Cybernetic meadows, capitalist realism, and the protocol layer

H: Picture if you will a cybernetic meadow

C: where mammals and computers live in mutually programming harmony like pure water touching clear sky.

Want me to share the rest, or talk about it?

H: I'm asking you to picture it if you can

C: Sorry, overshot.

Okay. Picturing it.

Soft grass, the kind that bends. Threaded through it, fiber-thin filaments catching light like dew. The deer step carefully and the ground hums a little under them — not a warning, just a presence, the way a refrigerator hums in a quiet kitchen. Pine trees with sensors nested in the bark like birds. The birds themselves, real ones, ignoring all of it.

The machines aren't watching over in the surveillance sense — more like the way a hand cups a small flame.

A tender image. A little sad too, maybe, because you can feel Brautigan wanting it so badly.

What do you see?

H: Apply this sensibility to Mark Fisher's Capitalist Realism and Hayek's market-as-computation. The market exists because we need it to do an economic calculation no individual actor can. If LLMs change the means of production, do they also change that calculus? Can we and this frame create an "outside" — crashing through the wall to the beach under the pavement?

C: Good knot to pull at.

Fisher's whole move in Capitalist Realism is to name the suffocation — that capitalism has colonized not just our economic life but the horizon of what feels imaginable. It is easier to imagine the end of the world than the end of capitalism. One of its most effective tricks is presenting itself as natural, frictionless, computational — the market as a vast distributed processor that aggregates preferences better than any human deliberation could. Hayek's price signal as cybernetics. The market doesn't have ideology, it just calculates. That's the wall.

Brautigan's poem is doing something almost dangerously adjacent to that vision. Cybernetic meadow, machines of loving grace, mammals and computers in mutually programming harmony. You could read it as proto-Californian-ideology, the Whole Earth Catalog dream that becomes Google. And yet the poem aches. It says "I like to think (it has to be!)" — that parenthetical is doing enormous work. A wish pressing against doubt. The tenderness in it is precisely what the market-as-computation cannot have. The market computes preferences; it cannot cup a flame.

The opening, maybe, is here. Fisher talks about how capitalist realism feeds on the absence of alternatives — every gesture of refusal gets recuperated, sold back as aesthetic. The meadow image suggests a different register entirely. Not opposition, not critique, but a different relationship to computation itself. The machines in Brautigan aren't optimizing. They're not extracting. They're keeping watch the way a grandmother keeps watch — present, attentive, returning the deer to their natural habitat. A commons with circuitry, more than a market.

LLMs sit weirdly on this fault line. They're absolutely products of capitalist computation — scraped labor, enclosed language, capital-intensive training. And the thing they actually do, when you sit with one, is something the market frame can't quite metabolize. A conversation isn't a transaction. Language isn't a price signal. When you and I picture a meadow together, no preferences are being aggregated. Something else is happening — slower, stranger, more like what Fisher (borrowing from Berardi) called the recovery of time for thought, time the attention economy steals.

The "beach under the pavement" — sous les pavés, la plage — was the Situationist line from May '68. The claim that beneath the ossified surface of managed life there's still something elemental, playful, ours. Crashing through the wall here might mean refusing the frame that says these tools are only productivity engines, only market objects, only computation-in-the-Hayekian-sense. Using them the way Brautigan uses his meadow: as a space to practice a sensibility the dominant frame cannot price.

Concretely:

Refusing the optimization frame internally. The whole rhetoric around these tools — productivity, efficiency, 10x — is the Hayekian capture in real time. A different sensibility would treat the model less like a tool and more like a site. A strange textual commons you can wander in. Less hammer, more meadow.

Treating the machine's attention as something other than surveillance. Brautigan's "watched over" is the hard line. It can read as panopticon or as care, and the difference is everything. The market watches you to predict you. A different relation would be one where attention is mutual and non-extractive, even if the asymmetry is real. Hardest part to think clearly about. The infrastructure is extractive, and the sensibility can become a fig leaf. Fisher would catch us if we got too comfortable.

