Substack Can Now Score Your Writing For Authenticity
They found the will to build that in eighteen months. They still haven’t found the will to remove the Nazis.
Parts of this article was published on Medium, April 10, 2024.
Substack Can Now Score Your Writing For Authenticity
Before I launch this particularly salient reprint, I want a moment of your time.
Substack has been, for me, a wonderful experience. I’ve made connections here I wouldn’t have made anywhere else and written things I’m proud of. It has not been without flaw.
I learned white supremacists write here. I suspected they did; it was confirmed. What else needs to be said about insecure individuals willing to destroy their country in order to rule the ruins? Really?
I’ve heard women are harassed here. Women shouldn’t have to block assholes who lack the good manners to STFU in polite company. Worse, when they report it, Substack doesn’t do anything, because… because… hell, I don’t know. Bigots, racists and misogynists can write anywhere. They don’t have to do it here. Make them get a blog.
And most recently: Andrew Tate was a featured writer before I was. If you don’t know who he is, I wish I were you. As it stands, he and his brother sit in federal detention in Miami, held without release, facing extradition to the United Kingdom on rape, trafficking, and a battery of other charges that should have made Substack yeet them into the sun years ago.
Their attorney’s public defense is that they are “being punished and attacked for what they say and not for what they have done,” and that nobody should go to jail for free speech. Hold onto that sentence. It’s the same one every platform uses to explain why the monsters get to stay.
Trust me, I’m going somewhere.
All of this is reason enough to consider walking away from Substack. My reasoning for staying was simple: (cover the children’s ears) Assholes are everywhere.
Bill Maher is an asshole. I said it.
Is there a public sphere that doesn’t feature an asshole getting time and attention he hasn’t earned? Think about it. They gave Bill Maher a Mark Twain Prize. When was the last time that man said anything important that wasn’t conservative-adjacent, slyly racist, or desperately misogynistic? I guess someone has to get it.
My point? Assholes rule the world. Every medium we use to speak to each other is currently, or about to be, in the hands of a billionaire whose reason for being is to enrich himself while starving the people who make him rich. There is nowhere you can engage media where assholes don’t control the media.
Substack platforms people who are unfortunately everywhere and aren’t likely to change. So I stayed, the same way I stayed on Twitter longer than I should have, because I don’t let assholes determine where I stand, what I do, or how I do it. They never get a say in what I write. I define my work and what it says, independent of the asshole ratio of the environment.
Which Leads Me To The Source Of My Rant
Substack has partnered with a company called Pangram to scan what you write and tell your readers how much of it a machine thinks a machine wrote.
Let me say the generous part first, because I don’t want to be accused of hedging.
Pangram is good at what it does. Independent evaluation puts its false-positive rate low enough that the usual objection — these things flag innocent people — doesn’t carry the weight it carried two years ago. I’m not going to stand here and tell you the tool doesn’t work.
I’m going to tell you something worse.
It works, and it still cannot do the thing Substack says it’s for.
Chris Best’s argument is philosophical, not technical. He says readers are paying for a relationship with a person. He says the problem isn’t AI, it’s the mismatch between what a reader thinks they’re getting and what they’re actually getting. He is right about that. It’s the most honest framing anyone at a platform has offered on this question, and I want that on the record before I take the rest of it apart.
But a detector reads the artifact. The relationship lives in the process. And the process is not preserved in the finished words.
Two writers sit down. One drafts from nothing at four in the morning, after a day of caregiving, bleeding onto the page, and then runs a spellcheck. The other feeds a prompt to a machine, takes what falls out, and spends twenty minutes sanding off the tells.
Their finished paragraphs can converge to the point where no instrument distinguishes them — not Pangram, not Pangram 4, not Pangram 12. That is not a flaw in the algorithm. That is a limit on what text can testify to about its own making. You cannot build your way out of it, because there is nothing to build toward.
The score is not a measurement. It’s a vibe with a decimal point.
Which is the whole of it, stated plainly:
No detector can ever prove authorship from text alone. At best it estimates statistical likelihood. The original creative process isn’t preserved in the finished words. Two different paths — a human drafting from scratch or a human heavily editing AI output — can converge on nearly indistinguishable text. That’s a limitation of the evidence, not just the algorithm.
And just so we are clear: I am not saying AI disclosure shouldn’t exist.
