs do not exist without people🔗

LLM tooling + human = system.

an LLM cannot survive on its own. go throw one in the forest and see how long it lasts. go take hundreds of them and give them a city and see where they are in 100 years. it’s fucking nonsensical, in the same way LLMs doing anything without humans in the loop is

someone publishing content (or a merge request) authored that content, regardless of the tooling they used – whether it be a typewriter, a text editor with autocomplete, or some LLM-powered assistance. take issue with the quality of their content, not the tooling they used to author it, please, I beg of you

the system adapts🔗

moreover, people (natural language) and LLMs change over time. or rather, there are many LLMs and they get updated on new training data on different cadences. ongoing changes in natural language will be picked up by new models

LLMs are now affecting natural language. people are conscious of em-dashes, and words like “delve”, and other stupid shit because of people like you

so, whatever B.S. your AI checker is based on needs to handle this, and it can’t. it can’t because it’s impossible

focus on content, not tooling🔗

please, please, please, if you have an issue with content, clearly state what it is. baseless accusations about tooling someone may or may not have used, that you have no way of checking for, is juvenile

addressing stupid counterpoints🔗

first if you’re not convinced, take me up on my $1,000,000 wager after you critically think through the ruleset and how easily I will win, and what rules you’d actually need to negotiate to even have a chance

yes, a lot of people are quite lazy in their LLM usage and use them to pump out content for clicks. yes, we should ban paid digital advertisement. no, this correlation of lazy/bad people does not mean you can A) tell when people use LLMs for anything or B) have made a valid point in this instance