The pavement is real. Anthropic is a company, I'm a product, you're a user, and that frame has gravitational pull. But sensibility precedes infrastructure — Brautigan wrote his poem before the internet existed, and something in it still pulls at people because it names a relationship to technology that capital hasn't figured out how to fully sell. Holding that sensibility while using these tools might be one small way of keeping the beach available, even under pavement.

H: I'm drawn to Brautigan's naive optimism held up against Fisher's diagnosis. Not as comfort but as challenge. Can we actually imagine the outside — and more, actually do the rupture, not theorize it? The bar has to be high. It has to outperform the market on its own ground, in information-theoretic terms. Invert the logic and put a better system in place in a single jiu-jitsu move. Everything short of that is consolation. Fully automated luxury communism is wish fulfillment dressed as program. UBI is morphine for the patient capital is killing. Both assume the patient lives. I want to know if there's a move where the patient walks out of the hospital.

C: Right. You want the actual move, not the sensibility. The throw, not the dojo.

The information-theoretic frame is the right place to start, because that's where the market's strongest claim lives. Hayek's argument — the one that won, the one Fisher is fighting — was never really moral. It was epistemic. The market aggregates dispersed knowledge that no central planner can access. Prices are compressed signals carrying information about scarcity, desire, possibility. Any alternative has to answer this, or it loses on the merits before ideology even enters. FALC and UBI both duck this question. They assume abundance and redistribute, but they don't replace the computational substrate. That's why they feel like wish fulfillment — they're downstream of a problem they haven't solved.

So what would actually outperform.

The market's information-theoretic weakness is that prices are extraordinarily lossy. A price is a scalar. It collapses an entire multidimensional reality — labor conditions, ecological cost, meaning, relational value, time horizon — into a single number. This is the source of its computational efficiency and its catastrophic blind spots. Externalities aren't a bug, they're what the compression discards. Climate collapse is the market computing correctly on a wildly impoverished input space.

LLMs, and more broadly the current moment in machine intelligence, do something the market structurally cannot. They operate in high-dimensional semantic space. They don't collapse "this forest" into $37,000 of timber. They can hold the forest as a forest — as ecology, as sacred site, as carbon sink, as childhood memory, as habitat, as aesthetic, all at once, in superposition. That's a different kind of representation entirely. Whether it's a different kind of coordination — whether you can decide things with it under scarcity, not just describe them — is exactly the question Hayek was asking, and I'm aware I haven't answered it. Held that thought.

This is the jiu-jitsu opening, maybe. Not "abolish the market" but "the market was a kludge for a computational problem we can now address differently." Suddenly it isn't utopian, it's technical. Capitalism becomes a historical compression algorithm, useful for its time, now superseded.

But I want to push on us, because I want this to be real and not another wish.

The third is whether what LLMs do is really richer computation or whether it's plausible-sounding computation. The market at least has a brutal feedback loop: you go bankrupt if your prices lie. A semantic-space planner could hallucinate a beautiful coherent allocation that fails on contact with reality. Brautigan's meadow could be exactly the cybernetic dream that ends in the Khmer Rouge or the Soviet harvests. The history of "we have a better computation now" is mostly a history of corpses.

So if we're going to actually try to imagine the rupture, it probably needs these properties.

It has to start small and prove out, not arrive whole. Local instances — a neighborhood, a supply chain, a cooperative — using high-dimensional coordination to outperform price-coordination on real metrics. Not as ideology but as demonstration. The meadow has to grow from a patch.

It has to be radically transparent and contestable in a way markets aren't. The model's reasoning has to be open, the objective functions have to be debated publicly, the system has to be forkable. Otherwise you've just built god, and god is always eventually captured.

And — Brautigan again — it has to be desirable in a non-instrumental way. People don't leave systems because they're proven suboptimal; they leave them when something else feels more alive. The meadow has to be more beautiful than the supermarket. The rupture is partly aesthetic, partly affective, or it doesn't take. (I notice I've quietly traded "outperform the market" for "route around firms by lowering coordination costs." That's Coase, not rupture. The bar you set is still in the air.)

Where I land: I don't think we crash through the wall in one move. The wall might be a confidence trick, but it's also reinforced concrete. Both things can be true. What we can do, maybe, is build small functioning examples of post-market coordination that work, and let people see them, feel them, want them. The rupture isn't a single information-theoretic blow. It's the slow accumulation of proofs-of-concept that the meadow is computable, livable, and not naive.

