Twitter is compromised. What comes next? A plan for zero-knowledge anonymous reputation.

The most interesting and ideologically cutting edge Twitter posters and demographics are still systematically suppressed and often lose their accounts. Elon pretends ignorance but may have made a deal with the advertiser cabal to continue the old network disruption policy and promote populist fyp sludge in exchange for deescalation. Mass report brigades also play a role on key topics. Clearly we need better platforms.

Here's my solution: a fair algorithm where you keep the clout you earn and no one can take it away from you or censor you. You might get hit hard and lose a bunch in one engagement, but never more than you know you're risking. You can't get shadowbanned because the server doesn't even know who you are. Dox risk is minimized by multiple identities and on-tap anonymity, and there being no identifying records on the server that even could be leaked. With those measures, friendly moderation would be a cherry on top, not a last resort.

I'm going to call it zero-knowledge anonymous reputation: the principle is that all reputation information is held client side as anonymous spendable notes, nym keys, and zero knowledge merit credentials. When you post, your client stakes some of your reputation notes, maybe invokes some of your credentials with zero knowledge proofs, and spins up a new nym. As far as the server or anyone else knows, you could be anybody. De-anonymization and user tracking becomes extremely difficult: they would have to hack your client or get you with stylometry.

Moderation is still possible: the server controls the reputation algorithm that defines who gets the notes and credentials, and what kind of behavior burns them or earns them. But client-owned anonymous reputation makes this fair and transparent: you're not getting banned, just losing a stake. There is no two-tier policing because the mods or server can't tell what tier you're in except by exactly what you did. But the mods can still shape what kind of discussion they are trying to drive on the platform. But their options for cloak and dagger network suppression become limited.

I'm sure in the hands of the trust and safety operators at Twitter, this could still be made to suck. A truly hostile platform could just move away from it. Mostly it's a credible signal from the platform of commitment to real anonymity, real fairness. It's a supplement to friendly moderation. It's also a credible signal to would-be hackers, discovery lawyers, and legal censors: we don't have that kind of information, nor the power to do what you are asking. Go play stasi with someone else. It also takes us closer to a fully censorship-resistant protocol so that if push came to shove with a regime hostile to free speech, it would be easier to go there than back in the box.

I've mostly finished the design for a next-gen sofiechan reputation and identity protocol that would have these properties. I didn't used to think it was worth it: most of what we're banking on here is *social* anonymity and friendly moderation. But it has recently seemed more possible and is now on the long-term roadmap.

Of course the short term roadmap (if we ever get the time again) is still dominated by usability and operational improvements. But I'm curious if full technical anonymity would as exciting for you as it is for me.

Very interesting.

On a tech level, I love it.

For the short term: I think the 80/20 of unlinking identities from email login after, say, a month would go a long way. Implementing that is a single Claude prompt away vs frontier applied crypto research for full anonymity.

I’m curious who’s still getting banned on X. There are plenty of high volume hard right accounts. There are self identified NatSocs like Emily Youcis and 0hp. There’s also no shortage of race/iq and race/crime posters, some quite high quality. Beyond cosmetic rules like not writing “nigger”, what are the red lines in practice?

>>5223
Not quite as simple as one claude prompt. The current system actively relies on the account linkage. Architecture change either way. But yes we probably get to unlinking identities first. I think prototyping the new algorithm will be the time to do it and can already involve massive improvements to privacy. But the big fancy tech version would be very cool long term.

As for who gets banned, i think the best summary i’ve seen is edgy anime youth. Also, big known accounts like youcis and 0hpl don't get banned because they are the network anchors used for surveillance or honeypot or whatever. For whatever reason the praxis seems to be to ban the small and midsize accounts. The infractions are not particularly out of the ordinary afaict. I think the best model is actually a semi-sophisticated adversary doing targeted bannings rather than just dumb rules.

Even if you don’t like those guys or can tolerate the current order the principle of it matters too. We feel relatively free right now but one legal challenge to elon’s rule at twitter under president AOC and we’re all back in the cuck cage. We need politically and technically robust freedom of speech, privacy, and access, not just friendly mods and government right now.

There's two threads here, clout and zero-knowledge nyms.

I still fail to see the vision with clout. What do I use it for? Does the average Sofiechan number care if number is bigger? Or do they care that they read something insightful or novel? Is a small Twitter account producing ideological cutting edge content in a hope to grow their clout, or are they moved by what they feel is true? Twitter is a pretty good case for the two being anti-correlated.

> You might get hit hard and lose a bunch in one engagement, but never more than you know you're risking.

Having to make a decision about "risking my clout" before making a post is terribly annoying friction if your userbase doesn't care.

Obviously I'm not the target market for the type of anonymity you're aiming for. Nothing against those who want it, it's a useful feature. But: a pure anonymous board has no memory. On a traditional forum if I'm arguing with someone I've argued with before, I'll adjust my approach to better get at what's driving our disagreement. Or perhaps I won't engage at all because I know this person rubs me the wrong way -- a metric clout can't capture. In a purely anonymous setting each person is a blank slate, there is no slowly making progress together over a long period of time.

Further consider that the end goal of any ideological work must be real world action, or the ideology is pointless and weak. I meet someone IRL and they tell me they have 800 Sofiechan Clout Points -- okay, so what? I meet someone and they tell me they are xenohumanist on Sofiechan -- ah, now I actually know something about how they think! We can continue a discussion already started, bond or clash.

In my experience communities are stronger when they prefer persistent nyms. Zero knowledge schemes are useful when your opponents are state level actors. I doubt this is the case for most users!

>>5225
These are good points and a good case for why persistent identity should always be allowed. The point of anonymous “clout” is that you should be able to leverage your reputation with the platform even when you want to be anonymous, which you sometimes do even if you have a persistent nym. And when you want to be anonymous, you should be actually anonymous. It doesn't have to mean much socially, but insofar as the platform itself has some language of reward, it should be available even across disconnected nyms. The ability to have zero knowledge nyms matters to some people and i think its an important thing especially for the platform. That platform potentially *does* have near-state “adversaries”. Technical anonymity protects the platform as much as the users.

As for what you can do with clout, mostly it represents moderation authority to promote and possibly to hide posts. But maybe also other things. In some contexts we may only want to allow posters who can put up a lot of clout, or things like that. It’s also not just clout as such. I think anonymous zero knowledge credentials like “this anon works at X company” are more interesting.