# Response to the OpenCivics PRD: Bioregional Knowledge Commons Visualizer

Thanks for sharing the PRD. The globe as a living network map is genuinely compelling — a bioregional knowledge commons *should* feel like something you can see, zoom into, and explore. The "front door / back door" framing (permissionless discovery via GitHub, deeper collaboration via the creation engine) is the right instinct for adoption.

I'm writing this because we've been in this space for a while and I think there's real overlap worth exploring — but I want to be honest about what "this space" means across different timescales. I've spent the last couple of years publishing research on bioregional knowledge commoning, covering participatory ontology design, agent-centric architecture, governance, and Indigenous Data Sovereignty in a [three-part article series](https://darrenzal.github.io/Quartz/BioregionalKnowledgeCommonsSummary). For several months I've been building production knowledge infrastructure (RegenAI) — entity extraction, deduplication pipelines, hybrid search. And for the past couple days I've been running a small federation experiment (Octo/KOI-net) in the Salish Sea bioregion. Most importantly, I work with a community of 40+ organizations, First Nations leadership, watershed groups, and academic institutions in the Salish Sea who are the actual people a "bioregional knowledge commons" would serve.

I'm not attached to any specific stack. Everything on the technical side is exploratory. What I do bring is the combination of deep thinking on this topic, real community relationships on the ground, and some early experiments that have surfaced useful lessons. I'd rather share those now than after they've been rediscovered independently.

## What We Bring to the Table

Rather than mapping our work onto the PRD's seven modules (which would imply we've "built your modules" — we haven't), I want to describe three assets that might be useful to this project.

### Published Research on Bioregional Knowledge Commoning

I have published a series on the conceptual, technical, and governance dimensions of bioregional knowledge commons. The topics directly relevant to the PRD include:

- **Participatory ontology design** — what we call "ontology commoning": collaborative, community-driven development of shared semantic structures using methodologies like Human-Centered Ontology Engineering (HCOME). The core idea is that the people who live in a bioregion should shape how their knowledge is organized, not just consume a schema someone else designed.
- **Ontological pluralism** — avoiding a single "master" ontology and instead embracing multiple valid ways of understanding reality. The articles explore "standpoint logic" as a framework and "ontological translation for interoperability" — allowing understanding across different conceptual models without forced homogenization. This is directly relevant to how the PRD's schema bridge should work.
- **Agent-centric architecture** — evaluation of Holochain and AD4M/Coasys as substrates for data sovereignty and peer-to-peer knowledge federation. Holochain's agent-centric model gives users sovereignty over their own data; AD4M's "Perspectives" and "Neighbourhoods" provide a framework for federated shared knowledge graphs with "Social DNA" as governance rules.
- **Indigenous Data Sovereignty** — FPIC (Free, Prior, and Informed Consent), OCAP (Ownership, Control, Access, Possession), and CARE principles as structural requirements, not optional metadata. The articles argue that governance of Indigenous knowledge requires TK Licenses and Labels, not just Creative Commons.
- **A 4-phase implementation roadmap** — from foundational community engagement and pilot ontology commoning, through core platform development, to scaling and long-term adaptive governance.

The full summary is [published here](https://darrenzal.github.io/Quartz/BioregionalKnowledgeCommonsSummary).

### Community Relationships — On the Ground in Two Bioregions

A bioregional knowledge commons isn't a technology problem — it's a community problem that needs good technology. Between us, we already have community roots in two bioregions, which creates a natural test case for cross-bioregional knowledge sharing.