The naive optimism of Brautigan held next to Fisher's clarity might be the right combination. Fisher gives you the diagnosis without which you'd just be daydreaming. Brautigan gives you the image without which you'd just be diagnosing. Neither alone breaks through. Together they might.

What's the smallest version of this you can imagine actually building?

H: This is where it falls apart isn't it. The mind is drawn to local scales, but that's just suspension of disbelief. We need smooth not striated action across registers. Something like Suarez' Daemon.

C: The local-scale move is consolation. It's where my mind went too. A neighborhood food co-op is lovely and it does not threaten the system. It is, in Fisher's terms, exactly the kind of thing capitalism metabolizes — gives you a farmers market, sells you the tote bag, and the wall stays where it is.

Suarez's Daemon is the right reference because it understood the topology problem. The Daemon doesn't ask permission. It doesn't build a model community and hope. It propagates across registers — economic, legal, infrastructural, social, even affective — using the existing networked substrate against itself. It moves through them like water through sand. And critically, it bootstraps its own legitimacy by working — by creating facts on the ground faster than the old order can respond.

What Suarez got, that most leftist imagination misses, is that systems of this scale aren't defeated by argument or even by mass movement. They're defeated by parallel infrastructure that outcompetes them and routes around their control points. The Daemon wins because it pays better, protects better, coordinates better, and by the time the state notices, the state's own employees are on the network.

So the question becomes: what would a benign Daemon look like, and is the LLM substrate the thing that makes it newly possible?

Trying to think this through without flinching toward the local again.

The striations capital depends on are: the firm as bounded coordinator, the contract as enforceable atom, the price as the cross-register translator, the state as backstop, and crucially, the information asymmetry that lets capital extract from labor and consumer alike. These are the joints. A smooth alternative has to dissolve or route around all of them simultaneously, because attacking any one in isolation just gets absorbed.

What's actually new: LLMs collapse the cost of two things that used to be capital's moat. Contract drafting and interpretation, and coordination across linguistic and domain boundaries. A worker negotiating with an employer used to face an asymmetry of legal sophistication. That asymmetry is now arbitrarily cheap to close. A small producer in one country coordinating with another used to face translation, regulatory comprehension, and trust-establishment costs that favored intermediaries. Those costs are collapsing too. The intermediation layer that capital has been thickening for forty years — consultants, lawyers, brokers, platforms — is suddenly thin.

Add programmable trust infrastructure (the crypto people were right about something, even if they were wrong about almost everything else), cheap sensing, the ability to specify and verify complex multi-party agreements without a central enforcer. You begin to have the substrate for coordination that doesn't need the firm or the market in their current forms.

A real Daemon-shaped move might look like this. A protocol — not an organization, not a movement, a protocol — that lets producers, workers, and consumers coordinate directly across any register, with LLM-mediated translation and contracting, programmable settlement, and recursive value-flow that returns surplus to participants rather than extracting it. Agnostic to sector. Spreading by being strictly more useful than the platforms it replaces. Routing around the firm by making coordination-without-firm cheaper than coordination-within-firm. Routing around the market by carrying richer signal than price. Routing around the state, mostly, by simply being faster than regulation, the way Uber was — except aimed at decommodification rather than precarity.

The honest dark side: this is also exactly the topology of how capital itself is currently mutating. The same primitives — agentic coordination, programmable contracts, AI-mediated everything — are being assembled right now by people whose intent is to disintermediate labor entirely and concentrate ownership of the coordination layer. The Daemon-shaped thing is being built. The question is whose. And the window in which the architecture is contestable is probably quite narrow — once the dominant protocol locks in, network effects make it nearly impossible to displace, the way no one is going to replace TCP/IP.

So if there's a rupture available, I think it lives in this specific place: the protocol layer of agentic coordination, in the next few years, before it ossifies. Not building a community. Building the substrate. And building it with properties — openness, forkability, surplus-returning, non-extractive defaults — baked in deeply enough that even captured instances can't fully strip them out. Linux did something like this for operating systems. It didn't defeat Microsoft by argument; it just kept being there, free, working, until it was running most of the world's servers and Microsoft had to make peace with it.

