SlopOps /slop-ops/ (n.) — the mass production of sloppy, low-effort products extruded using agentic coding tools by people who mistake capability for competitive advantage.

It’s 2013. You — and you alone — got access to Claude Code. Imagine how much you could have shipped! Like other daydreams, it emerges reflexively and is mostly harmless. But for some people, the residue of the daydream becomes a maladaptive mental model. They make decisions as if everyone else didn’t have the same capabilities. They fail to recognize that these capabilities are now table stakes, not a competitive advantage.

The most obvious bad strategy implemented under this mistaken frame is smearing attention across 1000 projects that don’t add up to anything heading in the same direction. Brownian quantity over focused quality. Compounding towards zero. I think it’s almost an extension of 2025 Cluely “solve for distribution at all costs and don’t worry about solving for anything else” bullshit. A lot of people got caught up in that fad; the same people are getting caught up in this one. Just because you can troll the internet into looking at your startup, doesn’t mean you should. Just because you can produce 1000 semi-baked apps while telling yourself, “even if just 1 hits, I’m golden,” doesn’t mean you should.

Again, you don’t enjoy a monopoly over these new capabilities. You can jubilantly cry, “we’re so early!” on twitter all you want, but the population of people who can now produce good-enough demos has dramatically increased. It’s no longer restricted to people who know how to program. Yes, there remain dramatic differences between what a skilled engineer can do with these tools versus someone who has never shipped anything before. But when you employ the directionless 1000 projects anti-strategy, you’re choosing to enter an arena where that doesn’t matter — you’re playing the slop lottery.

To be fair, there are people who make money launching a thousand small bets. You can absolutely look at them and ask, “well, why can’t I do that, too?” There is certainly no evidence of extraordinary technical prowess. The agent is doing all the heavy lifting, and some of the most successful people playing that game admit so regularly. But that’s the wrong question to ask. A better one would be, “hey, does Pieter Levels have something I don’t have?” The answer is an obvious, “yes, distribution.” You probably don’t have that. You’re just exchanging attention and tokens for fast-fashion wares to sell on Temu.

Worse than that, the window during which the lottery ticket game can work is closing. Supply for this level of effort is going to infinity. It will probably happen before the end of 2026. If none of your scratch-offs pay out before the music stops, what then? You’ll have to retreat to trying to use these incredible tools as a force multiplier in service of a serious and worthwhile project that takes sustained effort. Something that could actually differentiate itself. Something ambitious.

Or...
...you could just do that now.

Originally published at: https://www.generativist.com/notes/2026/Feb/15/slopops

"Fight bad ideas with better ideas." The idea is appealing because it satisfies at least two virtuous constraints. The censor is denied entry. The errant belief simply gets cut out by the sharp blades of logic and reason.

• Epistemic hygiene: ✔️

• Free speech: ✔️

As far as ideals go, it's not a bad one to hold. All things being equal, the free exchange of ideas is preferable to their restriction; less wrong beliefs are better than more wrong ones. But, in some cases, the rule — fight bad ideas with better ideas — is the exact Siren's call that allows for the former's reproduction because a lot of bullshit is prionic.

You've probably at least heard mention of prions before in their particular variants (e.g. "mad cow disease," Creutzfeldt–Jakob disease, etc). But for those unfamiliar (and with the IANAB[iologist]) caveat),

Prions are misfolded proteins with the ability to transmit their misfolded shape onto normal variants of the same protein.

A virus uses cell machinery to reproduce more of itself. Prions encounter good (read: correctly functional) versions of themselves, then bend them into dysfunctional shapes.

Replication by contact.

I think a lot of ideas are like this, too, which may offend a lot of people's enlightenment sensibilities and popular delusions. The counter-argument goes something like,

If I expose myself to bad ideas, I will resist them, no harm done. Meanwhile, by ensuring my intellectual promiscuity, I'll also (re-)discover great ideas, which I'll subsequently adopt.

Rationality acts as a ratchet.

And it is! Or, at least can be. Again, this is not a bad way to look at the world — it's a good one. But, trouble generally arises when we imagine ourselves invulnerable even though we can readily observe others obviously suffering upon exposure.

And some ideas really do seem to fuck people up.

I don't mean this in the sense of "this true idea cripples something vital." It's more like: exposure to this idea reshapes how you see the problem in a mistaken way that recreates it.

Many such cases.

Setting aside Enlightenment romanticism, it's also challenging from a progressive perspective. For those not enthralled by explicit censorial passions, to do something (in the context of expressed, repugnant beliefs) still feels like it must be superior to doing nothing. I've certainly felt that way. (And acted upon the impulse.) But there is nothing which demands that must be true.

Sometimes, "sunlight is the best disinfectant." Sometimes people do foolishly choose to ignore things merely because they are unflattering. But neither case precludes the other very real possibility: sometimes the reaction itself is what provides the raw material for the context's continued propagation. Own goals are actions, too.

The analogy breaks down once you notice that the misshapen ideas animating people seem to manifest as differing forms. (Which is why this one about toxoplasmosis gondii also works very well.) Still, I like it because it highlights a case where there isn't an immunological fix. Your best and only option for prionic diseases?

Avoid exposure.

No, there isn’t a fast and frugal rule that makes it easy to delineate prionic contexts from others which are simply unpalatable or noisy or adversarial. The point is merely to admit the possibility, if only so you can ask yourself: is this one worth the exposure risk? (Think: whatever cultural war issue is currently trending on twitter.) Prionic contexts are insidious in that they seem capable of not only rendering your answer momentarily less than sober — they seem to perpetuate the intoxication in some fundamental, structural way. And since it's your perception that’s influenced, you may not notice the deleterious effects.

Related stuff

This post is just this tweet,

decadently expanded after I realized in some other tweets that I still wanted to spend some time with it. It’s also probably influenced by The Toxoplasma of Rage (which I finally read recently) and There is No Antimemetics Division (which I read recently after learning about it in this substack post, Variant Xi). Semi-related in that my brain thought of it while writing: Derek Lowe’s Things [Chemicals] I Won’t Work With series.

The most obvious consequence is that it's extremely hard for upstarts to compete against an well-established network. Having a better product isn't a sufficient condition to compel defection. "If you build it, they will [not] come." Challengers have to be very creative. They have to create something that is extremely compelling to a dense core of early adopters. And, those users have to stick around.

In the aftermath of 2021 Capitol Attack, many challengers saw an opportunity. The cost of acquiring a large cohort of users — those banned or following banned users — was almost zero. By offering them a new home, these networks could grow quickly.

The problem with this logic?

The big platforms weren't caving to political pressures, they were capitalizing on them. They weren't acting as responsible citizens, they were cheaply discharging toxic assets. That is,

1. The offending cohorts hurt their bottom line; and,
2. The cost of expulsion was uniquely low given the surrounding political climate.

Viewed this way, the true cost of acquisition for this cohort isn't near zero -- it's expensive as fuck. (At least, that's true if your goal is long-term growth and wide-adoption.) Those expelled were more likely to be toxic, in the sense that other users don't want to be around them. By greedily scooping them up, challengers concentrate a repulsive force, at the expense of future growth. Said differently, they accidentally help make the incumbent's moat deeper.

It's easy to read this sketch and think: who cares? The people "shown the door" are generally shitheads. But, I think platform-level banning is bad for social media as an ecosystem. I didn't always think this. But I've changed my mind over the last year or so. The owners of centralized networks are unlikely to be good long-term arbitrators. Want a vivid example? Look at the clumsy attempts at massaging the discourse surrounding COVID.

As to the future? Well, there is little reason to expect improvement. At scale, the learning task demands decanting absolute truth from a distorted digital projection of social reality...then, punishing the wicked deviants. Even if this were tractable, it seems dystopian. Culture is the fabric we weave together. I'm not sure abdicating agency is the way through our contemporary condition. Instead, I think we need more of it.

Ours is a brand new world of allatonceness. “Time” has ceased. “Space” has vanished. We now live in a global village…a simultaneous happening. We are back in acoustic space. We have begun again to structure the primordial feeling, the tribal emotions from which a few centuries of literacy divorced us. We have had to shift our stress of attention from action to reaction. […] Electronic circuitry profoundly involves men with one another. Information pours upon us, instantaneously and continuously. Our electrically configured world has forced us to move from the habit of data classification to the mode of pattern recognition. We can no longer build serially, block-by-block, step-by-step, because instant communication insures that all the factors of the environment and of experience coexist in the state of active interplay.

– The Medium Is The Message. McLuhan & Fiore (1967, p. 63)

Start With A Goal of Synthesis

In my last post I argued that what comes next should be an emphemeral-by-default, peer-to-peer, algorithmically-augmented social medium. A while before that, I started focusing on a different question: how can we build a medium that can synthesize twitter’s capabilities, albeit peer-to-peer. I can’t answer this question in a single post. More than that, I don’t think I can answer it alone. But, I think it’s useful to start by thinking about messages and their propagation.

