Jay Graber Confuses the Gas and Brake Pedals

04 Oct, 2025

So, since my last post, there have been some developments. This post isn't really about updating readers on what's gone on – it's about analyzing what Bluesky should be doing to further their stated goals – but it's worth starting with a recap.

First, Jay doubled down and (alongside Jesse Singal) seems to have caused large numbers of users to read pictures of waffles as a transphobic dogwhistle.

Jay followed this up with a more self-serious statement...

And then, hours later, she followed this up with a bad AI-generated poster bearing the slogan "dec/acc decentralization accelerationism."

The central statement of intent in this bout of posting through it comes down to this:

We’re system architects at core. We built a decentralized network so you could run your own moderation [...]

The 'decentralization accelerationism' slogan seems to originate in internal conversations within Bluesky; a Bluesky engineer posted the same thing a full day before Graber.

The intent seems pretty clear: Bluesky (the company) wants to implode Bluesky (the service) to drive users to decentralize onto other app views and moderation services on the at protocol, thus achieving their ultimate aim of promoting the "protocol, not a platform" or whatever it is they say.

I'm going to write most of this post from a sort of maximal good faith perspective; I'm going to take them at their word that that's what they're trying to do. Unfortunately, what they're doing is counterproductive to their goals. If Graber thinks she's 'accelerating' adoption of decentralized at protocol services, she's doing the opposite – slamming the brakes.

Admitting you have a problem

So here's the thing. Why won't it work? Why is antagonizing the existing bluesky userbase harmful to bluesky inc's goals?

It feels odd to even have to answer this question, but it's pretty clear that the bluesky team doesn't view their userbase as either their customers or their product – more as a sort of incidental consequence of operating their little proof-of-concept that is Bluesky. That proof of concept is meant to eventually wind down, and the users eventually go away, at least in part, decamping for a broader decentralized 'ATmosphere', as some people call it.

So, why should they care about the users? The userbase isn't the point.

Except, of course, they are. Who is supposed to build all these third-party services? Who is supposed to populate them and give them a reason to exist? Who's supposed to financially support them? It's pretty obvious that you need users, but fundamentally you need users who are positively inclined towards Bluesky before they become positively inclined towards atproto.

You are atproto's 'brand ambassadors', to use something of a greasy phrase, and right now you're telling large numbers of people: atproto isn't for you; it's made by and for people who view you with contempt. What you're going to drive users towards is not alternate decentralized app views, relays, or labelers. You're driving them towards Threads. You're driving them to go back to the fediverse, if they can deal with that. You're driving them to quit social media entirely and retreat into more private forms of online communication.

The more you do this, the less valuable and worthwhile it is to build on the at protocol. If I were a developer on Northsky social, I'd be watching all this with dismay. When that project finally launches a widely available app view, who's going to be left on Bluesky who wants to switch over?

Are people going to trust projects like Northsky, or are they going to see the words 'sky' and 'atproto' and immediately go "nope, not interested" in the same way that a lot of people now recoil at any suggestion of joining a fediverse instance? You are destroying the brand of the very thing you're trying to build, before it even properly launches.

Because guess what, there isn't really an opportunity for people to actually move away from bluesky. None of the alternate app views currently available have their own, comprehensive and independent moderation – and moderation is by far the thing people care the most about in this whole fracas. Blacksky's app is still downstream from Bluesky moderation.

There's nothing wrong with that in regards to Blacksky, but it does mean that users that are motivated by concern and unhappiness with Bluesky's labeling are not inclined to use that one alternative – and of course, it's one alternative. For this whole 'decentralization' thing to work, there need to be several alternatives.

The supposed promise of atproto is that it's kind of a mix and match system. You decide how you want to host your data. You decide on what app view to use, which controls what you see (through default labelers); you decide what extra labeling you want on top of that. Other users also make those decisions for themselves, which means that your posts are seen or not seen based on a diffuse set of nuanced decisions by different independent actors – instead of what is currently the case, which is that your posts are 99.9% at the mercy of Bluesky.

But for that 'buffet' style of social media to actually exist, there has to be a damn buffet! Right now there's a sad salad bar with maybe two or three niche options, one large bowl of soup that 99% of diners are eating from, and one smaller bowl of much more niche soup. This is not a buffet.

Most of the alternate app views are things like Tangled; apps that use the same underlying protocol but have completely different purposes from a microblogging service. And they're mostly still in alpha. This ecosystem isn't an ecosystem, it's not even a terrarium. It's some leaves in a shoebox.