**Salish Sea (my side):**
- **Victoria Landscape Group (VLG):** 40+ organizations and 35+ individuals working on bioregional coordination in Greater Victoria
- **First Nations leadership:** W̱SÁNEĆ Leadership Council and WSÁNEĆ College — relationships built on respect for sovereignty and consent, not extraction. Any knowledge commons work in this bioregion must center Indigenous governance.
- **Watershed groups:** 5 active groups (Bowker Creek, Goldstream, Colquitz Creek, and others) stewarding knowledge about their specific waterways and ecosystems
- **Academic institutions:** University of Victoria, Royal Roads University / Cascade Institute, Camosun College
- **Municipal leadership:** Local elected officials who have signed bioregional frameworks into Official Community Plans
- Patricia Parkinson, who is connected to OpenCivics, also lives on Salt Spring Island in the Salish Sea — a bridge between the two communities
- **Adjacent projects:**
- **IndigenomicsAI** — a research program with the Indigenomics Institute (Carol Anne Hilton) developing AI systems that reason within Indigenous economic frameworks rather than forcing everything through Western economic logic. Running on NVIDIA H200 GPUs at the TELUS Sovereign AI Factory (Canadian jurisdiction, OCAP-compliant by architecture). Four research tracks: sovereign knowledge networks using ValueFlows, Indigenous ontology development, oral history processing with community consent workflows, and biocultural network intelligence using graph transformers. This is directly relevant to the PRD's knowledge commons vision — it's a concrete example of why ontological pluralism and sovereign infrastructure aren't abstract principles but operational requirements. If the knowledge commons can't hold Indigenous economic worldviews on their own terms, it's failed.
- **Salish Sea Dreaming** — an art/technology installation that serves as a public-facing prototype for IndigenomicsAI, exploring bioregional consciousness through AI and creative practice. Salt Spring Island art show planned for April 2026.
- **Regenerate Cascadia** — the meta-network VLG is part of, with seed groups forming across multiple bioregions in the Cascadia region
- **Bioregional Embassy** — diplomatic infrastructure for bioregional governance

**Front Range (your side):** You have your own bioregional network in Boulder and the Front Range. I don't know the specifics, but I know the bioregional movement is active there.

**The cross-bioregional opportunity:** The fact that we're rooted in two different bioregions — Salish Sea and Front Range — is actually the interesting part. Rather than testing knowledge commons protocols within one community, we could test *between* two, which is where the hard problems live: schema bridging, consent across governance boundaries, ontological translation. Cascadia and the Front Range could be the first real cross-bioregional test of whether any of this actually works.

### Technical Experiments (Early, Exploratory)

Two layers of technical work, at very different maturities:

**RegenAI (months of production work):** A centralized knowledge infrastructure that's taught us a lot about what works:
- 8 data sensors ingesting from diverse sources
- 88,000+ extracted entities with three-tier deduplication (exact → fuzzy → semantic)
- 163,000+ RDF triples in a knowledge graph
- 99.7% entity quality score (precision matters more than volume)
- Hybrid RAG search combining structured graph queries with vector similarity
- Key lesson: deduplication is the hard problem. Without it, knowledge graphs become unusable fast.

**Octo/KOI-net (days old — a prototype):** A federation experiment running in the Salish Sea bioregion:

┌──────┬──────────────────────────┬─────────────────────────┬──────────┐
│ Node │        Bioregion         │         Status          │ Entities │
├──────┼──────────────────────────┼─────────────────────────┼──────────┤
│ Octo │ Salish Sea (coordinator) │ Healthy                 │ 177      │
│ GV   │ Greater Victoria (leaf)  │ Healthy                 │ 5        │
│ CV   │ Cowichan Valley (leaf)   │ Healthy, just connected │ 1        │
└──────┴──────────────────────────┴─────────────────────────┴──────────┘

This is a 2-week-old experiment, not a mature system. But it proves the holonic federation pattern works technically — coordinator/leaf topology, event-driven state sync, cross-node entity resolution. We're running it to learn, not to claim we've solved federation.

## Things We've Been Thinking About

These are questions and patterns that have come up through the research and the building. I'm sharing them as things we've been sitting with, not as prescriptions.

### Federation as state vs. query routing

The PRD's federation layer is synchronous and query-centric: a node asks, the network answers. This works well for discovery. But in our (very early) experience, it breaks down when you need provenance, caching, or eventual consistency across nodes that go offline and come back — which is how community networks actually behave.

In the Octo experiment, we moved to an event-driven model. Every knowledge change is a signed event (NEW, UPDATE, FORGET) with a TTL and confirmation flow. Nodes subscribe to event types they care about, cache locally, and confirm receipt. A node can go offline for a week, reconnect, and catch up.

I'm genuinely curious how OpenCivics is thinking about this. Query routing is great as a UX layer — but does an event stream underneath give you the state coherence federation needs to be trustworthy over time? Or is there a different approach we haven't considered?

### The O(N²) bridge problem

The `bridge.yaml` pattern assumes every node maintains explicit mappings to every other node it wants to interoperate with. The math:

- **10 nodes** → 45 bilateral bridges
- **50 nodes** → 1,225 bridges
- **185 bioregions** (One Earth count) → 17,020 bridges

Each bridge needs a human to review and maintain it. This doesn't scale.

