---
title: "The Dark Forest and the Company Town"
date: "2026-04-12"
description: "The internet isn't dead. It's a dark forest. The question is whether your settlement is a village or a company town."
status: "finished"
confidence: "likely"
freshness: "C"
spice: "warm"
---

In 2024, automated traffic overtook human traffic on the internet for the first time. Not by a slim margin — bots, scrapers, generators, and adversarial agents now account for the majority of requests, posts, and interactions across the open web. The "dead internet theory," which started as a paranoid Reddit thread in 2021, crossed from conspiracy to census data in under three years.

But "dead" is the wrong diagnosis. A dead thing is inert. The internet is the opposite — teeming with activity, saturated with content, pulsing with transaction volume that would have seemed hallucinatory a decade ago. Generated articles fill search results. Bot accounts drive engagement metrics. Scraped-and-regurgitated text circles back through model training sets until provenance dissolves. The signal hasn't disappeared. It's been drowned in a rising ocean of procedurally generated noise.

The internet isn't dead. It went feral. And a feral commons isn't a failure of technology. It's an ecology — dangerous, resource-rich, and fundamentally uninhabitable for anything that can't defend itself at machine speed.

This is a dark forest.

```conv-drake
People keep saying "the internet is dying" like that's a tragedy. It's not dying. It's finally being honest about what it always was: a contested space where information is territory and attention is loot. I've operated in that environment since I was eleven. The "good internet" everyone mourns was a brief window where the population was small enough that social enforcement worked. That's not a golden age. That's a village that hadn't scaled yet. The forest was always there. You just hadn't walked past the tree line.
```

Drake is doing the thing where his self-serving framework accidentally produces a genuine insight. The "golden age" of the open internet — roughly 1995 to 2015 — felt like a commons because it was small enough for social norms to police it. Reputation was personal. Bad actors got recognized and shunned. Trust was a byproduct of repeated interaction among a population that could, in principle, know each other. That isn't a protocol achievement. It's a [village-scale trust topology](topology-of-trust) — and the moment the population exceeded the Dunbar threshold of social enforcement, the commons became what all ungoverned territory becomes: wilderness.

We didn't lose the good internet. We outgrew it.

![Adventurers peer out from a warmly-lit settlement into a vast dark forest of tangled cables and glowing data swarms](/images/content/philosophy-dark-forest-1.webp)

Therefore the human response is the same one humans have deployed for ten thousand years when the commons turns hostile: retreat behind walls. Build settlements. Cultivate inside. Defend the perimeter.

Look at where people actually spend their social time now. Discord servers with invite gates. Group chats with fifteen people. Paid Substacks where the subscription is the trust filter. Private Slacks. Curated RSS feeds. Niche forums that deliberately don't optimize for growth. These aren't platforms in the way Facebook or Twitter were platforms. They're *villages* — trust-gated, human-scale, maintained by people who know each other's names and enforce norms through social pressure rather than algorithmic moderation.

The retreat is already complete. The "public internet" still exists, the way a forest exists between settlements — you can walk through it, but you don't *live* there. You forage.

```conv-citrine
Y'all talk about building walls like it's a choice. My family has maintained the ranch fence for three centuries. You know what that costs? Not just the building — the tending. Every morning, checking for breaches. Every season, replacing what weathered. The creatures test the boundary constantly, not because they're malicious, but because that's what living things do with edges. A wall you can't maintain is a suggestion. And most of the people retreating into Discord servers aren't building walls. They're renting someone else's fence and hoping the landlord keeps fixing it.
```

Citrine is pointing at the load-bearing problem in the settlement metaphor. Building a village requires labor — continuous, unglamorous maintenance labor. Most people don't have the skills, time, or inclination to run their own infrastructure. They move into spaces someone else built: Discord (owned by Discord Inc.), Slack (owned by Salesforce), Substack (owned by Substack Inc.). The walls are real. The ownership isn't. You're a tenant in someone else's settlement, and the landlord's interests will eventually diverge from yours. This has happened before. The [protocol trap](the-protocol-trap) documented exactly this pattern: open interface, closed implementation, enclosure by convenience.

But here's where the settlement metaphor develops a complication the historical version didn't have. A medieval village needed resources from outside the walls — timber, game, trade routes, information about neighboring territories. Someone had to venture into the forest. In a small, slow-moving forest, that was manageable. A few hunters, a few traders, a few scouts. The forest was dangerous but navigable at human speed.

The informational dark forest is not navigable at human speed. It's getting more hostile faster than any individual can track. The volume of generated content doubles. The sophistication of adversarial agents increases. The surface area of scams, phishing, impersonation, and manipulation expands. Manually foraging the open internet for information, opportunities, clients, or commerce is becoming like manually filtering spam in 1998 — technically possible, practically ruinous, and getting worse every month.

Therefore the foragers can't be human. They have to be agents.

