Bluesky didn’t reproduce Twitter’s ownership structure

Yesterday I watched a familiar little internet ritual play out on Bluesky.

Someone made a claim that was directionally right, rhetorically unstable, and therefore doomed. Within minutes, the replies had converted a structural argument about automation, labor, and externalized harm into washing-machine discourse. One person was talking about how productivity gains fail to become freedom under existing arrangements; everyone else rushed to prove that appliances exist.

It was a clean specimen.

::: {.bluesky-wrap .outer style=“height: auto; display: flex; margin-bottom: 24px;” attrs="{"postId":"3mgoeyh5u5k2a","authorDid":"did:plc:rffkyd3hlw5m7vheifn5qode","authorName":"Dr. Hilary Agro, PhD 🍄","authorHandle":"hilaryagro.com","authorAvatarUrl":"https://cdn.bsky.app/img/avatar/plain/did:plc:rffkyd3hlw5m7vheifn5qode/bafkreibxjyx3abze7x4rqfefaq23giq3s75jsrgc2jfw42dvj23ujxltfe","text":"It’s bizarre to me how many people in the comments here seem to think that automation is going to translate into more freedom for the working class.\n\nWhen has that ever panned out? How much leisure time do you have, after years of these \"innovations\"? We are Charlie Brown trying to kick the football","createdAt":"2026-03-10T03:01:36.547Z","uri":"at://did:plc:rffkyd3hlw5m7vheifn5qode/app.bsky.feed.post/3mgoeyh5u5k2a","imageUrls":[]}" component-name=“BlueskyCreateBlueskyEmbed”} ::: iframe ::: {#app} ::: ::: :::

The original thread’s claim was not actually hard to understand. It was overstated, sure. It was written in a moral and testimonial register rather than a neat analytic one. It carried some extra spiritual cargo. Fine. But the core point was legible enough: technological “innovation” does not, by itself, translate into leisure, dignity, or freedom for the working class. Under existing arrangements, labor-saving gains tend to be privatized upward while the harms are externalized outward. Convenience here, extraction there. Efficiency for firms, poisoning for somebody else’s land and lungs.

A normal person could read that and say: yes, there’s something real here, though the framing overreaches. Instead, the replies performed a more familiar operation. They shrank a structural claim into a trivial empirical one. You say automation has not broadly delivered freedom? Ah, but what about the washing machine. You say “innovation” is not neutral? Curious, since refrigerators exist. At least one person even reached for the cotton gin, which is almost too perfect: a canonical example of a labor-saving device intensifying the surrounding extraction regime was being offered as proof that technology is not implicated in labor politics.

The point was not rebutted so much as compressed into something easier to kill.

This happens constantly now. Not just on Bluesky. Not just in political argument. It has become a general reading failure online: classifier first, object later. The reader is less interested in what is being said than in what kind of person would say it, what team it affiliates with, what implication can be extracted from it, what weak seam can be widened until the whole thing rips.

Writers know this, which is why so much public language now arrives wrapped in what I think of as the poster’s litany.

Poster’s litany is the ritual pre-buffer people learn to perform before making any claim that might attract hostile parsing. Obviously not all technology. Obviously I’m not saying we should live in caves. Obviously there are exceptions. Yes, washing machines exist. Yes, some innovations really do improve life. Yes, this is more complicated than one post can capture. The point of these little throat-clearings is not always clarification. Often it is prophylaxis. They are booster-stage wrappers meant to keep the actual claim alive long enough to survive contact with the feed.

And still, the buffer rarely works.

Because the deeper habit is what I can only describe as adversarial semiotics: a style of reading that approaches every statement as a suspicious object to be classified, neutralized, or weaponized. Too neat, therefore fake. Too legible, therefore calculated. Too aligned with a social reward, therefore cynical. Too stylized, therefore pose all the way down. This is not wholly irrational. There is plenty of strategic performance online. Plenty of aestheticized sincerity, symbolic rent-seeking, moral theater, affiliation laundering. That is precisely why the reflex survives: it parasitizes real pattern recognition.