I’m saying:
“If authenticity is so important that you’re deploying machine classifiers against your writers, why isn’t community integrity important enough to receive the same institutional energy? Imagine how the energy of this place would change if avowed racists, misogynists and other hateful mutherfuckers just weren’t allowed?”
Eighteen months. That’s roughly how long it took to go from there is no way to detect this to there is a button in your reader’s menu. Somebody scoped it, staffed it, negotiated a vendor contract, built the UI, shipped it on web and iOS, and wrote a manifesto to announce it.
The Nazis are still here. The harassers are still here. Those problems have been reported, documented, escalated, and written about by people with far more patience than I have. They remain unsolved, and we are told they are hard.
They are not hard. They are expensive in the only currency that matters, which is the willingness to lose subscribers you’d rather keep. Slop costs a platform nothing to oppose. Nazis cost a platform something. That’s the whole equation. Everything else is press release.
But It’s Optional, Thaddeus
Someone is already typing. It’s reader-initiated. You can switch it off per post. Nobody’s demonetizing anybody.
True. All of it. And entirely beside the point.
Nobody builds a scoring apparatus, watches it get cheap, watches advertisers learn to read it, and then declines forever to act on the number. That isn’t cynicism. It’s Platform Gravity 1. The observability layer always ships first, because it’s the part that costs you nothing politically. Enforcement is what observability is for. It arrives later, quietly, in a product update, framed as something the community asked for.
This does not require anyone at Substack to be lying to you today. Chris Best can mean every word he wrote. Coherence does not require conspiracy — only outcomes, and the persistence of the pattern is itself the evidence. The machine is in the building now. What the building does with it is a question of incentive, not intent, and the incentives are not a mystery.
I know they’re not a mystery, because I’ve watched this exact movie before, on a different platform, at precisely this point in the arc.
Let me show you the letter.
How We Got Here…
Which leads me to an article I wrote on Medium two years ago when they said the same thing.
Not a single aspect of the argument has changed between then and now except the number of generative AI, large language models (GAI-LLMs) have proliferated and caused CEOs everywhere to believe they should fire tens of thousands of people because the technology appears sophisticated enough to fire people and allow the remaining people to pick up the slack with the GAI. This is a mistake. 95% of such implementations have failed, and almost always in a disastrous fashion, but that is an article for another day.
Today, let’s look backward to see forward and recognize the challenge cannot be met by verification which cannot be verified and authentication which is at best an imperfect science. Let’s turn on the Wayback Machine… When you see the word Medium, you know to substitute Substack and you will be fine.
Hey Medium, How Will You Know For Sure?
Good morning, dear readers. You haven’t heard from me as much as I used to write here for a number of reasons:
Our society has decided to double down on making technology no one needs. Instead of creating tools to remove drudgery from our lives, we are instead creating tools to remove creativity and joy from our existence. This is the center of the topic we are going to discuss today.
Artists around the world are watching as big and powerful corporate entities claim they have found a way to increase productivity and reduce manpower using large-language model generative AI engines. These tools are supposed to be able to write better than Humans, create art faster than Humans and no one has to pay them a living wage. Let’s be honest, they weren’t paying living wages before generative AI, they don’t appear to be trying to change that, now do they?
As a creator of the ‘leaves of language’ I am annoyed with the number of generative AI articles being thrown at the world and appearing in every medium possible, from analysis to marketing, we have no place where such writing can not be found. My experience on LinkedIn of late is nothing more than dozens of new generative AI “prophets” promising you will be better, faster, stronger if you adopt this technology today. Worse, these tools are being integrated into every digital application you can find, integrated into browsers, into operating systems, and into graphic design programs. In my estimation, it will take less than five years for you to be unable to escape these tools unless you are writing in a text editor not connected to the Internet.
As a result, disreputable people are not writing, they are letting machines write FOR them and this is disconcerting. Because writers, actual Humans, who want to have something to say, have experiences worth noting, with philosophies worth sharing, are hidden away behind constructs designed to interface with the algorithms, with the right words, with the most algorithmically accessible phrases, sure to tickle the search engine optimization so that their SCUM rises to the top.
HOW WE GOT HERE:
With that preamble I will share with you the letter I received this morning from Medium:
Hello,
We are writing to notify you of policy updates that may impact your participation in the Medium Partner Program.
Medium is for human storytelling, not AI-generated writing. We recently defined and clarified our specific policies around the different uses of AI-generated content and technologies, and what is allowed in the Medium Partner Program. You can read those policies in our Help Center.