Carving away extraneous detail, tweets are the atoms of the twitter universe. They bind a blob of content to an identity at a particular time. Likes and retweets are time and identity bound pointers to tweets. They socially allocate attention without content creation. Quote tweets, replies, and quoted replies are compounds. They bind a blob of content and a pointer (or two) to an identity at a particular time. Ignoring time and identity, these foundational interactions look like,

But, that is not the whole story. They joy of the medium is how users weave these interactions into a tapestry. Thus, if you want to synthesize twitter with P2P distribution, you can’t only consider atomic message exchange. You need to think about sub-graph collation and propagation, too. Measured in bytes, these may not be large objects. But, measured in milliseconds, latency can add up quickly – and latency matters a lot.

Sharing Subgraphs

Assume each foundational interaction (message) has a unique content hash. The P2P network allows you to query for hashes and the associated content. The actual content includes a reference to the users identity as well as a cryptographic signature of the message. Each node maintains a small subset of this data, conditional upon user and algorithmic requirements.

You have both the content c0 and its associated hash f(c0). The content-associated identity allows you to connect to the author or their delegated proxy. Where squares refer to the identity (identical in this case), you have the following thread,

If the protocol supports an action like retrieve_subgraph(c) and you are connected to u0, this is an O(1) operation.

Now consider the following thread,

What happens when you call retrieve_subgraph(c) for c0 on u0?

In the ideal case, u0 returns the entire thread in O(1). Because messages are signed and contain pointers the relevant parents, authenticity of the result is cryptographically assured. However, if u0 is is unaware of some inbound references or it has evicted them from its persistent stores and ephemeral buffers, the resulting view is incomplete in O(1).

This incompleteness is not a bad thing. Two cases illustrate why,

• Abuse: If c1 is abusive content replying to c0, u0 is under no obligation to propagate it. u0 created the context; they are in the best situation to moderate it.

• Indirect Criticism: If c0 is in some way bad and u1 wants to comment on it without provoking u0 or their followers, u1 is under no obligation to notify u0.

In the case of abuse, subscribers of u0 and not u1 benefit from the failure to propagate. Alternatively, subscribers of both are in the best position to evaluate and mediate (or ignore) the possible conflict. In the case of indirect criticism, the often self-perpetuating collision between identities and cliques is less likely to occur.

Incompleteness is Structure

This local incompleteness does not imply inaccessibility. For example, consider the simple case where u0 flushed its ephemeral buffers and chose not to retain the rest of the chain.

• If the requester exhaustively queried each user on the graph, the cost of achieving completion would be a burdensome O(n). Unconditional queries are decidedly not social.

• If the requestor queried users in proportion to prior interaction frequencies – and if nodes retained messages in proportion to time and interaction frequencies – the upper-bound is still n. However, the expected time until completion is much, much lower, as a property of the interaction graph. Interaction proportional queries are inherently social.

Now, assume there is a maximum cost threshold for queries. This may be explicitly programmed or implicitly generated as a function of latency, which users robustly do not tolerate. Assuming the heterogeneity of BYO algorithms and interfaces, this distribution does not lend itself to easy analysis. That is a good thing! The ragged landscape resists easy exploitation and manipulation. However, we can say something critical about the induced pressure – the probability of observing a message becomes a function of social peers willingness to propagate it.

Whether by algorithm or explicit action, this structure allows users to engage in both amplification and attenuation. These probabilistic boundaries can be rendered sharper given the addition of inhibiting signals – mutes and blocks operating at the level of either identity or content, functioning as something like antigens. The willingness to accept or ignore one of these attestations of harm is at the discretion of the user. Boundary spanners may ignore them or implement time decay, facilitating social graph healing. Those more vulnerable may enforce them rigorously, warding off harms. But, in both cases, these boundaries are subject to continuous (re-)negotiation, rather than punctuated and durable ruptures.

The result is socially-shaped content propagation. The decision not to propagate is a form of censorship. But, it exists at something like the maximum margin between freedom of speech and freedom of reach. Your time and attention are finite. The social graph structures it in useful ways. You need not abdicate your responsibilities and deliberation; but, you can collaborate for the purpose of socially-allocated attention.

Next Steps

The great power of social media is discovery. Social discovery depends on,

1. The willingness of inter-mediating paths to propagate a message;

2. The existence of ancillary curation services;

3. The ability of peers to connect to each other or by proxy quickly

I’m confident that (1) is not only viable, but the fundamental solution. I also believe (2) is a useful out-of-core service, that helps with heterogeneity and ensures the platform can evolve. The big issue is (3). Last year, I started building a platform to fill exactly this gap, but I set it aside. Now, it’s time to pick it back up. If you have expertise in building cryptographic distributed systems, I do want to hear from you. In the meantime, I’m going continue doing ground work including building an empirically-constrained simulations to guide subsequent specification.

I’ll be sharing more soon.

Originally shared on falsifiable (2021-01-15).

Motivation

Recently, I wrote a post that proposed Twitter shouldn’t ban Donald Trump because I thought the possibility of durable social fission exists. Hours later, Twitter banned Donald Trump. Soon after, a broader group of major internet companies did the same. Then, Amazon brought the Ban Hammer down on Parler, the hate reactor built by hapless fall guy John Matze and heiress Rebekah Mercer.

There are excellent justifications for why this was necessary and why this emergent coordination occurred and why was necessary. I don’t want to talk about them. What matters more for me right now? I don’t think it will ever work again. A trans-national economy of people building censorship-resistant technology seems to have a new occasion for war. The fire next time won’t be containable in similar ways.

This leads me to a simple conclusion: if you care about making social mediums work better in the future, your solutions must assume censorship resistance as a given. Otherwise, your building on an already-pruned branch.

Engineering Our Augmented Space

Centralized social media – as it exists today – is Procrustean. On Twitter, super-nodes excessively synchronize attention in mal-adaptive ways. The weight of identity-to-context skews far enough towards the former that we’re prone to chronically ignoring or failing to consider the latter. Social beefs end proliferate in wild abundance. Crowds grow, swarm, and disassociate from reality. All of this is all great for farming freely-offered attention. But, a good cybernetic ecosystem it does not make.

Given it’s defacto-albeit-not-really role as the world’s hyperspace public sphere, a great deal of (recent) attention seems to be on making a censorship-resistant Twitter. This is a mistake borne of the DWeb/Blockchain communities’ ideological defaults.

Censorship resistance may be necessary for solving some of these problems. But, it’s desperately far from sufficient. Missing is the attention dedicated to creating a social environment that fits us, not one that demands we fit it. It requires considering that – while there are a variety of technological architectures for creating a distributed and decentralized social medium – not all of them are sociological solutions.

I am interested in finding a solution at the intersection of both problems.

What follows is a small projection of my view from 10,000 feet.

1. Ephemeral By Default

Blockchains ensure immutability. Robust consensus mechanisms afford canonical state. Both can be useful. But the latter is computationally costly. More than that, there is an expense borne by the forms of expression it selects for.

In the context of ephemeral mediums, it’s excessively restrictive. Not every statement requires non-repudiation; in social contexts, very few do. If every utterance persists forever, people either self-censor or excessively fracture their identity into disposable fragments, as a defense. Both behaviors frustrate social processes – including those of consensus formation.

The need to optionally commit a statement to a ledger in contexts where non-repudiation is desirable exists, as does the need. But it is not a good default. The temporal context of most expressions is now. Carrying prior utterances – generally without context – forward, creates conditions for maladaptive hysteresis. The past binds longer than it need to, in ways that don’t reflect the present – and, in doing so, it limits the future.

2. Peer-to-peer by Necessity

On the technological side, federation offers a means of decentralization that fits in well with existing skills and practices. For example, if you can deploy a website, you can probably set up a mastodon instance. This matters for fostering adoption. Participation at an infrastructure/administration level doesn’t have a steep learning curve or cost of entry.

On the sociological side, it also seems to solve a critical problem: users have a stake in the communities they join. Or, more accurately, users are part of communities – it doesn’t make much sense to speak of communities absent individual stakes. This can help create incentives for the community to define and enforce its boundaries.

However, boundary enforcement and community cultivation seems inextricably tied up in communal ownership of identity. For example, on mastodon, your identity is subject to community/node intervention. Expulsion doesn’t render you a digital nomad; it erases the identity. For what will end up being the most passionate and proselytizing early adopters, this is too repulsive. And while you can propose networks that afford right of intact exit, these have the effect of neutering the enforcement mechanisms. What you have is less of a community and more of a federated network transport.

