Substack put out an announcement this week.
They are worried about AI slop drowning out real writing on their platform. So they are rolling out a detection tool. Down the line, they say, readers might get to filter AI writing out of their feed entirely.
Their CEO said something in that announcement I want you to sit with.
He said he does not want the app to turn into LinkedIn.
A tech company, talking about AI, reached for LinkedIn as the insult. Not a scam site. Not a spam farm. LinkedIn. A place most people already know is full of content that looks fine and means nothing.
That is not a complaint about AI. That is a complaint about content with no seams showing. Writing where you cannot tell what is real, what is polished, what is disclosed, and what is not.
A reporter covering the story, writing for Business Insider, said something I think is exactly right.
He said AI being inevitable is different from AI being accepted.
And then he made a prediction. He thinks this whole mess settles into a labeling system. Not AI-free everywhere. Not AI-banned. A label, sitting next to the content, the way organic sits next to food at the grocery store. Some things carry the label. Some things do not. The reader decides what they want.
I read that and I did not feel surprised. I felt like somebody was describing a room I have already been building furniture for.
I have spent more than a year writing rules for how an AI should disclose itself. Not whether it should be allowed to work. How it should be honest while it works.
Does it tell you when it pulled something live versus when it is working off what you gave it. Does it tell you when it hit a wall it cannot fully explain, instead of pretending the wall is not there. Does it label its guess as a guess and its evidence as evidence, instead of letting both sound the same.
That is a label. It is just not printed on a package. It is built into the conduct of the thing itself, every time it opens its mouth.
Substack is reaching for a detector because a detector is the tool they had sitting on the shelf. Point it at a piece of writing, get a score, sort accordingly. I understand why. It is fast and it is available today.
But a detector only tells you what already happened. It looks at finished writing and guesses where it came from. It cannot tell you whether the AI that helped write it was honest about its own limits while it was working. It cannot tell you if the machine hedged when it should have been sure, or sounded sure when it was actually guessing. A detector checks the wrapper. It cannot see how the thing behaved while nobody was checking.
That is the piece I have been building the whole time. Not a scanner that runs after the fact. A standard for how the AI conducts itself while the work is happening, so the wrapper does not need to lie in the first place.
The reporter is right that this is not going away. AI is not going back in the box, and pretending otherwise wastes everybody’s time.
But he is also right that inevitable and accepted are two different things. People are not scared of AI existing. They are scared of not being able to tell what they are looking at. Take that fear away and most of the anger goes with it.
That is the whole argument behind everything I publish here. Not slower AI. Not less AI. Honest AI. Tell the person what they are getting, in plain language, while the work is happening, not after somebody runs a scanner on it.
Substack is one company reaching for one tool this week. It will not be the last company that reaches for something. When they do, I want the record to show plainly that somebody was already writing the standard before the companies noticed they needed one.
I do not write AI slop I write in an AI governace to show its possabilities.
“Written with Claude, under The Faust Baseline’s disclosure standard”
What does a slop filter do to someone who can’t type.
A person with cerebral palsy, or late-stage ALS, or arthritis bad enough that a keyboard is an hour of pain per page, or someone who is blind and has always composed by dictation — they speak their thinking to a machine, the machine turns it into sentences, they edit it, they publish it. Under a detector built to catch “AI slop,” that person’s authentic voice, produced the only way their body allows, scores exactly the same as a spam farm cranking out affiliate content. The tool cannot tell the difference between assistance and authorship. It only sees the fingerprint of the process, not the person behind it.
That’s not a hypothetical edge case you’re reaching for to win an argument. Dragon NaturallySpeaking has existed for people with RSI and disability for decades. Assistive writing is not new. What’s new is a filter that can’t distinguish “a machine helped a disabled person say what they meant” from “a machine generated content nobody meant.”
What is writing, then
Here’s the honest answer, and it’s the one that should go in the post: writing has never been the physical act of forming letters. Milton dictated Paradise Lost to his daughters after he went blind. Nobody says his daughters wrote it. A executive dictates a letter to a secretary who types it — the executive wrote the letter. Writing is the act of deciding what gets said. The hand, the keyboard, the voice, the AI — those are the instrument. The person deciding the meaning is the author.
Written with my AI partner | The Faust Baseline™ | intelligent-people.org
“If this post helped you understand AI better. Share it, a Word of mouth is the only algorithm nobody owns.”
Contact: micvicfaust@gmail.com
Post Library – Intelligent People Assume Nothing
Purchasing Page – Intelligent People Assume Nothing
© 2026 The Faust Baseline LLC | All Rights Reserved