Trying to drive users to switch to other app views is incredibly premature right now. What Graber is doing by posting through it is strangling the thing she ostensibly wants to build in the crib.

But I do think there's still a path forward. There is a way you could fix it. Here's how.

Every day is shut the fuck up Friday

You need a head of comms and a head of trust and safety. Bluesky actually has a Head of Communications job open, and has had it for some time. I might just apply; I have relevant experience and surely it can't be worse that the current status quo if I was making those calls.

You desperately need someone inside the building who can take the CEO's phone away. And not just the CEO, but everyone's phone. Bluesky's entire team, which seems to contain 0 soft skills among the lot of them, has been posting through it. Bluesky's CTO Paul Frazee made a statement about Jesse Singal himself in a reply, something that the CTO has no business doing – why are you discussing individual moderation choices with an user?

Random bluesky engineers keep weighing in, sometimes trolling, sometimes boosting users who are siding with Jay in the fracas. This is a fucking madhouse. You need message discipline. You need to be professional, for god's sake. When you write posts on a social media platform that you run, those aren't posts, they're company comms. Sorry. You can't be a mod and a poster at the same time.

This is like 90% of the problem. You keep amplifying ill will. You keep getting yourself ratioed, inducing people to try to figure out why you keep getting ratioed, bringing more attention into the controversy. When Lowtax' ex wife is telling you that you remind her of someone, it's time to stop posting.

Moderation doesn't have to please everybody

But it does need to be coherent. You need an actual set of goals for your moderation standards that make sense to what you're trying to do.

You need to understand that not banning Jesse Singal, or the New York Post, is a proxy for a broader problem of inconsistency and capriciousness.

Singal has spent the last several days gloating and boundary-pushing and generally being a dick, which is his internet persona now that the mask is fully off that he's an unserious grifter. This may not be against your rules, but the problem is that you're saying "well it's not against the rules" against the backdrop of Gretchen Falker-Martin being banned for saying "lol" at Charlie Kirk's demise.

You had a whole wave of really bizarre bans mere weeks ago, and then you edited your ToS in a way that seemed defensive and meant to justify those bans. That same ToS edit also moved the ratchet towards eventually banishing NSFW artists and sex workers off the platform.

Again: you don't need to ban Jesse Singal, though that is certainly part of the easiest path to fix this.¹ But you do need to 1. apologize for being weird and inconsistent 2. build forward on some kind of clear direction.

One possible direction: the wild west

Is Bluesky supposed to be a minimal ground foundation of moderation that others build on with additional labelers? Great. You need to minimize your rules. Rules about non-con porn have to go. Rules about 'glorifying violence', whatever the fuck that is, have to go. You have to actually come down to a set of rules where you basically only ban stuff that is outright illegal or so universally reviled that anyone who doesn't want to see it banned needs to be banned.

Doing this would, in fact, turn your service into kind of the wild west. It would become more toxic. It would become worse. People would feel unsafe. But if you're really, really committed to doing it this way, this is honestly the only way to justify an infamous hatemonger like Singal. I'm not exaggerating here; people view Singal as an architect of the modern transphobic movement.

If he can have an account on your service, then you also need to accept people who post things like "I think every last israeli government official should be executed" or "every morning I wake up and open palm slam my VHS of charlie kirk turning into a human gusher".

You can either be a knitting circle or you can be 4chan. What you can't do is expect knitting circle behavior from ordinary users while clearing the runway for figures like Singal – or an example I think is even clearer, the New York Post, whose headlines in my view surely do actually violate any reasonable moderation standard against transphobic speech.

For this whole idea of "composable moderation" to work, people need to feel secure that they're not going to be severed from having their posts seen by their friends because they posted the wrong thing even though other people on the network are posting comparably spicy things. It needs to be credible that moderation is something users have choice over.

Right now, Bluesky is moderated fairly strictly in some ways, fairly loosely in other ways, and if you think it's not strict enough you can sprinkle some extra moderation on top, but you can never scrape away the excess of moderation that leads to things like the waves of post-Charlie Kirk suspensions.

Option two: fix your shit

If the idea is that Bluesky is a space with a specifically enforced tone built around "productive" conversations, you need a way better definition of what those are and how to achieve them, and you need to be consistent and strict in actually driving off people who are not contributing to productive conversations.

Like Singal, whose activity on the site for the past several days can really only be described as trolling while skirting the line of the rules. That's fine if other people get to do similar things, but not if you're mass-labeling your critics as 'rude' and banning people for saying 'lol' wrong.