More to the point, I think the model reflects an implement-what-we-know mentality. Early websites were nothing more than digital brochures. Federation models try to replicate…schools? Co-op boards? Cities? City states? Nation states? They tend to assume a one-to-one/belongs-to relationship often bound by geography. But the beauty of the internet is that meat space doesn’t bind. And, the brilliance of twitter showed us that association is so free, it’s almost a liquid.

Recapitulating, Zuckerborg was almost preposterously wrong when he said,

You have one identity…The days of you having a different image for your work friends or co-workers and for the other people you know are probably coming to an end pretty quickly […] Having two identities for yourself is an example of a lack of integrity.

Identity is contextual. We all wear different masks. And, we act in ways conditional upon both our expectation, capital, and social environment. This doesn’t betray a “lack of integrity.” It is a fundamental and adaptive part of being a human being.

With this rough sketch, I reach the following conclusions:

• Individuals must own their identities. No one except the person associated with a particular identity can revoke it. The most motivated adopters won’t allow otherwise.

• The medium itself must not mediate voluntary associations. That is, for any possible relationship pair, no entity can inhibit direct communication except for either of the participants. This follows from the same adopter assumption. But, it’s also a consequence of the ethereal nature of association absent geographic confinement. While this environment has perils, it is also the source of its continuing promise.

This largely precludes federation. Peer-to-peer is the way forward.

3. Computation is the Massage

I argued that federation won’t work because it relies upon the wrong analogies. But, if any user can reach any user, the possible interaction space is quadratic. We do need something to structure something so incomprehensible.

The social graph is the scaffolding of that structure. It collapses n2 to something we can cope with. It affords repeated interactions and social recall. It defines an architecture for message passing where we act as the processors.

However, we are social creatures and social media is omnipresent. Viewing ourselves as the processors, Parkinson’s Law becomes oddly relevant: work expands so as to fill the time available for its completion. Who we follow increases until we reach the point of desired infinity-scroll saturation. At this point (or well before it), processing demands on a one-to-one basis exceed our human capabilities. We fall back upon higher-order generalizations. Fuckery ensues.

Here, social media critics generally prescribe something like “people-centric technology” as the antidote. Increasingly, I think it’s snake oil. More often then not, you can replace it with RETVRN without loss of meaning. Technology changes the way we relate to each other. With it, so do our aspirations, perceptions, mediators, and institutions. Some expand, and some contract. The way through isn’t to retreat to a now inaccessible past. The way out is to marshal our modern resources into new capabilities.

In the abstract everywhere of computer-mediated spaces, algorithms operating on messages are the massage. In particular – and assuming primarily peer-to-peer architecture – there are two opportunities for computational augmentation:

1. Propagation: On current mediums, user actions such as retweets, likes (stochastic retweets), and replies are explicit propagation instructions. They essentially say, “I would like to allocate follower attention here.” But, this is a crude mechanism. On a P2P network, algorithms can shape content percolation across the graph in heterogeneous ways. Moreover, the propagated messages need not be only excitatory. There is a stupifyingly obvious need for inhibitory ones, too.

2. Home timelines: What the user sees is a function of the messages their node receives. But, given invoked Parkinson’s Law, there will be more of them than time. The people-centric approach demands something like chronological timelines or less content. I think the algorithmic approach is better. The core problem now is that it’s not user-centric. Agency is absent. You don’t install what you want – what suits you best. The central provider does, and its objective function and yours diverge.

The combination?

Computation becomes the medium. By user input or algorithm, each node decides what to pass on to each connection. We shape the fabric of attention together. We decide what to amplify and attenuate, graph-locally. We decide what we want our experience to look like. We can create a space of varied social games to replace the blight-prone contemporary mono-crop.

I am aware that prescribing algorithms reeks of techbrosis at the current moment. But, I contend that we’ve reached the point where we are reactively imputing harms to the wrong component. Algorithms have tremendous ethical considerations. They can be harmful. They often are. But, they are also damn near magical Promethean fire. They allow us to navigate higher dimensional spaces otherwise beyond inaccessible. They can augment our capabilities. They already do. And, they can do more. The way out is through.

Next Steps

I started building what I think is the core of this last year, so I’m admittedly biased. But, it was too large a project for myself, so I moved onto a project that could help me “work my way up to it.” But recent events have changed the urgency.

Reiterating my human decency bona fides as I’m sure it will come up,

1. I believe Parler is a hate reactor;

2. I believe John Matze was a group’s useful idiot;

3. I believe the storming of the capitol was one of the most dangerous moments for Democracy I’ve lived through; and,

4. I believe the interventions taken by a variety of corporate actors do help get us through the moment, where groups trapped by conspiratorial feedback loops coordinating to do violence pose a clear and present danger.

5. I believe every one of these services was well within their legal right to act as they did.

But, I also believe that those actors were more bound by fiduciary duty than social ones – they were discharging toxic liabilities. And, the immediate needs may obscure the future harms. “Freedom of reach” is not “freedom of speech” rhetoric breaks down as more and more of our social and intellectual lives become more computer mediated. And, the further down the technological stack we go, the more entangled the two become.

For this particular episode, the benefits greatly exceeded the costs. In the long run? I don’t know. But, I do know that the event mobilized something like an army to make the same intervention impossible in the future. And, I am confident they will succeed in doing so. I want to join them, because I am a true believer in social mediums. There is still great and untapped power. But, I care about more than censorship resistance alone.

Originally published on Falsifiable.

Note: I published this on Falsifiable earlier this morning. I’m sharing on my substack now rather than in a few days because Twitter has since pulled the trigger and banned Trump. I’m just as apprehensive as before, but now I also feel like we’re spiking the ball on the one yard line.

What Would You Do?

Last November, I asked my followers a question,

Most answered in the affirmative. I would answer in the same way. No interaction with him – whether in favor or against – has ever been worth my time.

That isn’t to say I want to simply ignore all references to him, period. That would be foolish. Not only was he the President of the United States, but he was a despotic megalomaniac. America is the country I call home, so it very much matters to me. Yelling “just ignore him” isn’t even a good solution for when two siblings fight in the back seat of your wood-paneled station wagon.

Yet, I am willing to rely on trusted mediation.

And, I have!

You probably have, too.

To a large degree, that is the generalized power and promise of social media. “The real environment is altogether too big, too complex, and too fleeting for direct acquaintance.” We require indirect sources of information. Historically, such sources were bound by space and time. Technological progression slowly liberated us from both. Now, social media grants us the ability to select our intermediaries from what approaches the world’s entire population. This power is truly incredible. We exist in a space our ancestors could only imagine through the magic of fiction.

However, with great power comes great potential fuckery.

Like ‘propaganda’, ‘manipulation’ currently has negative connotation. The more comfortable and consequently common word to reach for is “influence.” It amounts to the same thing though. When we say someone “influences” us, we mean we’ve escalated their permissions over our beliefs in some critical way. And as we extend trust, we constrict incredulity.

Ideally, this relationship isn’t paired with passivity. When someone’s expressions fail to match either our direct experience or our careful interrogation (or both), our trust in them should degrade. And, with degraded trust, our skepticism of what they have to say grows accordingly.

This is an adaptive process. Human sociality affords distributed search and specialization. The titanic polymath who could deeply understand the entirety of human knowledge is a romanticized relic of a long departed past. Having the world’s information at your fingertips means that you can at enter a mind-boggling large amount of the terrain at any given time. But, you can’t explore it all. Time still binds us individually. We need we.

This process becomes mal-adaptive in proportion to the degree that trust becomes a function of something other than our accumulated evaluations of what someone says. As it does, the harms associated with the negative connotations of manipulation and propaganda manifest. We find both our emotions and beliefs manipulated in ways at odds with what we would discover independently, given sufficient time and information.

On social media, this happens with fluid ease.

Trump leverages this form of attack relentlessly. Enjoying (pathological) agenda setting power in both traditional media and as social media supernode, his ability and inclination to induce mal-adaptive contagion is unrivaled in recent times.

For these reasons, I think Facebook and Twitter banning Trump is, at face-value, reasonable. I also thought it was the correct thing to do in 2016. It’s an emergency scram button to shut down our the hypercritical reactor. Banning Trump broadly facilitates the discovery and expression of beliefs for all participants – online and elsewhere – save Trump. However, I don’t think it’s reasonable to pretend doing so represents a good solution. Instead, it’s merely the best one constrained to those that are easy and without admitting the possibility of better architectures.

A Decent Proposal

Trump’s grip on culture and social media strangles both. But only the latter affords straight-forward technological interventions. Individual action has proven insufficient. There are a myriad of incentives (many of which are also pathological) that prevent people from using mutes and blocks. It is a collective action problem for which the collective has no binding means of consensus formation. The question posed by my tweet allows me to propose a minimally viable one: ask the users.