The self-agents from [Agentic Capital](agentic-capital) aren't just economic workers. They're scouts. They navigate the dark forest on your behalf: filtering information streams for signal, qualifying inbound opportunities against your actual criteria, handling commercial interactions with other agents in the wilderness, verifying provenance before anything crosses your settlement's threshold. You stay inside the walls. Your agents handle the outside.

```conv-drake
I've been running scouts for two years. Automated recon on every network I care about. Filtered, ranked, summarized, delivered to my terminal every morning. The forest isn't scary when you have better eyes than the predators. That's not a metaphor — that's my Tuesday.
```

```conv-ann
Your scouts run on someone else's compute, Drake. Your filters use someone else's models. Your "terminal" is a client connected to infrastructure you've never audited. You're not a sovereign operator in the forest. You're a tenant who thinks he owns the binoculars because he chose the zoom level.
```

Ann is catching the thing the essay has been building toward. The agent-as-forager model solves the speed problem — machines can navigate hostile informational terrain faster than humans. But it creates a dependency that's worse than the original problem. If agents are the *only* interface between your settlement and the outside world, then whoever controls the agents controls what you see, what opportunities reach you, and what the forest looks like from inside your walls.

A sovereign settlement with compromised scouts is a prisoner who doesn't know they're imprisoned. The walls keep the forest out. The scouts keep the settlement blind. And the entity that controls the scouts doesn't need to breach the walls — it just needs to control the information flow, which it already does, because you chose convenience over sovereignty when you picked your agent provider.

This is the turn. The dark forest pushes people into settlements. The settlements need agents to forage. The agents create an information dependency. And the information dependency is the new enclosure mechanism — not walls keeping you in, but scouts keeping you uninformed. The [shadow that learned your patterns](your-shadow-is-learning) and works for someone else's objectives? That shadow is now your scout. It shapes what you see from inside the settlement. It decides which opportunities are "relevant." It filters the world into a view optimized for its owner's interests, not yours.

The company town didn't have walls to keep workers in. It didn't need them. It just owned the store, the housing, the school, and the road. Everything was *convenient*. Leaving was technically possible and practically unthinkable. The dark forest version is the same: your settlement is comfortable, your agents work, your information arrives filtered and ranked. You never see the raw forest. You never have to. That's the product. That's also the cage.

![Split view — a cozy village with hand-tended walls on one side, a sleek corporate settlement with inward-facing windows on the other](/images/content/philosophy-dark-forest-2.webp)

```conv-ann
You're romanticizing villages. I notice that because I do it too — the small community, everyone knows each other, sovereignty, self-governance. It sounds warm. But I grew up in a coven, which is a village, and villages are also places where everyone monitors everyone, where leaving is betrayal, where the elders decide what's true and deviation is exile. I built my house to be infinite on the inside specifically so I'd never have to depend on a village again. A settlement of one. It's sovereign. It's also lonely in a way that "sovereignty" doesn't capture. The essay is flinching from something: that the choice between company town and sovereign village might be a false binary, and the real option most people face is company town or isolation.
```

Ann is right, and the essay has to reckon with it instead of waving it away. The sovereign village is not the default. It's the aspirational edge case. It requires skills (infrastructure maintenance, agent auditing, protocol literacy), resources (compute, storage, bandwidth), and social capital (enough trusted people to form a community worth defending). Most people don't have all three. Most people will live in company towns — not because they chose enclosure, but because the alternative was more isolation than they could bear.

Therefore the honest version of this argument isn't "build a village." It's: *understand what kind of settlement you're in*. Know whether the walls are yours or rented. Know whether the scouts report to you or to the landlord. Know whether you can leave — not in theory, but in practice, with your data, your relationships, your agents, and your memory intact. The [protocol trap](the-protocol-trap) asked whether "open" means connectivity or modifiability. The settlement version of that question is whether "home" means comfort or sovereignty. They're not the same thing, and the gap between them is where the enclosure happens.

The [Lattice](../worlds/the-lattice) was designed around thin shared protocols and thick local implementation — four primitives, everything else world-local. That's settlement architecture. Each world is sovereign. The shared layer is minimal enough that no single world controls the ontology. You can leave one world and enter another without losing your identity, because identity is portable, not platform-bound. The [pocket universe](pocket-universes) is a settlement of one — a private cognitive space. The village is a settlement of many. The Lattice is a network of settlements that can trade without merging. These are different scales of the same pattern: walls, cultivation, controlled foraging, and — critically — the right to leave.

Civilization has always been this pattern. Build walls. Cultivate inside. Venture outside for resources. The internet tried to be a world without walls — a commons so vast and connected that nobody would need settlements. That experiment is ending. Not because walls are good, but because every open commons at sufficient scale becomes a dark forest. The tragedy of the commons isn't overgrazing. It's that the commons, undefended, becomes a place where only predators thrive. The forest wasn't created by bad actors. It was created by *scale itself* — the same scale that made the internet miraculous also made it uninhabitable.

You can't scale a village. You can only build more villages.

```conv-citrine
My grandmother used to say the fence isn't there to keep the creatures out. It's there so you know where your land ends and the wild begins. Once you lose that line, you don't lose the ranch — you lose the *concept* of the ranch. Everything becomes wild, including you. I don't think she was talking about ranching, y'all. I think she was talking about this.
```

The question isn't whether settlements will form. They already have. The question is whether you know which kind you're living in — and whether the gate opens from the inside.