Substack Didn’t Launch a Feature. It Made a Confession.

There’s a version of platform compromise I can live with.

Every institution you operate inside makes decisions you disagree with. Tolerating that is the cost of participation. I’ve been doing it my whole career. This is not a purity test, and I’m not especially interested in the usual departure theater where one compromised platform gets exchanged for another and the writer calls the trade integrity.

But Substack’s Polymarket integration clarified something.

It wasn’t just a bad feature launch. It was a structural confession.

The old deal was vampiric: Substack takes a cut, surfaces your work to an audience, extracts attention. You know what’s being taken, and the extraction has a ceiling. You can live with a vampire because the vampire needs you alive.

Polymarket changes the organism. Once a platform imports a prediction market into the writing surface, it stops merely monetizing your work and starts using your work as substrate for someone else’s financial instrument. Your writing becomes carrier medium for speculation infrastructure. The parasite does not need you specifically. It needs tissue. If your publication dies, the parasite moves to the next host.

That is not a normal product evolution. It is a change in the extraction relationship: from “we monetize your audience” to “we monetize the environment your work creates, and your work is now a disposable input.”

If you want to see what that regime looks like in practice, look at the Maduro market. In early January 2026, a newly created Polymarket account built a cluster of positions around Nicolás Maduro’s removal and possible U.S. military action in Venezuela, then cleared roughly $410,000 when the raid became public. The obvious question was the right one: at what point does a “market” stop being a market and start looking like access asymmetry with a glossy front-end?

That is not an edge case. It is the diagnostic running in public.

Polymarket’s pitch to writers is that prediction markets make journalism more honest. Skin in the game. Accountability through exposure to reality. Numbers as epistemic discipline. Substack has already leaned on uptake among high-earning publications as evidence that this is working.

But the writing that actually holds up six months later tends not to look like market content. It tends to be slow, hedged, revision-friendly, willing to say “I don’t know yet,” willing to stay unresolved while the object remains unresolved. That kind of writing makes terrible market substrate. It does not produce clean binary arcs. It does not generate the crisp probability curves that make prediction markets feel like they are discovering truth. It resists resolution.

The writing that works for prediction markets is different. It is timely, confident, and oriented toward outcomes that resolve cleanly. Sometimes that certainty is earned. Often it is not. The format does not care. It only cares that the bet closes.

That is the epistemic problem. Prediction markets do not merely sit next to a writing environment. They exert pressure on the kind of writing that flourishes around them. They reward a style of thought that is prematurely legible, prematurely decisive, and prematurely binary. They are structurally hostile to slow uncertain work.

I doubt anyone had to say “let’s corrupt the epistemic environment.” Nobody needs to be that theatrical. The mechanism is more ordinary than that: trend pressure, investor pressure, product pressure, then a rationale assembled after the fact. Prediction markets feel rigorous in the way anything with numbers attached feels rigorous. Once somebody influential calls that “the future of media accountability,” the idea moves faster than the critique.

And that was the point at which I finally had to admit I had been making the wrong category error.

The problem was not just that Substack made a bad decision. Platforms make bad decisions all the time. The problem was that I had let a platform occupy the role of home when it should only ever have been infrastructure I do not control.

That is the mistake.

Writers on platforms routinely collapse two things that should remain separate:

Where the work lives.
How people find it.

Those are not the same layer. One is archive. One is distribution.

Substack had been trying to be both for me, and it was a bad fit from the start. My output is irregular, dense, and revision-heavy. Frameworks evolve. Models get updated. Threat assessments get refined. Papers accrete. Arguments get stronger or collapse under their own weight. That is not really a newsletter cadence. It is closer to an archive of versioned artifacts.

And an archive does not belong inside someone else’s business model.

So I changed the relationship. Eighteen papers on Zenodo, a live mirror at neutral.zone, revision history visible to anyone who cares enough to look: the move already happened.

The canonical work moved onto infrastructure I control: GitHub repositories, static mirrors, Zenodo releases with DOIs. Version-controlled, citable, permanent, owned. The work now exists independently of any platform’s continued good behavior. If I want a clean reading surface for people who do not speak git, that is solvable. But the underlying artifact is no longer hostage to a platform’s incentives.

That shift solved more than the Polymarket problem.

It solved the cadence problem: no more performing “newsletter” on a schedule that doesn’t match how the work actually develops.

It solved the revision problem: on a platform, substantive updates are often invisible edits; in a repo, revision history is part of the artifact.

And it solved the dependency problem, which is the only one that really matters. If a platform enshittifies, I lose a traffic source. I do not lose the work.

The parasite cannot feed on tissue that is no longer present.

This is the lesson I should have started with. Not “leave every compromised platform.” That just leads to endless migration theater. The real move is demotion. Stop treating platforms as homes. Stop confusing discoverability with custody. Stop letting someone else’s product roadmap define the persistence layer for your thinking.

I can tolerate platform compromise. What I no longer want is platform custody.

That is the distinction Polymarket forced into the open for me. Not because prediction markets are uniquely evil, though I do think they are corrosive in exactly the wrong way for the kind of work I’m trying to do. But because the integration made visible what the platform had become. I was not writing inside a neutral publishing tool that happened to have some bad ideas attached to it. I was writing inside an environment increasingly optimized for other constituencies, other incentives, and another theory of what writing is for.

My work does not fit casino logic. It is slow, uncertain, willing to be wrong without becoming a tradeable position. It needs revision history more than it needs virality. It needs custody more than it needs vibes.

So I’m not especially interested in flouncing off Substack. That would only repeat the same confusion somewhere else. The more useful move was to demote it. Substack, Bluesky, whatever comes next: these are distribution layers now. Edge caches with mood swings. I’ll use them tactically while they’re useful and stop pretending they are the place where the work truly lives.

That part is settled.

The archive stays where I control it. The platforms circulate summaries, links, and fragments. If they become hostile, I lose a route, not the corpus.

That is a much calmer way to relate to compromised infrastructure. Not purity. Not exit theater. Just custody, finally assigned to the right layer.