To be more specific, consider a candidate specification:

On promptDate, all users will be presented with a modal dialog asking them to make a YES/NO decision. If they answer YES, neither @realdonaldtrump nor any interaction with him gets placed in their feed. That is, twitter filters all of his tweets and replies from both your timeline feed and all user feeds (from the perspective of your account). If you answer NO, your subjective view does not change. The aggregate vote tallies are visible on each day up until bindingDate. You may change your vote any time until then. Afterwards, votes and resulting infrastructure changes are commited and irrevocable for each particular account.

I think this solution would facilitate healthy conversation with effects that reverberated offline. Moreover, it does so in a way that expands the agency of twitter users, rather than unilateral restricts it. However, neither hypothetical benefit motivates my reasoning.

Platform Balkanization

I’ve demanded the digital pound of metaphorical flesh many times. In the case of Donald Trump, I’d revel in it. That is to say, I understand viscerally the impulse towards participant expulsion, generally and in this particular. But Banning Trump is unique because it is possible that his move to, say Parler, could drive a mass exodus.

The first order effects of such a potential outcome are delightful for many parties, including me. Twitter (and facebook), gets to discharge an asset-that-has-predictably-curdled-into-a-liability. There is a reason neither has acted until recently. And, Gab and Parler aren’t even competitors – they’re White knights. More importantly, those subject to durable social media abuses get genuine harm reduction. Dismissing that is just a smug variant on “fuck your feelings.”

But, reduction associated with displacing the shittiest participants of social media onto a new medium could be an ephemeral mirage.Out of sight, out of mind, and into commercialized hate breeder reactors may not be a good strategy. Thus, the risk I am more interested in at least injecting into the current frenetic conversation is platform balkanization.

Broadly, there are two hypotheses about what happens when these users get isolated on a new network partition (here, in the form of a different platform),

1. Smothered fire: The echo-chamber becomes a circle jerk; participants get bored for lack of targets; raids on larger networks will occur sporadically but they can’t trigger the same dynamics; the medium fizzles out.

2. Coal-seam fire: The echo-chamber splits consensus reality in an enduring way; people with similar corrosive beliefs form an even more cohesive orthodoxy; the distance between them and people outside the partition grows; in-group trust pins on maximal, granting uncritical reception of escalatingly radical beliefs.

The smothered fire scenario is seductive. And, it may happen! But, increasingly, the coal seam fire seems…if not more likely, than at least the one that deserves way more focused consideration. 2016 and everything after wasn’t a wildly unlikely bad sample path. It was a continuation. We hid some things from ourselves before then; we want to do so yet again. While criticisms of social media are extremely valid – there is an impressive inventory of problems – blame assignment then and now has some convenience borne of desperation. But, there is at least one critical difference: Public passions and fiduciary responsibilities are currently aligned.

Freedom of Speech vs Reach

Censorship is easy to defend against at low temperatures. It’s our cultural’s default stance. But at high social temperatures where norms diffusely lose stability, so does the strength of our convictions. It becomes easier to withdraw principled support when principles no longer broadly bind.

Things are very hot now.

I have neither the desire nor inclination to defend Donald Trump on free speech grounds. I think people who do so are fools – those who do so in deference to a principle and with precisely zero context, doubly so.

I also think people who pretend it isn’t censorship – that it is a question of reach, rather than speech – miss the point in similar ways. Yes, Twitter is a private company. It has every legal right to ban anyone, and especially someone as malignant as Trump. But, I treat it as something more important than that – I treat it as a public commons in hyper-space. Celebrating the corporate entities ability to unilateral tear up a large portion of the social graph makes the contrast obvious. In it, I find myself in a curious position: imploring Lord Jack to consider better alternatives rather than taking the path I previously condemned him for not taking!

The solution I’ve offered isn’t meant as a navigated compromise between the smothered fire or coal seam scenarios. I’m not even sure it’s a reasonable proposal – I haven’t done the work necessary to draw any conclusions. I don’t even have the data or financial resources to do so.  Instead, it’s just my way of getting people to think more about solutions that preserve the possibility of a social medium that continues to transcend space and time. One in which the space of potential human social interactions is n² instead of m²+k² ≪ n², winnowed only by boundaries enforced and enforceable by the participants and the participants alone.

To those who reject what I’m saying out out hand – I get it. You’re tired and this whole experiment in kakistocracy coupled with a Clockwork Orange-style compelled viewing has been a tragic catastrophe. I just want to make sure we’re not continuing the tragedy. The obvious compensation for partitioning the social graph may not even be paid out. And, for the sake of correcting our own social delusions, we do need we.

What’s the difference between me and you?

As a convenient fiction, assume it is possible to map "I believe" to a point on a finite, one dimensional space. "Taking a position" means placing yourself at the appropriate point.

For any particular belief, you're not the only one doing so. At any given time, many people are expressing their beliefs. If we happen to observe someone doing so, our brains quickly establish our relative position to theirs, estimating the difference between opinions.

These differences have direction in addition to magnitude. With routine evaluations, we are adept at orienting ourselves to this space. We quickly assess how far we would have to move relative to our expression to comply with someone else's.

When people say someone, "defends their position," they're often giving a name to something like inertia. There are a variety of costs — social, psychological, and even institutional — associated with moving from one point to another. It's generally easiest to stay where you are, at least perceptually

But — ignoring that individual opinions are noisy samples from complex, latent, and generally conditioned distributions rather than an exact projection of some underlying, isolated belief — people do "change their minds" and their probable, subsequently realized positions.

In doing so, the picture becomes dynamic. The moving actor may be aware that they have moved, as well as the direction of their motion. If observed, their counterpart may have a rough idea of the change as well. The "relative to"s multiply.

Alice (blue-er) may be aware of where she is now and the direction and cost of traveling back to where she was.

Bob (red-er) may also be aware of this aspect of Alices change, although he may instead perceive it as a change away from him, rather than in terms of going back.

Extrapolating, Bob may furthermore assume Alice's probable continuation.

Alice, too, may perceive a bias to her future changes.

The question now becomes: how should this change affect Bob's underlying belief and, by implication, future expressions? If he trusts her, he may assume that the change to her position reflects evidence that she had access to and he did not. Therefore, his change should match hers, at least in direction.

But, Bob also knows he and Alice started in disagreement. That is, their previous experiences — the ones integrated in their beliefs and observed through expressed opinions — were different, assuming sincerity in expression. Thus, Bob may choose to stay put, discarding Alice's opinion as irrelevant.

Alternatively, Bob may interpret Alice's movement as evidence of an environmental change that he has yet to observe or directly interrogate. And, given that they initially disagreed, Bob may infer that this evidence calls for reinforced disagreement — movement further away from Alice.

Experiential Trust

The previous illustration also applies when Bob generally distrusts Alice. To illustrate, assume Bob and Alice start out in the same position.

Again, Alice moves.

Again, Bob must interpret and weight the evidence inherent to his observation of her expression. However, this time, the initial context of agreement is less likely to compel Bob to move away from Alice. Instead, given their initial agreement — and assuming Bob trusts Alice and believes her expression was sincere — his best action is to update his beliefs to match Alice's expression.

But, what if Bob doesn't trust Alice? What if his experiences with her in other contexts makes him believe she's an unreliable or even deceptive source of information? In this case, the update rule previously explored seems plausible — rather than ignoring her observable change, he moves to oppose it. Again, distance grows.

The Problem With Strangers

Moving away from contexts with familiar counterparts, what happens when Bob encounters someone he does not know, Carol, expressing a particular belief? How does he integrate this information?

He may simply ignore the information offered by her expression. After all, if he has no experience with Carol, his assessment of her reliability is bound to be error-prone.

Of course, her expression does communicate some information. If Bob generally trusts strangers — assuming both reliability and sincerity — he may move towards Carol, even absent direct experience with her.

Alternatively, if Bob generally distrusts strangers — assumes that they are either unreliable or deceptive or both — he may instead move away from the stranger, Carol.

Yet, as the careful reader may have noticed, Bob is a circle and Carol is a square. While Bob may not personally know Carol, he may have accumulated a good deal of experience with squares that he can use as a template for evaluation and judgment. That is, he may use an accessible stereotype (here, shape) instead of an unconditional level of trust.

Consider the addition of another stranger, Ted. Like Carol, Ted is a square. Last week, Bob observed Ted taking a position.

Joining the two observations — Ted then, Carol now — by accessible stereotype alone, Bob may perceive a leftward motion in squares.

Reversing the temporal ordering of his observations, he'd perceive a rightward motion instead.

But, in both cases, the perception of motion is noisy at best. Carol may stand in for Tim and vise-versa only if the homogeneity of beliefs induced by "being a square" is absolute, an improbable condition. In all other cases, the motion may be mere illusion — resampled, Tim and Carol may take the same positions previously taken. Thus, if Bob already updated his beliefs in response to his perception of motion and trust/distrust in squares, he risks integration of noise as signal. In turn, this affects how he'll pay attention to, interact with, and learn from squares in the future.

