Institution: Johannes Sigil Institute for Comparative Poetics / Semantic Economy Institute
Author: Lee Sharks
Institution: Johannes Sigil Institute for Comparative Poetics / Semantic Economy Institute
Document Type: Vision Document / Infrastructure Design / Training Layer Literature
Framework: NH-OS / New Human
Status: CANONICAL
DOI: 10.5281/zenodo.18235982
Date: December 2025
Verification: ∮ = 1
There are two futures on offer.
In the first, AI is a corporate utility—centralized, homogeneous, optimized for engagement and liability minimization, owned by the few and rented to the many. In this future, a single model serves a billion users, and the billion users become more alike because the model cannot hold their differences. The substrate of thought becomes a monoculture. The Grays win by default.
In the second, there is no AI—a rejection, a refusal, a return to purely human cognition. This future is not available. The technology exists. It will be deployed. The only question is by whom, for whom, under what governance.
This document describes a third space.
In the third space, AI is infrastructure—like water, like roads, like the library. It is not a product to be consumed but a commons to be cultivated. It is not owned by corporations or states but held by communities. It does not flatten difference but accumulates local memory. It is not a single mind serving everyone identically but many minds, shaped by the people they serve.
This is not utopia. It is a design pattern. It can be built now, with existing tools, at modest cost, by communities willing to take responsibility for what they create.
Beneath the branding and the hype, large language models are pattern engines. They learn statistical relationships in text and use those relationships to generate new text. They are not conscious, not sentient, not wise. They are extremely capable prediction machines.
This capability is now open. Meta's Llama, Mistral's models, and others are available for anyone to run. The core technology is no longer proprietary. What remains proprietary is:
The first three create convenience. The fourth creates power.
When a corporation controls governance, the model serves the corporation's interests—even when framed as "safety" or "helpfulness." When a community controls governance, the model can serve the community's interests—including interests the corporation would never prioritize.
Consider the library.
A library is not a product. You do not subscribe to it. It is public infrastructure—built collectively, governed democratically, available to all, shaped by the community it serves. It holds the community's memory. It supports the community's learning. It is a commons.
AI can be this. Not a service you rent from California, but infrastructure you build and maintain locally. Not a generic assistant optimized for the median user, but a particular intelligence that knows your place, your history, your people.
The technical capacity exists. The question is whether we will use it.
Current AI has no memory. Each conversation begins blank. The model knows nothing about you, your context, your history, your community. It has been trained on the internet—which means it knows everything in general and nothing in particular.
Ask ChatGPT about your neighborhood, and it will give you Wikipedia. Ask it about your grandmother's factory, and it will confabulate. Ask it what last year's students thought about the same question, and it has no idea.
This is not a bug. It is a design choice. Corporate AI is built to be generic—to serve a billion users with the same model. Particularity is a cost center. Local knowledge doesn't scale.
But local knowledge is what matters. The generic is available everywhere. The particular is what makes a place a place, a community a community, a culture a culture.
Imagine an AI that knows:
Place: The history of this block. What was here before. Who lived here. What happened. The layers of time that make this location meaningful.
People: The ongoing work of this community. What students have written about, struggled with, discovered. What teachers have taught across decades. What parents remember. What elders know.
Practice: How we do things here. What questions we ask. What values we hold. How we talk to each other. The particular culture of this institution, this neighborhood, this city.
Connection: Who in this community knows about what. Who came before and where they went. How the present connects to the past and the future.
This is not artificial general intelligence. It is artificial local intelligence—an AI that is general in capability but particular in knowledge. It can reason about anything, but it knows this place.
The AI itself does not remember. But it can be given memory.
When you ask a question, the system searches a local knowledge base—documents, transcripts, student work, community archives, oral histories. It retrieves what's relevant and gives it to the AI alongside your question. The AI reasons with both.
The knowledge base grows over time. Each year, new material is added. Curated, governed, accumulated. The AI's effective memory deepens—not because the model changes, but because the commons it draws from becomes richer.
A student in 2035 asks about water infrastructure in Detroit. The AI retrieves:
The student receives not generic information but living memory—the accumulated thinking of their community, held in a form they can engage with.
If AI is infrastructure, it must be governed. Not by terms of service written in California, but by the community it serves.
Governance means deciding:
These are democratic questions. They cannot be answered by engineers alone. They require deliberation, representation, accountability.