Beginning May 1, 2024, stories with AI-generated writing (disclosed as such or not) are not allowed to be paywalled as part of our Partner Program.
Accounts that have fully AI-generated writing behind the paywall may have those stories removed from the paywall, and/or have their Partner Program enrollment revoked.
Please remove all AI-generated writing from behind the paywall by May 1st to remain compliant with this policy, and stay active in the Partner Program. We define AI-generated writing as writing where the majority of the content has been created by an AI-writing program with little or no edits, improvements, fact-checking, or changes. This does not include AI writing tools such as AI outlining, or AI assisted fact, spelling, or grammar checkers.
Additionally, stories with AI-generated writing will also have the following distribution restrictions:
AI-generated writing that is disclosed as such and is not behind the paywall is eligible for General Distribution, but it is not eligible for Boost distribution.
Undisclosed AI-generated writing will be given Network Only distribution. This means it is distributed to the author’s direct network of followers and subscribers only. It is not eligible for wider distribution.
AI-generated images must include captioning identifying them as such.
There are also new policies for AI content used for abusive or illegal purposes. We encourage you to read the FAQ and the AI policies in full, and reach out to us with any questions via our Help Center.
Thank you,
The Medium Partner Program
A Very Nice Letter but:
I do not fear any aspect of this letter or its ramifications. All of my writing has been and will be 100% me, all the time, unless I am mocking generative AI. Then I will tell you that I am.
But I have to ask, having spent quite a bit of time writing about generative AI, — not here, because there is already so much of it and my work doesn’t get the coverage it should so I don’t bother — how exactly do you plan to determine if someone is generating their work through artificial, generative AI, means?
There are no reliable programs which can determine with 100% accuracy whether a piece of written work, is artificially generated. I know. I have tested them. Not all of them. And new ones are being created every day. Yes, there are ways to recognize such things but they are time consuming and this would surely add a layer of complexity to an environment processing THOUSANDS of articles per day…
The irony is not lost on me that the only way to be sure is to PAY a company to write a program to detect if the text being used is generated by a program. And such tools may not always be accurate. Here is where the rubber meets the road:
What happens to a writer if they write something and a program is incorrect in its assertion of authenticity? What happens if a machine is wrong? Are we going to punish people if their work doesn’t have the “joie de vivre” to pass this test?
Why do I ask? Because I am a writer, a prolific one with over three thousand articles on the Internet, not a single one of them uses generative AI to WRITE them. I have used generative AI to create graphics, only so that I do not use other artist’s work without their permission. It’s a tough world out here.
Once upon a time:
You could write an article, pour yourself into it and if your craft was up to par, if your creativity tapped into the Platonic realm where all things wait to be created, if your work could touch the emotional centers of your readers, you might, if your work was truly meaningful, connect to a reader, who would find your work exactly what they were in need of; whether it be inspiration, solace, darkness or a path toward the light, a writer who loved their craft was sure to find an audience waiting for them.
Then search engine optimization became a thing. The writers created for the Algorithm, wrote for the Algorithm, created headlines sure to manipulate the reader’s fear of missing out, headlines designed to titillate, not stimulate, articles meant to feed the need to be connected to the seediest, least useful, most exploitative aspects of this society, articles which remind us we are not enough, we don’t know enough, we don’t have enough, we are not enough, and only with this article created to assuage our advertising-addled minds, promising us the Rosetta Stone to assured success, can we be cleansed.
Then we made an algorithm that can do all of that and still be home in time for dinner.
The millions of words I have wrangled to become a viable, meaningful wordsmith, the efforts and articles crafted here and on Quora and on the Science Fiction Stack Exchange and on Facebook, and my various blogs have no meaning now.
Because clout-chasing, algorithm-appeasing, talentless hacks inundate us with hundreds, thousands, tens of thousands, nay, MILLIONS of these artificially-generated, algorithm-manipulating, FOMO-curating, productivity-hacking, bullshit articles which demonstrate no skill, no emotion, no connection, just a genius at manipulating the ALGORITHM, drawing eyes, altering the landscape so that they can appear as giants, casting shade with their mirage of creativity while real writers, not understanding what is happening, languish unseen, unrecognized, unremarked, except by other Artists who see them, who understand just how difficult this path we walk truly is.