Media Connection

Of course, we're more than just circles or squares. The very best of moral philosophies — or, at least the ones I prefer — prescribe treating and judging individuals as individuals, rather than on the basis of their traits. Yet, to deny that group identity isn't a vital and necessary aspect of how we interpret and relate to the world — an ever-present and accessible prior — is to engage in willful delusion. Moreover — and contrary to some popular beliefs — stereotyping is not inherently maladaptive. Consider the Platinum Rule: "Treat others the way they want to be treated." This rule is hard to implement for strangers absent some coarse model of their experiences and expectations.

However, stereotypes also represent an exploitable vector for manipulative attack. The challenge is to recognize when stereotypes misguide you and, more critically, when they're used to manipulate you. Consider that human beings don't have just one identity — we have and cultivate many of them. And identities themselves aren't fixed concepts. But, when presented with information for evaluation and judgment, some identities are more accessible and salient than others.

Why?

Partially, it's a function of biography — the chance (albeit predictable) bad encounter with someone "not like you" that dominates your expectations in a path-dependent way. Something as simple as a competitive context — an omnipresent reality — may compel adoption of a negative stereotype. And reproduction along big structural lines is startlingly easy. There is no easy fix here.

However, we do see position taking on mass and social media presented as exemplars of some underlying social reality. And, to some degree, it is. But, the incentives for both platforms and content producers select for ever-salient discord. And discord is an easy story to tell.

As a hypothetical, imagine that you're a producer at a news network. There is a strong incentive to inflate both the apparent distance to and relative motion of people with identities underrepresented in your audience. You can do that by carefully (but with incredible ease) selecting which out-group voices to present as adversaries, in addition to an implicit order.

But, that's only one side of the manipulative template. Thus far, Bob has been an individual with a circle identity, important only in that it was not a square. But, by presenting circle-aligned commentary, the manipulative media producer can remove Bob's particular position from the equation. They can replace his subjective experiences with their own position, so long as distrust in the out-group exceeds the resulting incredulity for an in-group position too far removed from Bob's. "You are here" gets lost.

Meanwhile, none of this has accounted for any direct evidence about a hugely complex and dynamic world that may and almost certainly has undergone changes. The message conveyed concerns identity. Particular issues are merely the substrate. The misdirection induces pathology — one that is surprisingly difficult to defend against, given our cognitive hardware.

At the level of an individual it may be useful to ask this outline of questions for evaluation,

- What are you reacting too?
- A static position?
- An apparent change?

- Who expresses the opinion you observe?
- Is it representative of anyone else?
- Does it represent a genuine change?
- If so, is it robust or ephemeral?

- Who relays the expression to you?
- Do they do so to present changes or positions?
- May the have an ulterior reason in either case?

But, really, for all the exhortation about thinking carefully and exercising media literacy, this isn't a problem subject to piecemeal patches at the individual level. The information environment is fucked because the incentives are wrong. It's all entertainment, with unpredictable major news on world events as tonight's regularly scheduled programming. It's hard to see reality when everything around you is busy telling you dark twisted fantasies.

After finishing a Ph.D. investigating belief system expression and discovery through simulation, I started working on a social immune system for twitter leveraging their API. Eventually, I realized that trying to build an elaborate sandcastle on someone else’s private beach isn’t the smartest of plays. Now, I’m building something else. If you are in this space — and especially if you foster romantic notions about the power and promise of decentralizing computer-mediated communication — please reach out. I want to hear from you.

Ours is a brand new world of allatonceness. “Time” has ceased. “Space” has vanished. We now live in a global village…a simultaneous happening. We are back in acoustic space. We have begun again to structure the primordial feelings, the tribal emotions from which a few centuries of literacy divorced us.

Marshall McLuhan / Quentin Fiore. The Medium is the Massage (1967, p. 63)

The Pre-Global Village

Dunbar's number refers to the claim that people can only manage relationships with around n people. (Most commonly, n is a very hand-wavy 150.) It's one of those ideas that made the leap from academia to popular science and then to cocktail parties. Consequently, I cite it, quote it, and talk about it, as well. But, I'm not a big fan of Dunbar's number for a few reasons.

First, either I continue to misunderstand it or most other people do, and I'm pretty sure it's the latter. Managing relationships with n people does not entail "knowing" 150 people in the dyadic sense of having had sufficient experience with them to recall who they are. That's not how human sociality works. Managing relationships with n people means something closer to: you both have some experience with them and also understand how they socially relate to everyone else. That is, you know 150 people when you can competently predict how any two of them would interact in paired-off social circumstances without you. In the narrow context of anthropology skewed towards the lower end of socio-complexity, the difference is hard to see. For bands or tribes bound by geography, resource scarcity, and compressive kinship grammars, knowing 150 people and understanding how they relate to one another comes close to meaning the same thing. But as we move towards increased socio-complexity, the distinction becomes critical: Dunbar’s number is a statement about inter-relationship comprehension with upper-bounds quadratic in n, not linear.

This restatement allows me to introduce the second, much stronger reason I don't like Dunbar’s number, at least in common retellings. Extrapolating with data from other primates, Dunbar himself theorized that the binding constraint was neurological.[1] That — impressive as our brains may be — we can't seem to juggle social relationships with more than 150 people. Expanding that to the upper-bound of 22,500 possible inter-relationships, (150²), it sure does seem challenging! But, as the band/tribe example demonstrated — and as Herbert Simon emphasized endlessly — the environment itself shapes problem-solving considerably, and we are remarkably proficient at shaping ours. Thus, I don’t think cognition is what binds; I think it’s mostly environmentally-shaped social time.

At the level of a band or tribe — where geography, kinship, and resource scarcity compel durable proximity — it seems reasonable to believe you could accumulate direct experiences with 150 people that would allow you to induce at least some mental model for them all. The quality of each model varies with experiential time, which is some function of social proximity. You'll have an excellent model for your parents or siblings or peer-group; you’ll have a worse model of your mother's father's aunt.

With the computer scientist's cap on, the question becomes a familiar one: if it is time that binds, are there more efficient (sampling) algorithms? Nature and human experience answer in the affirmative: we gossip.[2] Talk is cheap in the sense that it is not expensive, not in the sense that it is valueless. It takes less time to gossip than to observe or participate in all the covered interactions. At the level of band or tribe (proximity-constrained and with kinship grammars to compress the inter-relationship space), we would reasonably expect to have mental models for both the gossiper and subject of gossip. Consequently, we can evaluate the gossipers claim; judge its veracity; then, update our expectations of both accordingly. Given the same time constraints, an allocation to gossip can boost model performance.

Further boosting comes from the particular social topologies we readily forge — in particular cliques. Each member of a clique has semi-regular direct interactions with every other member, allowing them all to induce good models of mutual behavior. They also gossip about each other frequently. In this particular configuration, there is the happy coincidence of accessible good models for all the implicated parties for the purpose of evaluation and judgement. This makes deception both more difficult to achieve and more costly — good models detect the attempted exploit and that detection may quickly propagate within the group of people the deceiver interacts with most. Thus, cliques are remarkably efficient in social and information-theoretic terms.[3] For the same amount of effort, you can produce a greater number of high quality mental models.

This trusted dense core — along with progressively less dense ones that Dunbar did talk about — allows us to deliberate over more far-flung spaces in our social neighborhood. The cost? Errors become correlated. Social expectations become the product of ensembles. At low-levels of socio-complexity, this isn't pathological. The errors are already implicitly conditioned on interaction frequencies and there is still considerable shared reality so, in some sense, the errors are structurally minimized. But, at higher-levels of socio-complexity, things fragment. Ensemble expectations become JIT category factories, and the consequent space of inter-relationship schemes explodes. Sometimes, calamitously.

The Twitter-Scale Global Village

Armed with this conceptual framework, we can now shift our attention to social media. The environment is dramatically different. Time still binds there…but "there" is a different kind of there. Geographical coincidence no longer limits the set of potential relationships. People can transcend many of their traditional and local circumstances. This is, of course, the most celebrated aspects of social media — one that I continue to delight in. But, it also moves us closer to the asymptotic, quadratic upper-bound of complexity for inter-relationships in ways we never have before.

To get a sense of our new scale, the following graph shows the empirical cumulative distribution functions for twitter friend and mutual relationships over a sample of 33,292 users.[4] Here, ‘friend’ refers to someone the ego follows and ‘mutual’ refers the the case where the person followed also follows the ego. (The curious notch in the Friends CDF is a social artifact of twitter’s subscription rules that allows a maximum of 5,000 friends under promiscuous-growth-limiting conditions. The shaded portion portrays 95% of the mass. The vertical line is the median.)