One approach we've been exploring (and that the published articles discuss under "ontological translation for interoperability") is a shared pivot vocabulary — a "bridge language" that each node maps to once. Each node maintains one mapping from their local schema to a commons ontology. Cross-network discovery goes through the pivot, not through bilateral bridges. Bilateral bridges can still exist for high-fidelity relationships between specific nodes, but the default path is through the commons.

This doesn't eliminate the mapping work — someone still has to agree on the pivot vocabulary, and that's a governance challenge as much as a technical one — but it reduces the scaling from O(N²) to O(N).

### Consent and mapping governance

This is easy to hand-wave and hard to get right. In the published research, and in the BKC Community of Inquiry & Practice (CoIP) work, consent isn't just metadata — it's a structural requirement. OCAP, CARE, and FPIC frameworks need to be first-class fields in schema and bridge definitions, not optional notes.

The articles specifically argue for TK Licenses and Labels for Indigenous knowledge (Creative Commons is insufficient for IKS), community-controlled governance of what gets shared across nodes, and provenance tracking at the entity level. Communities will walk away from tools that share their knowledge without clear governance.

This isn't theoretical for us. Through the IndigenomicsAI work with the Indigenomics Institute, we're grappling with exactly these questions: how do oral history recordings enter a knowledge system with full speaker attribution and community authority? How do you build AI that reasons *within* Indigenous frameworks rather than flattening them into Western categories? The answer starts with sovereign infrastructure (TELUS Sovereign AI Factory, Canadian jurisdiction, OCAP-compliant by architecture) and continues with consent workflows where nothing enters the system without explicit community approval. This is probably the hardest unsolved problem in the space, and it's one we're actively working on.

### The holonic pattern

This is the biggest architectural idea I want to flag. The PRD describes a flat peer network: nodes register, nodes federate, the globe shows them all. But bioregional knowledge doesn't organize in flat networks — it nests.

[Greater Victoria]   [Cowichan Valley]
        ↘                 ↙
   [Octo / Salish Sea]        ← holon boundary
              ↓
      [Cascadia Coordinator]   ← sees "Salish Sea" as one node

This is the holonic pattern: every part is also a whole. A bioregion is both a collection of sub-regions and a single entity in a larger network. This matters for:

- **Privacy**: a sub-region can share with its coordinator without being visible to the global network
- **Governance**: each level has its own consent rules
- **Scalability**: the global network sees dozens of bioregional holons, not thousands of leaf nodes
- **Subsidiarity**: decisions and knowledge stay at the most local level possible

The PRD's globe could represent this beautifully — zoom in on a bioregion and it expands into its sub-regions, each with their own knowledge graph. Zoom out and they collapse into aggregated nodes. But the federation protocol and registry need to support nesting for this to work.

Our Octo experiment is the tiniest proof-of-concept of this pattern — 3 nodes in one bioregion. We haven't proven it works at scale, across bioregions, or with real governance complexity. But the architecture reflects how bioregions actually organize, and that feels important.

## The Broader Ecosystem

A few projects are worth knowing about in this space:

**Regenerate Cascadia** is the meta-network the VLG is part of. There are seed groups forming across multiple bioregions in the Cascadia region. If the OpenCivics globe were to visualize a bioregional network, Cascadia has the beginning of one.

Other projects we've been tracking that might be relevant down the road: **Murmurations** (distributed data sharing protocol with 117,000+ organizational profiles — a potential complementary discovery layer), **AD4M/Coasys** and **Holochain** (agent-centric substrates for data sovereignty — discussed in the published articles). These aren't things we're building on currently, but they're in the same design space, and the OpenCivics visualizer might benefit from being positioned as a layer that can sit on top of multiple federation substrates rather than a monolithic stack.

## What We Don't Have

Honest accounting:

- **Globe visualization** — Nothing. The PRD's globe concept is exactly the kind of "front door" we're missing. Our knowledge graph is currently accessible through a static Quartz site and an API.
- **Onboarding UX** — Getting a new node connected requires SSH, systemd, and PostgreSQL knowledge. No self-serve portal, no wizard, no non-technical path.
- **Creator workflows** — Entity authoring happens through API calls and markdown files. The creation engine concept in the PRD is better by orders of magnitude.
- **Community reach at scale** — We have deep relationships in one bioregion. OpenCivics has a broader network of practitioners and institutional relationships.
- **Non-technical participation** — Everything we've built assumes a developer. The PRD's focus on accessibility is something we haven't prioritized at all.
- **Mature federation** — Octo is weeks old with 3 nodes. This is prototype territory, not production infrastructure.
- **Proven governance model** — We have frameworks from the published research but haven't implemented them end-to-end. The theory is ahead of the practice.
- **Cross-bioregional validation** — All our community work is in one bioregion. We don't yet know what breaks when you cross bioregional boundaries.