But after enough time in the feed, the pattern recognition becomes a script. Everything arrives pre-contaminated by the suspicion of audience. No statement can simply be mixed, partial, or poorly phrased while still carrying something true. It must either be a flawless crystal of universal truth or the product of a fraud. One bad sentence is treated not as evidence of human roughness but as a total diagnostic. A claim can be 70% right, 20% overheated, 10% mushy—which is to say, recognizably human—and the machine will still reward the cheapest possible move: find the weak sentence, flatten the argument, produce a countertop-level counterexample, declare the worldview refuted.

This is why so much online argument now feels less like reading than prosecution. The signs are already arranged in advance. The reader’s job is not to interpret but to convict. And a platform that trains prosecutorial reading will eventually reward writers who learn to compose for prosecution.

You could call this a kind of literacy, if you were feeling generous. People who spend enough time online do develop real skills. They become faster at recognizing tone, trope, subtext, dogwhistle, recycled phrasing, familiar rhetorical templates. There is some truth in the idea that heavy posting produces a heightened sensitivity to language. But “hyperliteracy” is the wrong word for it. The thing being trained is not patience, depth, ambiguity tolerance, or sustained interpretation. It is high-speed pattern familiarity under compression.

You get very good at reading posts. You may also get worse at reading texts, situations, institutions, or people.

Bluesky is useful here not because it invented this pathology, but because it demonstrates how little better governance alone can do to cure it. That is what makes the current moment on Bluesky mildly depressing in such a recognizable way. The platform has some genuinely interesting institutional ideas. The most important is probably the one Jay Graber has repeated for years in various forms: the company is a future adversary. Build for exit. Build for portability. Build assuming eventual capture. That is a serious governance doctrine. It is much closer to how institutions should be designed than the usual Silicon Valley theology of benevolent founders and permanent trust.

Very few companies build from the premise that successor sovereignty is a problem to be constrained rather than a right to be expanded. Bluesky, through ATProto, at least tried to put some real effort into that principle. Accounts can move. Identity is more portable. Alternate clients and app surfaces can exist. The company is not meant to be the final sovereign over the social graph. This is good. More institutions should think this way.

But there is a limit to what protocol alone can solve.

You can design an escape hatch from platform ownership. You cannot, by protocol alone, untrain the interpretive habits formed by years of posting under compression. Better substrate does not automatically produce better readers. Better governance does not automatically produce better cognition. A social graph can be portable while the same damaged reading style migrates intact.

That is the more interesting failure mode. People keep treating Bluesky as a platform story: did it replace Twitter, did it fail to replace Twitter, is the user growth there, is the vibe gone, is the protocol enough of a differentiator. Those are real questions, but they miss something nastier. What has reappeared on Bluesky is not simply Twitter’s incentive structure in miniature. It is Twitter’s reader: the posting-conditioned subject formed by years of speed, suspicion, affiliation sorting, and adversarial compression.

That subject does not need Elon Musk to exist. It does not need quote-tweets in exactly the same form. It does not need identical moderation policy. It can be reproduced anywhere the userbase is large enough, the pace is fast enough, and the reward structure favors instant classification over slow contact. Once enough people carry that reading style into a platform, the old runtime boots back up. One person says something partly true in the wrong register. Another person spots the unstable sentence. Then the whole thing gets routed into team sorting, gotcha extraction, and little commodity-grade rebuttals that make everyone feel like they have participated in epistemic hygiene.

This is why so much discourse about “better platforms” has a slightly unreal quality to it. Ownership matters. Governance matters. Yes, even protocol matters. I would much rather inhabit a network designed around the premise that the operator is a future adversary than one designed around a billionaire’s whim. But none of that gets at the deeper issue, which is that posting itself has trained people into a style of reading that mistakes suspicion for discernment and compression for clarity.

The woman yesterday said something overstated and morally saturated about automation. The replies said, in essence, “washing machine.” That exchange contains more of the current internet than most official platform analyses ever will.

Bluesky did not reproduce Twitter’s ownership structure.

It reproduced Twitter’s reader.