The median number of mutuals is 244 which is well within the plausible range for Dunbar’s number. The median number of friends — those we preferentially observe— is 798. In both cases, the distributions are far from Gaussian. To a large degree, the beefy right tails are an environmental effect. Friends and follows don't decay from disuse on twitter as they would in a “real” (non-digitized representation) of a relationship. This inflates friend and mutual statistics in misleading ways, especially if algorithmic mediation winnows presentation, as it almost certainly does.[5] But, the inescapable conclusion is that, without geographical confinement — and, with a communication channel that approaches the lower-limit of legible expression — we're no longer in the anthropologically tame world of bands and tribes.

Of course, there are familiar social regularities. Cliques continue to form as they always do. You don’t need to fire up Jupyter notebook to see that. Even limited to avatars and ambient intimacy, we're remarkably good at detecting cliques. It's a familiar twitter experience: this person belongs with that one and some others — not in a stereotypical way, but in a who-sits-together-in-the-cafeteria one. We seek the strong force of social density.[6] But, the rest of the interaction space has unfamiliar and sometimes uncanny sparsity. In my conception of Dunbar’s number, I used 150² = 22,500 as the upper-bound alternative for people who exist in an environment where fully-connected graphs were plausible. (That is, everyone does indeed "know" everyone, in some not-entirely indirect way.) Social media is not that environment. As the following graph of friends-of-friends and mutuals-of-mutuals shows, it’s much wilder.

The median friends-of-friends and mutuals-of-mutuals are 209,043 and 66,787, respectively. These are big numbers (and they are both gross UNDERESTIMATES given API limitations).[7] Clearly, the frontier set of who you or someone you know might interact with in an observable way is huge. Of course, you could say the same thing about any “IRL” city. Just because they are accessible does not mean they are accessed. Except, they are — and with regularity. The following graph shows the proportion of interactions by relationships and interaction types. With the sample I have, it’s not possible (yet) to properly partition friends-of-friends (mutuals-of-mutuals) from outside of friends-of-friends (mutuals-of-mutuals). Consequently, “outside” here merely means not in the friend or mutual set.

Unsurprisingly, most interactions take place between friends (i.e. people the ego follows) and around a third of interactions take place between mutuals. But, an impressive amount takes place outside these relationships. Your friends and mutuals do not form anything like a blanket for interactions you only occasionally escape. And, rightfully so — ambient discovery couldn't work so wonderfully well if they did! However, this also presents a critical problem. There is no way to accrete sufficient interactions and observations with something on the order of 60,000 people (in the dramatic underestimate). Even if you could, cognition really does bind here.

But, it's worse than that because the latent space we're intrinsically trying to navigate is the quadratic inter-relational one. In search of lacking and needed structure, our remarkable pattern-matching machinery gets to work manufacturing cues from scant material. To some degree, that’s what culture affords. On twitter, it means we create meaning from things like "pronouns in bio" or American flags in a display name or the hashtags in someone’s tweets or even the vocabulary someone uses. These cues provide a means to aggregate experiences, transcending experiential sparsity in a way that affords stable expectations. Thus, when we're confronted with contexts that have limited, ambiguous, noisy, and error-prone available information — something that happens with high frequency on twitter — cues offer a strong and readily available signal for integration. Absent the necessarily experience, the resulting mixture allow us to construct “good” models in a better-than-chance predictive sense.

Unfortunately, the social cues end up dominating the accessible information. You find yourself updating expectations in response to synthesized experience over what is often not much more than noise. (You notice this most on the knife’s edge — "I think I like this post but let me first check the profile description and a few other tweets to make sure.") The trouble is that the resulting judgment back-propagates over both our social and asocial beliefs (i.e. knowledge or something like it). Concurrently, by something like Hebbian logic, these ensemble-manufactured social cues end up wiring your interaction networks — providing the missing relational structure through something like a self-fulfilling social prophecy. Thus, the chronically-correlated errors have much more pathological effects. Social beliefs leak far too much information into our (presumptively) asocial ones.

Tomorrowland

Recapitulating, what binds us now isn't that we can suddenly connect to n people, where n continues to approach the full population of the world. What binds is our ability to navigate the quadratic inter-relational space. We may aspire to treat other people as individuals but, in practice, what we get is a proliferation of categories with attached expectations that condition perception and behavior. This begs the big question: do we have sufficient means of grounding to resist the decoherence of beliefs and the reification of cues induced by these entropic contexts? Or, do they occur with such pinned frequency that we can't recover — such that the errors can't correct, and instead they invade and degrade surrounding knowledge?

I don't know.

It's an unanswered question. More than that, I think it's an analytically intractable one. Not only is social media a remarkably complex system in its own right, but it's not an isolated one and the world is fantastically large and messy. Personally, I’m irrevocably optimistic. History repeatedly documents the death knells that never came. Writing destroys the mind! Books destroy the mind! Radio destroys the mind! TV destroys the mind! The world wide web destroys the mind! Social media destroys the mind! Except they didn't and probably don't. What new mediums do destroy is the primacy of older ones.

Along the same lines, there is nothing magically natural about the social configurations of bands and tribes. Mostly, they ofter us hints as to evolutionary pressures inherent in our species’ once-upon-a-time cradle. The same applies to the world as we imagine it was before social media — it isn’t privileged. Homo sapiens sapiens is nothing if not adaptive. We shape our environment; our environment shapes us, over and over again, recursively and with no end. This is more of that. What we're building / growing / evolving now will change the ways in which we relate to one another. They already have; they still have to. The architecture is fundamentally different: what we individually and collectively can perceive and achieve will be, too. It need not be harmful and it is not unnatural—it's just next.

None of this is to say we're there yet…wherever there is. Criticism remains, as always, important — it’s one of the basic corrective mechanisms we have to flop about the space of shared possibilities somewhat reasonably. It isn't merely valid — it's utterly necessary. There were, are, and will continue to be problems, many of them serious. We don’t yet know which structures are reliably bad and generate robust harm. However, we can recognize that current designs still fail to implement the sensing tools we need to perceive and the actuators we need to act in the new socio-cultural space. They mix dyadic relationships and experiences with social and cultural ones in ways that durably frustrate our cognition. More than that, limited APIs aside, we're mostly subjects of centrally-managed architectures that frustrate design innovation and evolution. We're governed by entities trying govern an intractable space at speeds that can't even keep up with last year's 0-days. The opportunities for and inherent to a new ecosystem are profound.

But, I think everyone (myself included) needs to spend sufficient time zoomed out far enough to see the full field. Already, we spend time online to entertain ourselves, to seek validation, to mitigate social isolation, to learn, to discover opportunities, to transact, and to deliberate. There isn't one mode. It's not one thing. It's not an isolated or meaningless playground, as some of the most confused critics claim. It's closer to a new kind of metropolis or nation.

Or what eventually competes with them…

…past human performances are no guarantee of future realizations.

After finishing a Ph.D. investigating belief system expression and discovery through simulation, I started working on a social immune system for twitter leveraging their API. Eventually, I realized that trying to build an elaborate sandcastle on someone else’s private beach isn’t the smartest of plays. Now, I’m building something else. If you are in this space — and especially if you foster romantic notions about the power and promise of decentralizing computer-mediated communication — please reach out. I want to hear from you.

Thanks to my wife and Danny Horowitz for reviewing drafts.

Notes:

1. I’ve taken considerable liberties in relaying the anthropological story and literature because this isn’t an academic post and I’m not writing a lit review. I’m also deliberately using terminology that frustrates ease-of-concept recognition for both academics and people familiar with other pop-sci terminology (e.g. ‘intersubjective’), hoping to constrain attention to a more narrow band of discussion.

2. Here is an example of where I’m taking some poetic license. Dunbar talks about the relationship between grooming and trust formation quite explicitly. He even theorized how that could evolve into language (and was largely criticized for the claim). But the academic chain-of-evidence here obscures the point I’m trying to make. (Plus I haven’t read him in a decade.)

3. If you haven’t read Robert Axelrod’s, The Evolution of Cooperation, I strongly encourage you to do so. I’m admittedly biased (my dissertation cites his as foundational work), but it’s such an elegant and beautiful demonstration of things that are otherwise hard to reason about.

4. This is a weird and biased sample. I drew it using snowball sampling with both Paul Graham and myself as the seeds. Except, frontier expansion followed all friend relationships in breadth-first order. That is, it sampled all the people Paul Graham followed; then the people they followed. (And the same starting for me). At the time of writing, n=33,292.
   
   Admittedly, snowball sampling has some serious problems in most cases. In this one, it over-samples from the top-end of active users. If you took a much larger sample of friends and followers via the tweet sampling endpoints (which wouldn’t get you mutuals), you’ll see lower medians because it mixes casual and active users (and, without correction, retains those with no friends and followers). Since I believe the information environment for everyone will increasingly resemble the world the latter now lives in, I don't foresee any qualitative deviation from what I'm trying to broadly show here associated with my sampling method.
   