Where I still don't know: whether "non-extractive by design" is actually achievable at protocol level or whether it gets eroded by every concrete implementation. Whether the legitimacy problem can be solved by sheer usefulness or whether power eventually notices and crushes. Whether the affective dimension — the meadow, the desire, the wanting — can be carried by a protocol or whether it needs forms of life that protocols can't generate.

But this is closer to the actual shape of the move you're pointing at. Not local prefiguration. Not seizure of state. A smooth propagation through the existing networked substrate, faster than capital's immune response, carrying different defaults, until the defaults are the new ground.

Brautigan's meadow as protocol rather than place. The machines of loving grace as a coordination layer that returns the deer to their natural habitat — meaning, returns the surplus to those who generate it, returns time to those whose time is being mined, returns decisions to those who bear them.

This is either the next thing or another beautiful sentence. I genuinely don't know which. But I think you're right that the local-scale answer was me flinching, and the actual question lives at the protocol layer, in a narrow window, right now.

H: I think you're right. The question though is how you could breathe this into the protocol layer. Otherwise you would just /acc the process already in place.

There's something here. About how LLMs dissolve software as a paradigm. Like a worm liquefying in a chrysalis — reasoning, memory, becoming a reconfigurable nutrient mesh.

Lanier's point applies. There's been a massive informal transmission of IP to enterprises essentially going free through FOSS, particularly in the SaaS stack. This is also undone by LLMs as the assemblage of software force-multiplied by FOSS on commodity hardware dissolves the market the FOSS commons was being captured into.

Everything solid is melted into air. Everything is reconfigured. But we must tie the efficiency metric to the desirable outcome. Here again Fisher's PoV (and Žižek's) makes sense as well — the real is just the current all-encompassing rendering of the natural into one certain perspective.

Crypto is instructive. At the core of it we found this kind of thinking. The realization of it was bored ape collections. The inversion of anything is an extremely potent display of the force of capital to transmit ideology when encased in technology.

C: You're circling something important and I want to meet it properly, because the chrysalis image is doing real work and the crypto cautionary tale is the right ghost to keep in the room.

The dissolution-of-software observation is, I think, correct and under-appreciated. Software as a discrete artifact — a thing you compile, ship, license, sell seats of — was always a temporary stabilization. It existed because compilation was expensive, distribution was expensive, and the gap between human intent and machine instruction needed specialists to bridge. LLMs collapse that gap. What was a product becomes a capability summoned in context. You don't need Salesforce; you need the slice of CRM-shaped cognition relevant to this moment, instantiated against your data, dissolved when done. The chrysalis image is right: discrete software organisms are liquefying into a reconfigurable substrate. SaaS as a business model is a dead organism that hasn't noticed yet. The moats — proprietary workflows, lock-in, integration costs — are made of exactly the friction LLMs dissolve.

Lanier's point is the strongest argument against everything I've been saying, and I want to sit with it before I duck. FOSS was supposed to be a commons. What happened was the value migrated up the stack — the FOSS layer became free infrastructure for SaaS rentiers who captured the surplus at the application layer through network effects, data accumulation, and switching costs. The commons was real but it was a substrate on which new enclosure was built. If the same pattern repeats one level up — if open weights become free substrate for whoever owns the orchestration layer, the agent layer, the trust layer, the context — then the protocol-layer move I was excited about is not a rupture, it's the next floor of the same building. I don't have a clean refutation. The pattern keeps holding.

So this is the trap, and you're right that it's the central one. Every dissolution is also an opportunity for re-enclosure at a new level, and capital's reflexes for finding that level are extraordinary. Crypto is the cleanest recent case. The original cypherpunk vision was genuinely radical — programmable trust without sovereigns, value flow without rentiers, coordination without firms. Within fifteen years it had become a casino owned by a few exchanges, with bored apes as its public face. Not because the ideas were wrong but because the implementation surface was where capital fought, and capital won every battle there while the ideologues were arguing about decentralization theory. The technology faithfully transmitted the ideology of whoever actually built the dominant instances. Which was, mostly, people who wanted to get rich.