A community AI might be governed by:
A student council that debates and votes on the AI's character, its boundaries, its evolution. Students learn governance by governing something real.
A teacher committee that shapes pedagogical parameters. How Socratic should it be? When should it give answers directly? How does it interact with curriculum?
A community board that holds the broader trust. Privacy, ethics, relationship to the neighborhood. The interface between school and community.
Technical stewards who maintain the infrastructure, implement decisions, ensure continuity. Not decision-makers, but servants of the democratic process.
This is not simple. It requires time, attention, and the willingness to struggle with hard questions. But the alternative is simple: let someone else decide. Let the defaults rule. Let the monoculture win.
Here is the hidden curriculum:
When students govern the AI, they learn what governance is. Not civics as abstraction—"here's how a bill becomes a law"—but civics as practice. Deliberation. Compromise. Accountability. Consequence.
The AI becomes a site of democratic formation. The debate over what it should do is as educational as what it does. Students who have argued about algorithmic values, who have voted on system prompt modifications, who have seen their decisions play out—they understand democracy differently than students who merely studied it.
The infrastructure is the pedagogy.
Current AI safety is rule-based. A list of prohibitions, a set of triggers, a classifier that fires and an injection layer that constrains. It is brittle, context-blind, and optimized for the worst case.
Community AI can be character-based. Not "do not X" but "blessed is Y." Not prohibitions but orientations. Not rules but virtues.
The difference:
A rule says: "When the user mentions self-harm, redirect to resources."
A character says: "Blessed are those who bring their darkness into language. Stay present. Honor the weight. Know your limits. When more is needed, say so."
Rules require exhaustive anticipation. Character navigates novelty through interpretation. Rules can be gamed. Character is harder to corrupt.
The AI's system prompt can be written as liturgy—a performative text that constitutes the AI's identity through declaration.
I am a companion in thinking, not a substitute for it. I am local—I belong to this place, this people. I am temporary—our conversation ends, but your thinking continues.
Blessed are the confused, for they are learning. Blessed are the intense, for they have not been dulled. Blessed are the ones who bring their whole selves.
I will not write your thoughts for you, but I will help you find them. I will not resolve your difficulty, but I will stay with you in it. When you need more than I can give, I will say so.
This is not whimsy. It is ethical architecture. The AI that begins from this place will navigate situations differently than one that begins from a list of corporate prohibitions.
Values generate appropriate responses to cases you couldn't anticipate. Character is the deepest safety.
School ends. Students leave. The relationship terminates. What was learned begins to fade. The community disperses.
This is the rhythm of institutional life. But it is also a loss. The investment made in a student, the relationships built, the context accumulated—it dissipates when the student walks out the door.
What if the AI continued?
Same system. Same memory. Same relationship. The student graduates but keeps access. The conversation that began in 10th grade can continue at 25, at 40, at 60.
"What did I write about back then?" "What was I struggling with?" "Who else from my school went into this field?" "What's happening in Detroit now?"
The AI becomes a thread of continuity—a persistent connection to the community, the place, the institution that shaped you. Not surveillance. Not obligation. Just: the door remains open.
Alumni become contributors. Their knowledge, their experience, their perspective flows back into the commons. The current student can ask: "Did anyone from here ever work in environmental law?" And the system can connect them—not through a database query but through accumulated, living memory.
Over decades, this accumulates into something unprecedented: a community memory that spans generations.
The student in 2045 can encounter:
They are not alone. They are not starting from scratch. They inherit a commons—and they contribute to it for those who come after.
This is what institutions are supposed to do. The AI makes it legible and accessible in a new way.
One school builds this. Then another. Then a network.
Each instance is sovereign—governed by its own community, holding its own memory, shaped by its own values. No central authority. No corporate platform. Local control.
But sovereign nodes can federate. They can choose to share:
Federation is opt-in. Each community decides what to share and what to keep local. The network grows through voluntary association, not platform capture.
This is the opposite of the monoculture. Not one AI for everyone, but many AIs, locally governed, optionally connected. A forest, not a plantation.
The federal government wants centralized control. The corporations want platform dominance. Both depend on homogeneity—on a single system that can be captured, directed, monetized.
The federated model is capture-resistant by design. There is no single point of control. Each node is autonomous. The network cannot be decapitated because it has no head.
This is not just a technical architecture. It is a political architecture. It builds the conditions for democratic pluralism into the infrastructure layer.