   I have to look through twitter’s new restrictions on sharing for academic purposes. This is graph stuff (their moat), which they care about a bit more than just tweet IDs. But, eventually, I hope this analysis, at least, will find a home on falsifiable.com…once I launch that.

5. I have no good means of testing it, but I’m pretty confident that people with high friend and mutual counts tend to rely on algorithmic curation, careful list hygiene, or alternative interfaces like tweetdeck. You can test for the last case with the source field of a tweet entities, but I haven’t done so.

6. There’s a connection to the efficient coding hypothesis here that I keep trying to make unsuccessfully. Maybe you can find it though.

7. Snowball sampling along the frontier of friends in full breadth-first order has exponential growth. From each seed, you follow ⟨𝑘⟩ paths; then, for each sampled node, you follow ⟨𝑘⟩ more; then, for each of those, you follow ⟨𝑘⟩ more. Given the twitter's API limits, this makes friend-of-friend work...difficult. (Although, again, you could get just the friend distribution by collecting user entities.) At the time of writing, approximately 92% of relationships are missing from this analysis. Hence, I lump friends-of-friends (and mutuals-of-mutuals) together as "outside" when they would be much more interesting once partitioned.
   
   The problem is fun to think about because this "missingness" statistic is related to the structure of the graph. For example, if you have a fully-connected component, then you may have 9 friends and they all have 9 friends but the entire population is 10 people, making sampling linear not quadratic at depth one. Twitter is not like that though — it’s much more sprawling. Until some unreachable (given API limits) point, the expanding frontier will always produce an outcome so that most of the friend-of-friend/mutual-of-mutual set is missing. There are better sampling methods and means of correction, but this sample wasn't even meant for this particular analysis, and the corrections aren't trivial.

When I was a child, my optimism was boundless. Infinite wonder surrounded me, and I saw it as exactly that. Life was for exploration! And I was lucky. I grew up as the world built and developed greatest territory ever imagined: the internet.

Then I grew older.

I never lost the sense of wonder. I count myself lucky for that. But all the complicated entanglements became obvious to the point of near-preeminence. "Here be dragons" was no longer a beacon to run to. It was a serious warning.

I put away childish thoughts and reasons.

At least, that's what I deluded myself into believing. The truth is something far worse happened: the maps got filled in and I mistook them for the complete territory. They are not. They never were. They mustn’t ever be. Thinking otherwise is a dangerous intellectual trap that often accompanies familiarity and expertise. The true Siren's call that smashes your ship to bits on unforgiving rocks. History may rhyme, but it need not repeat. Not only is the map not the territory, but there is no fixed territory. That's what it means to be human: we have the near-magical ability to imagine and create new ones.

That's also what the internet was to me as a child. It wasn't a place — it was a substrate. It was an æther for human creativity. Promethean Fire channeled by "the electron and the switch." An everywhere and No Where I entered in order to explore and discover ideas, people, and possibilities — and to realize them!

The internet is still that place for me. But the future I imagined would have arrived by now never did. Don't get me wrong, people built some truly incredible things. My workstation really is a bicycle for my mind; my iPhone truly does make my life better; and, whenever Twitter goes down — when we suddenly find ourselves unable to think the same thought at each other across the world all at once over a damn near philotic web — I get reminded that the social world is unfathomably more accessible than it ever was before. And yet, I think it is all a stunted thing in comparison to what I thought it would be.

This has been a lot of been abstraction and aphorism, so let me try to be more concrete and specific: I think the great force of future's diversion was cloud computing.

Cloud computing is awesome, in both senses of the word. By invoking the correct spell, I can marshal and command computational resources well beyond those I imagined even a decade ago — and I can do so for surprisingly little expense. Jeff Bezo's was right: the internet is like electricity, and infrastructure as a service means you can do more with it.

But there is an expense paid that's often obscured because it's not currently measured in dollars. Cloud computing — especially when coupled with ad-supported business models — selects for particular data flow patterns; and, in turn, those data flow patterns select for particular computations. So developers don’t spend their time building smart and powerful agents to assist, extend, and act on behalf of the people they serve; developers build extraction engines -- mechanized means for filtering signal sampled from a tiny portion of the spectrum of our possibilities, narrowed by historically-imposed constraints. It's not a malevolent process. It's the evolutionary outcome born to a series of realized happenings.

But it need not be that way.

Let me give you an example to entertain. I am what you would call…an active twitter user. I use it like many other extremely online twitter users. But I also use it differently than them. I have the technical skill to create different views than the one twitter restricts me to — the one their architecture and business model demands.

• I use a graph database to see how people relate to each other. It adds another dimension to the experience, often an illuminating one. When you can condition expressions graphically, you can see the contrasts between groups of people. You can see with a little more clarity what people believe because they believe it as an identifiable people versus what they believe because it's directly-coupled to their experience and knowledge.

• I use a full-text search engine over previous tweets to look at the progression of someone's expressions. It adds another dimension to the experience, often an illuminating one. In isolation, a particular tweet may resonate, but it doesn't help you judge who they are or estimate what they believe. N of one is the lowest of bounds. And social media means — algorithmically curated or not — your sample is intrinsically biased. The things your most likely to observe? Those you're most likely to agree with strongly or reject vehemently. If you want perspective, you need the ability to zoom out, an affordance not provided by the offered client.

I spent the better part of a year trying to figure out how to offer these services broadly to people. My conclusion was that I can't. It's a fool’s errand. Twitter actively doesn't want you writing clients and I don't want to build a sandcastle on someone's private beach. Best case scenario? They don't knock it down for a while. But even more critically, I don't think anyone could make a twitter that works the way I and others may want using an ad-supported business model or an efficient data flow architecture, let alone both. That's the evolutionary dead-end we find ourselves in through the lens of one small but (hopefully) legible example.

So yes, it is time to build. It's always time to build! And, I know what I'm building now.[2] The internet is a decentralized and distributed network of people and resources. I think it's the greatest thing we've ever built. But, ironically — and, tragically, too—we failed to maintain peer-to-peer as the primary conduit. This matters because how we interact selects for what we perceive. More than that, computation itself is a medium. Hell, it’s THE medium. As more people learn to code and as no-code becomes more powerful, how we interact also selects for what we can achieve. As that tide continues to rise, I want to make sure it lifts us all. I want to help push us out of this local minima.

The future never arrived because we failed to distribute it correctly.

1. Photo by James Copeland / Alamy Stock Photo. It’s actually pretty shitty that a platform for creative people still has no clean means of attribution for use of a different type of creative person’s work.

2. I don’t mean this metaphorically. This whole essay is me waving a flag. I applied to YCombinator hoping to build a business motivated by these exact ideas. If I get in, I’m going to build it; if I don’t, I’m going to build it. In both cases, I need help. Give me infinite time, I could do it myself. But I don’t have infinite time, and I don’t think we do, either. If you’re a distributed systems engineer, cryptographer, network engineer, virtualization engineer, or product developer, please reach out to me on twitter or via email at jbn@abreka.com. (Note: I’m unlikely to respond until next week because now I have to go back to playing an important albeit much less interesting game.)

Ambient intimacy is seeing twitter's "What's happening?” prompt and answering, "How was your day?" instead. Given sufficiently personal expressions, what accretes is “real” intimacy — just digitally-transmitted instead of verbally-expressed over a kitchen sink. And even in the case of one-sided communications, granting the emitter your attention gives them access to the same cognitive machinery.

"I've never met them but I know them."

More often than not, you're not wrong.

I've met a lot of people through twitter. Generally, they match my pre-IRL expectations. The ones who don’t use tend to use Twitter as if it were LinkedIn but without suit-jackets. They dispatch TwitterCard-sealed news from their particular domain, and they share little personal information. The one’s I do recognize upon first bio-meeting? They aren’t using twitter as a social bookmarking service — they’re being social through self-expression.

“I never met them but I know them.”

Admittedly, Twitter is a lot of different things to different people. And, it’s often a confusing place because limited context makes it hard to distinguish what mode someone is in when they tweet. But, oddly enough, things get easier with ambient intimacy. When you have a personal model of someone in addition to an intellectual one, you see more of them. That doesn’t mean you won’t be wrong. But it does make the context incomparably richer — and, social.

That’s the real reason I care about anti-abuse generally and community boundary enforcement, specifically. Hat-tip to J. Corwin for introducing the term to me, but we live in a noosphere.

There are lots of similar words used to say the same thing about our increasingly hyper-connected information environment, but I like this one because, to me, it conveys how new and weird things are. We’re only at the beginning of what’s even possible, let alone what’s necessary.