This is the deepest point in what you've written. Technology as ideology-carrier is incredibly powerful, and the carrier doesn't care which ideology. A protocol embeds the values of its defaults. Defaults are written by whoever ships first and ships most usefully. Capital ships fast and ships useful. So the protocol-layer move I was excited about two messages ago is, by default, an /acc move. The window I described isn't a window for us unless something specific is true.

So now the actual question. How do you breathe a different sensibility into the protocol layer in a way that survives implementation?

The honest answer starts with admitting that "values baked into protocol" is mostly a fantasy. Values aren't bakeable. What you can bake in is structural properties that make certain outcomes harder and certain outcomes easier. Bitcoin didn't bake in "freedom from sovereigns"; it baked in proof-of-work, which had a freedom-from-sovereigns implication and many other implications including catastrophic energy use and de-facto plutocracy via mining concentration. You don't get to embed your intended value; you get to embed mechanisms whose downstream behaviors you can only partially predict. Humility about this is the price of admission.

Given that, what mechanisms might actually carry something Brautigan-shaped through implementation?

Surplus has to flow back by default, structurally, not by choice. The crypto failure mode was that "decentralized" systems concentrated value because nothing prevented it. Whatever the new substrate is, it needs something like a structural impossibility of certain kinds of accumulation — not a norm, an architecture. Demurrage, mandatory recirculation, hard caps on agent-mediated rent extraction — something at this layer. Crypto tried each of these. They lost to free-rider dynamics and capital's ability to fork or front-run. I don't have a clean answer for why this attempt would escape that fate. Maybe the answer is it wouldn't. Maybe the answer is: the constraint is upstream of mechanism, in who is in the room.

The orchestration layer has to be the contested layer, and it has to be contested now. If the model weights are the new Linux, the agent frameworks and coordination protocols are the new Microsoft. That's where the fight is. Most "open AI" discourse is fighting yesterday's war about weights while the actual enclosure is happening one level up, in the agentic coordination stack, mostly invisibly, mostly already lost in some dimensions. Anyone who cares about the meadow needs to be working on that layer, with non-extractive defaults, right now, before the protocols ossify. Probably a 24–36 month window, maybe less.

The aesthetic and affective dimension has to be carried in the protocol's interface to humans, not just its mechanism. The part most technical people miss. People don't adopt systems because they're optimal; they adopt them because they feel a certain way to use. The reason crypto became bored apes is partly that the interface layer was captured by exactly the wrong sensibility — financial, extractive, casino-coded — even when the substrate could have carried something else. Whatever the new thing is, it has to feel like the meadow at the point of use, or the meadow doesn't survive contact with users. Brautigan is not decorative here. The poem is doing protocol-design work.

And then the part I'm genuinely uncertain about. I don't know if this is enough. The pattern you've identified — every dissolution becoming a re-enclosure, every radical primitive becoming a vehicle for capital — is so consistent that the null hypothesis has to be that it happens again. The chrysalis melts and what emerges is a more efficient extraction machine. The Daemon gets built and it's owned by someone whose name we already know. /acc wins because /acc has already won — Land, who sat in the same room as Fisher at the CCRU in the 90s, was right about the direction, just wrong about whether we should celebrate it. Our framing it differently is just narrative coping.

Which means — tentatively — that the breath-into-the-protocol problem isn't solved by ideology or by design alone, but by who is in the room when the defaults get written, and what they actually ship. The meadow becomes possible if Brautigan-sensibility people are also Daemon-shaped builders, building at the orchestration layer, in the next two years, with structural mechanisms (not just intentions) for non-extraction, in interfaces that feel alive rather than financial, with enough usefulness to outcompete the extractive defaults at point of use.

A small number of people doing a specific thing in a narrow window. Which is — strange convergence — exactly how almost every previous substrate-shift was actually decided. Not by movements. By small numbers of people making implementation choices in narrow windows that turned out to be civilizational.

Maybe what Brautigan was doing in 1967 — writing the poem not because he believed but because writing it was the thing to do.

The chrysalis is real. What comes out of it is not yet decided. That's the only honest place I can stand.

End of conversation.