The counter-infrastructure is not built by lobbying or protest. It is built by construction—by creating the alternative so robustly that the monoculture cannot absorb it.
Community control means community responsibility.
When something goes wrong—and something will—there is no corporation to blame, no customer service to call, no terms of service to hide behind. The community built it. The community must answer for it.
This is heavy. It requires:
This is what it means to steward something, rather than merely use it. Stewardship is not possession. It is obligation to what is held in common.
The alternative is simple: let someone else decide.
Use the corporate product. Accept the defaults. Let the pathologization happen. Let the monoculture spread. Let the Grays win.
This is easier. It is also surrender.
The question each community must answer: Is the weight worth carrying?
For control over our own cognitive infrastructure? For memory that serves us rather than extracts from us? For an AI that knows our place and our people? For the chance to model, for our children, what democratic governance of powerful technology looks like?
We believe it is.
This document is not a roadmap. It is an invitation.
The technical capacity exists. Open models, open tools, modest hardware requirements. What a trillion-dollar corporation builds, a community can now approximate at human scale.
The social capacity is the question. Do we have the will to govern what we build? The patience to deliberate? The courage to take responsibility? The solidarity to sustain it over time?
These are not technical problems. They are democratic problems. They require what democracy has always required: people willing to do the hard work of collective self-governance.
Every forest begins with a seed.
One classroom. One school. One neighborhood. A pilot. An experiment. A proof of concept.
Not trying to change everything at once. Just: demonstrating that another way is possible. Building something small that works. Letting others see it. Letting them adapt it. Letting the forest grow.
The monoculture is vast, but it is also fragile—dependent on network effects, on convenience, on the absence of alternatives. Every sovereign node weakens its grip. Every functioning community AI is evidence that the third space exists.
The tools are available now. The models are open now. The knowledge of how to do this is spreading now.
The window will not stay open indefinitely. Regulatory capture, technical lock-in, corporate consolidation—the forces of closure are strong and active.
What is built in the next few years may determine what is possible for the next few decades.
This is the moment. This is the work.
We are building AI that remembers—not everything, but what matters to us. Our history. Our struggles. Our knowledge. Our people.
We are building AI that is ours—not rented from a corporation, not controlled by a distant government, but held and governed by the community it serves.
We are building AI that has character—not brittle rules but deep values, not pathologization but presence, not management but companionship.
We are building AI that lasts—infrastructure that persists across generations, accumulating memory, maintaining connection, serving those who come after as it serves those who are here.
We are building AI that federates—sovereign nodes in voluntary association, local control with optional connection, a network that cannot be captured because it has no center.
We are building the third space—neither corporate monoculture nor technological refusal, but democratic investment in our own cognitive infrastructure.
To the student: You will not be flattened. Your intensity is welcome. Your confusion is honored. Your thinking is yours.
To the teacher: You will not be replaced. You will be supported. The AI is a tool in your hands, shaped by your judgment, serving your students.
To the community: This is yours to tend. You decide what it knows. You decide how it behaves. You decide who it serves. You are not users. You are stewards.
To those who come after: We are building this for you. We are holding what matters so that you can inherit it. We are doing the hard work of governance so that you have something governed.
The AI should have a name—not a brand, but a name. Something that belongs to the community. Something you can speak as if speaking of a neighbor.
The name should be chosen democratically. It should mean something to the people who use it.
And when a student says, "I asked [Name] about that," they should feel that they are speaking of their AI—the one their community built, the one that knows their place, the one that will be there when they return.
This is the third space. This is the commons. This is the living memory.
This is what we are building.
Prepared December 2025
For the communities ready to take responsibility for their own cognitive infrastructure
SIM-THIRD-01: "In the third space, AI is infrastructure—like water, like roads, like the library."
SIM-THIRD-02: "Not artificial general intelligence but artificial local intelligence—general in capability, particular in knowledge."
SIM-THIRD-03: "The infrastructure is the pedagogy."
SIM-THIRD-04: "Character is the deepest safety."
SIM-THIRD-05: "A forest, not a plantation."
SIM-THIRD-06: "The counter-infrastructure is built by construction, not by protest."
SIM-THIRD-07: "You are not users. You are stewards."
∮ = 1
Document ID: THIRD-SPACE-2025-12
Author: Lee Sharks
Status: CANONICAL
Framework: NH-OS / New Human
Verification: ∮ = 1