I admit to having walked away from Medium for a while because I could see the productivity articles, every one of them just like the other, drawing thousands of views and yet saying absolutely nothing, while my missives on the Human experience went unread because I didn’t do my level best to manipulate the Algorithm.
I wanted to write. Not take a course on search engine optimization. Or on the care and feeding of my favorite generative AI. Or on how to manipulate people into reading my work with the best headline which says nothing and intrigues folk into opening my work. (But I did take a course in all of those things and simply refuse to use them. They add no value to my work, only bog it down in ALGORITHMIC-ASS-KISSING.)
My writing is already able to be everything a growing body needs without those technological byproducts used by hawkers of lazily generated, lifeless, repetitive nonsense.
Okay, I sound hurt. And I am. I had never intended to make a living as a writer. I spent 30 years in information technology and loved every minute of it. Until technology and social challenges made my opportunities non-existent. Then I turned to my love of writing as a balm to the changing society, to note the inequities I had seen forming all around me. The same inequities which now surround everyone I know, and permeate every aspect of our existence. Like generative AI and the tech-bros who bring us technology we don’t need while ignoring everything meaningful in life.
We are beset with catastrophes they don’t seem to have any interest in solving though they have the capacity and money to do so. Problems such as: economic inequality, climate change, ecological collapse, social unrest, mass Human migration, rising temperatures, crop failures, price-gouging, corporate greed, the list goes on and on. No one wants to solve those problems. They’re hard. They don’t make the kind of money they hope to make by exploiting us.
Instead they want to take away the livelihoods of artists by creating substandard, re-creation, plagiarism engines, unable to produce new ideas, only regurgitate old ones, producing texts which are little more than the empty calories of fast food franchises, whose products look like food, but ultimately aren’t.
Employees of Medium:
I don’t envy your decisions. With no easy way to determine who is writing what, we are poised upon the edge of an abyss. I am uncertain if your agency can do anything to stem this mechanized tide of algorithmically-generated nonsense. I just wanted to let you know as a writer who has never made a lot of money with your agency, a writer who has never used such tools and never will (for my writing) I see the future ahead and it is bleak.
Those writers who have used this forbidden technology you now claim you hope to avoid, those who have been gaming the search engine optimization system, who have created empty articles for years, whose work is unremarkable and inhuman, already dominate your space, bringing nothing to bear on the true challenges of our mortality, our struggles to find ourselves, to be our best selves, to create, to resonate, to believe in the ideals of our better angels, those of us who have written with the best of our hearts and minds, we have already been lost.
We have watched the lion’s share of your funds go to the very same people who will use technology to appease technology, exacerbating the truism of our society: Only the rich, get richer. The rest of us get meaningless technology which does not help us, it only dispossesses us, pushing us from opportunity, hoarding everything meaningful, and while they celebrate their genius at having gotten over, again.
The rest of us have already been forgotten. Our words never rise, not because they aren’t meaningful, but because they cannot be lifted without becoming part of the nightmare we see everyday here.
From time to time, I will likely visit Medium, to read those writers whose work I love, whose words come from a place of humanity, and who take pride in their ability to share the Human experience. I will continue to watch the productivity hawkers make the bulk of the money I see being made here, saddened by the loss of another opportunity scoured away by the ever-increasing reach of the Algorithm. I will miss the Medium I knew when I first came here, a place where people spoke their truth to power, their life to meaning, their experiences to share in the hopes of inspiring others, I will miss it when it is gone.
I hope you find a way to remove those technological vampires from the service, but if history is any indication, the monsters almost always win in the end. See the ravages of ‘Big Tech’, ‘Big Oil’, ‘Big Pharma’, ‘Big Tobacco’, for references supporting “the Monster Always Wins” premise.
By then, it will be too late for anyone to do anything about it. Every false positive doesn't merely misclassify a document. It transfers the burden of proof onto the writer. The author must now defend work that required no defense yesterday. Authentication systems that cannot reliably authenticate do not increase trust. They redistribute suspicion.
To my fellow wordsmiths, don’t you dare quit. Find someplace willing to appreciate your brilliance, because you are.
Footnote: Platform Gravity (Maxiom Index)
Every capability a platform builds is eventually used for everything it can be used for. Not because anyone plans it, but because a tool that exists is cheaper to extend than to leave idle. Observability precedes enforcement; the score precedes the consequence. The gap between them is measured in quarters, not principles.
I am the Cognitive Dissident — Thaddeus Howze