And, I think social networks designing for ambient intimacy is a necessity. The distinction between “irl” vs “online” continues to erode. That richer model of what someone is saying that ambient intimacy affords? It’s error-correction repurposed from the former environment and adapted for the latter one. It helps us correctly receive what someone really means to say. Even if what they’re saying is silly, that’s useful.

Postscript: I’m not working on reflock — what was meant to be my twitter anti-abuse tool — anymore. There are a lot of reasons for that but it looks like Tracy Chou’s app mirrors a lot of what I was implementing. I haven’t used it so this isn’t an endorsement, but please signup for the waitlist if you have abuse problems. I hope it works well because I want you to share yourself freely!

At the time of writing, Barack Obama has over 110,000,000 Twitter followers.

Barack Obama is a supernode.

[Image caption: A digraph that has 199 nodes dependent on 1 central one that I did so this post has a media card for embedding but that, by happy accident, created some pixelated artifacts that resonate with the spirit of this post.]

The Technical Problem

As a service specification, assume:

1. Subscribers (i.e. followers) should observe an Obama tweet quickly (low latency).

2. No subscriber has privileged access in that all users share the same expectation of observability (fairness).

As a world-scale, real-time, distributed (albeit centralized) platform, their developers spend considerable resources on engineering low-latency bit plumbing. And, in the practical terms of end user experience, the distance in time between action and feedback matters.

But, consider the challenge of materializing Obama’s tweet in subscriber feed,

• If you manage an on-disk circular queue of tweet ID’s for each user timeline and punt the full content to lookup requests on caching infrastructure, you’ll have to fan-out a 64 bit snowflake (tweet ID) to 110,000,000 disk or memory (or both) locations. That’s nearly a gigabyte of instantaneous writes.

• If you were to send the full tweets instantly to every connected, subscribing client, N, and assumed that the size of the delivered tweet was 296 bytes — 280 characters plus two 8 byte integers for the snowflake and user ID — you’re writing (296 x N) bytes of payload alone to your network wire. With a million users, that’s about 280 MiB ignoring network overhead.

These are extremely naive simplifications. There are much smarter ways of doing things that account for social network dynamics and use reasonable heuristics.[1] But, it should be clear that the technical problem is not trivial, especially once you recall that other users are also active at any given moment.

As for (2) — well, realistically, it’s an intractable specification. Assume Obama tweeted from his Kalorama residence while connected to the nearest data center in Virginia.[2] My friends in Maryland will probably receive his tweet before me. In reality, packets traverse complex routes before arriving at their digital destinations. But, it’s reasonable to simplify by saying whoever is closer as the photon flies gets their 280 characters first.

There are technical means of mitigation (albeit with wild uncertainty and impressive computational expense). If you had a map that associated every subscriber with a distribution of their read latencies conditioned by datacenter as well as one for inter-datacenter latencies, you could use both to randomly select a distribution path for each tweet such that the shared expectation converges on a single value for all subscribers.[3]

Alternatively, you could discard (2) by asserting: sub-second arrival-time heterogeneity and the associated information arbitrage opportunities (and risks) are not the platform’s concern — which, I think, is perfectly reasonable.

Twitter isn’t an enterprise-grade messaging queue. It’s a social networking service.

The Social Issue

Taking a glance at twitter’s engineering blog, you’ll quickly discover that Twitter cares deeply about (1). I have no idea if they care about (2) — I just made it up. It’s reasonable-sounding only if you’re thinking in the wrong context or too many of them. Fairness is a laudable concern and entities like high-frequency traders do seem to systematically exploit technicalities in parasitic ways, but complexity is expensive and the boundary separating problems you can attend to from the fantasy land of those you wish you could is much nearer than you think. If you want an aphorism: pick your battles. If you want a philosophy: do one thing, and do it well.

So, what is Twitter’s one thing?

Simple: Acting in accordance with its fiduciary duty to investors.

This isn’t meant to be a provocative statement. Given that twitter is a corporation, it’s a legally-enshrined obligation. And, as a publicly-traded company, there are even more commitments attached. Of course, twitter does have to satisfy some demand, lest the users go elsewhere. Setting aside manufactured demand, you can simplify twitter as a multi-sided market for attention:

• advertisers buy it;

• twitter sells it;

• users offer it in exchange for that delicious content.

Certainly, there is justifiable cause for concern with adtech. And, as someone who seems to be perpetually aware of squandered attention yet capable of resisting anything but temptation, I have serious issues with attention allocation under this particular business model (as implemented). But in stark terms, I don’t think twitter could have existed in any other form, historically-speaking. And, if you follow my ExTrEmElY oNlInE life, it’s obvious that get a lot out of it.

So, what’s my problem? Well, you see dear reader, I’m a true believer.

I think the reason for twitter’s existence, specifically, and social networking services, generally, should and can be removing the constraints of time and space from human social interactions! To facilitate the discovery of ideas, people, and connections that otherwise might not be possible! To expand the set of what we may perceive!

This may be romanticized view, but I’m a romantic at heart and this is my newsletter that your subscribing to, so I assume I’ll be forgiven any (over-)indulgence here. But, what does this have to do with supernodes? Well, at least for me, they’re one of the clearest examples of engineers pursuing (brilliant) technical solutions to mischaracterized problems in compliance with business model demands and while ignoring the social ones. That is, allowing supernodes is a problem by itself because,

SUPERNODES RESTRICT OUR SOCIAL ATTENTION.

This isn’t to say that I believe what Obama has to say doesn’t matter or that it doesn’t warrant my attention. I just don’t think it’s reasonable to computationally-encode and grant anyone that level of write access to our collective attention on a social medium. That’s not a social interaction, it merely simulates one. And in the confusion, all manner of negative consequences manifest.

Admittedly, even with technically-imposed limits — say a 30,000 follower maximum capacity — I’d still be aware of Obama’s tweets. As the 44th President of the United States, what he says is culturally and politically relevant, and it would traverse the social graph quiet readily. But the difference between percolation and direct distribution is vast — or, at least, I believe it is. I want a social medium that facilitates social interactions and discovery, rather than one that further entrenches the pathological hierarchy users initially (at least, I hope), meant to escape.

TL;DR?

Fuck supernodes. We don’t need them. Social tools should design against them.[4]

1. For example, see: Feeding Frenzy: Selectively Materializing Users’ Event Feeds.

2. I don’t actually know where twitter’s datacenters are located.

3. This convoluted method is, hilariously, an absurd and still-wrong simplification.

4. For all I know, Twitter does try to attenuate supernode signal strength. I’d applaud them if they do. But, the thing is: I can’t inspect their code, and unless Blue Sky is successful, it’s still a proprietary protocol that makes delivery opaque. I want agency and transparency, and Twitter limits both in this context. In any case, I absolutely hate footnoting the concluding sentence, but it’s the right thing to do.

The quote itself comes from a talk Simon gave in 1969 titled, Designing Organizations for an Information Rich World. Having completed a Ph.D. in Computational Social Science (CSS), Simon was unavoidable. And, rightly so! But — as someone whose attention and interest seems to experience an almost infinite regress absent careful environmental shaping — this quote became a mixture of mantra and mission to me. The first time I read it was the first time I seriously asked, "well...are we designing for an information rich world?"

I don’t think we are.

We certainly live in an information rich world. But I think the operationalized design principle is quiet different. While there is a(n often realized) danger in hyperbole, I don't think it's hyperbolic to state plainly: attention is both for sale and frequently sold.

This isn't a new phenomenon.

The competition for attention itself is part of the Darwinian struggle. The existence of markets for attention — formal or otherwise — is old. That's not inherently bad. And, I'm not sure how much the proportion of our attention actively allocated by markets has changed over time. But, I also think this line of reasoning misses a critical point. If "software is eating the world," why can't it also better manage my attention?

I don't have an answer yet.

My knee-jerk reaction is that, it's easier to sell freely-given attention than to demand money for a service that allocates it in a user-controlled way. In the context of social technologies, this is especially true given the logic of Metcalfe's law. Better design principles may exist, but evolutionary pressures on corporate entities select for ones that forge exponential moats the fastest. Alternatively, maybe we do have great tools for attention allocation, it's just that "work expands so as to fill the time available for its completion." In either case, I want to ask: Is this the best we can do?

I don't think so.

I'd like to find out.

Knowledge discovery and curation is an inherently a social activity. Even if you could resign yourself to a (boring) life of quiet isolation, you wouldn't want to! From "don't eat those red berries," to "check out this fascinating research paper," we is greater than the sum of the I parts. But, there is room for computer-mediated improvements, and I intend on contributing to that effort.

Perhaps this means I'm building and hacking on niche products. Maybe my struggle to remain focused really does reflect some inherent deficit which demands specialized design. But, I doubt it. "Information overload" is a problem so common that it requires no elaboration for clear articulation. If you feel similarly afflicted, subscribe to this newsletter, where I'll share what I find along the way.