This has to come with a wave of way better communication about why you're doing what you're doing, where the lines are, and how to appeal in the event of moderator sanction. You need to start aggressively banning nazis. You need to stop being scared of bluesky being seen as a hugbox for lefties and liberals, because ultimately that's what any reasonably-moderated space will look like anyway.

This approach is, I think, inimical to the technolibertarian fantasies that Jay Graber et al still cling to. It's definitely the cleanest path to restore user trust and get back to doing more work and crashing out on social media less, though.

Ultimately, people would accept a degree of overbearing moderation if and only if Bluesky felt safe and pleasant. Right now, it doesn't feel safe and pleasant; it feels like the site is run by low-key bigots who are chummy with infamous bigots. Reversing this is a prerequisite for the at protocol you claim that you want to promote to get out of the bubble of tech industry people who already think exactly like you do.

You don't have a product

Finally: You don't fucking have a product. You need to have a product. I pointed out the other day that basically everyone who's using Bluesky's code has done so by forking it, and I got a reply from one of Bluesky's engineers:

Excellent! That is what we want to happen!

The problem with this attitude is that github repos are not a product. You're not actually furthering the adoption of the protocol; you're just throwing this thing out there like a dead fish and telling third parties to draw the rest of the owl for you.

Does Bluesky even have a head of product? Is there someone whose job it is to think about users' needs? Do you essentially view yourselves as an organization of programmers who exist only to serve other programmers? Because if you do, the technology you claim to be building can only succeed in spite of what you're doing.

I understand the idea that you want people to build on Bluesky, but the first step to hacking something is just getting it to run.

For all of Mastodon's faults, a lot of people did stand up Mastodon instances, and the fediverse network did substantially decentralize. This happened because standing up a Masto instance was easy. You didn't have to understand the codebase, just configuration and installation. Documentation was not the best, but it was good enough.

Nowadays there actually are viable alternatives to Mastodon that interoperate with it – I actually use one, GoToSocial, for my own azhdarchid.online instance. I don't think this would ever have happened, though, if Mastodon's ease of installation hadn't driven people to adopt it and create a ton of instances.

More to the point, Mastodon's developers actually promoted this do-it-yourself behavior. Bluesky's developers might think they're doing this, but really they keep saying "you can just make your own! just make your own!" without grounding that in any kind of practical reality. This is both a problem of comms and a problem of product.

You need a roadmap and a PR campaign. You need to be able to say "here's what expect to ship in the future and here's what you can do with it that furthers decentralization." Instead all your posts on the social media platform that you operate are vague gestures towards decentralization. Give people a fucking recipe. Give people a Docker container.

You need some kind of product, which is to say some kind of way in which people can move towards a more decentralized network – again, without having to do your homework for you first. Launch an alternate version of the bluesky app that isn't married to your moderation service and lets users easily pick their own labeler and relay. Work on a way to get the cost of running your own relay down. Launch a PDS hosting service. Do fucking something that isn't just committing code to github and assuming it'll materialize an ecosystem on its own.

And if you do want third parties to develop on at proto, you're doing less that the bare minimum right now. Bluesky used to have a grant program for third parties developing atproto apps, which disbursed a measly ten grand then stopped giving out grants this year. If you really want third parties to build shit on your platform, you need to do two things.

You need to first, stop giving them reasons to wonder about whether 'atproto' will be a tainted term by the time they launch; again, shut the fuck up Friday.

Second, you need to support those third parties. And yes I mean materially. You need to actively promote and push people towards those alternatives, and you need to give them money, outright.

Successfully driving the adoption of a new technology, especially one that competes with an established solution (ie, monolithic corporate social media), is incredibly difficult and it's as much a social task as a technical one. You're all under the illusion that code can change the world on its own.

But real talk now

All of this said: again, this post is mostly written from the perspective of maximal good faith; of assuming Jay Graber wants what she says she wants.

I don't know that I believe it. I think that it's hard to reconcile a lot of this behavior, and the vaguely transphobic backlash that's come along with a lot of pro-Graber posting on bluesky – where people talk about trans people asking to feel safe on the platform as "weirdos" who are "killing the vibe" – with a good-faith view of what's going on.

Ultimately, I think Graber is trying to drive queer users off the platform not as part of some 6D chess plan to push decentralization, but simply out of bigoted animus. She has contempt against the kind of people who did end up using the platform, and she wants them out of her sight – she doesn't care if they use other at proto services, so long as they quit hers. If that's her goal, well, I guess she should then just keep posting.

#bluesky #social media