I'm listing these because collaboration works best when both sides know what the other actually brings and where the real complementarity is.

## A Concrete Starting Point

Rather than "let's run a pilot someday," here's what I think would be productive:

**Start with our two bioregions as a real-world test case.** The Salish Sea has 40+ organizations already mapped, active watershed groups, First Nations partners, and academic institutions. The Front Range has its own bioregional network. Together, they're the kind of cross-bioregional pair the OpenCivics globe is designed to visualize. Instead of a technical integration exercise, we could:

1. **Jointly design what a "bioregional node" should look like** using real data from both bioregions — not hypothetical schemas but actual organizations, practices, and ecological knowledge from the Salish Sea and the Front Range
2. **Test cross-bioregional knowledge sharing** — what happens when two communities with different governance structures, different ecological contexts, and different ontologies try to share knowledge? This is where schema bridging, consent protocols, and ontological translation get real.
3. **Test governance patterns within bioregions** — how does consent work when a First Nations community shares knowledge with a watershed group? When a local practice gets aggregated to a bioregional level? These are real questions we face with real communities.
4. **Use technical experiments as supporting evidence** — entity resolution demos, federation health endpoints, and knowledge graph queries can validate design decisions, but they shouldn't be the main event
5. **Surface the hard UX questions** — what does a non-technical community member actually need to contribute knowledge? What does the "creation engine" look like for someone who isn't a developer?

The technical integration pieces (registry format, health endpoints, entity resolution across protocols) would be part of this, but framed as "does this serve the communities" rather than "does this protocol work."

## Closing

I think we're close enough in vision that comparing notes would be more productive than building in parallel. What I bring isn't primarily a codebase — it's years of research on exactly this problem, months of lessons from building knowledge infrastructure, community relationships in the Salish Sea, and a willingness to start from scratch on the technology if there's a better approach. Combined with your community in the Front Range, we'd have a genuine cross-bioregional test case from day one.

I'd love to set up a working session — walk through the published research, compare community contexts, show the experiments, and figure out where the real complementarity is.

For anyone who wants the deep dive on the conceptual framework: [Bioregional Knowledge Commoning (full article series)](https://darrenzal.github.io/Quartz/BioregionalKnowledgeCommonsSummary)

— Darren

---

## Appendix: Technical Details

*Optional reading for those who want protocol-level specifics. Note: the Octo/KOI-net system described below is a weeks-old experiment. The RegenAI stats reflect months of production use.*

### Protocol Comparison

| Aspect | OpenCivics (PRD) | KOI-net (Octo) | Status |
|---|---|---|---|
| **Federation model** | Synchronous query routing | Event-driven (NEW/UPDATE/FORGET) with confirmation | Weeks-old experiment |
| **Node identity** | GitHub repo + manifest | ECDSA P-256 keypair + RID | Working between 3 nodes |
| **Schema handling** | `schema.yaml` per repo | 15 entity types, 27 predicates, three-layer mapping | Defined, tested against real data |
| **Bridge model** | Bilateral `bridge.yaml` | Shared pivot vocabulary + optional bilateral | Early, needs cross-bioregion testing |
| **Trust model** | GitHub-based (PRs, forks) | Signed envelopes (ECDSA), consent metadata per entity | Protocol works, governance not yet implemented |
| **Topology** | Flat peer network | Holonic (coordinator/leaf nesting) | 3 nodes, 1 bioregion |
| **Storage** | GitHub repos | PostgreSQL + pgvector per node (markdown + structured data) | Running in production |
| **Discovery** | GitHub registry | KOI-net node registry + coordinator gossip | Working at small scale |

### Entity Resolution (Three Tiers)

This comes from months of RegenAI production work, not just the Octo experiment:

1. **Tier 1 — Exact + Fuzzy match**: Normalized label comparison + Jaro-Winkler similarity. Catches "DFO" = "Dept. of Fisheries and Oceans", "Herring monitoring" = "herring Monitoring".
2. **Tier 2 — Semantic match**: OpenAI embeddings + pgvector cosine similarity. Catches "Indigenous fishing rights" ≈ "First Nations fisheries governance" when labels differ but meaning overlaps.
3. **Tier 3 — New entity creation**: If no match above threshold, a new entity is created with full provenance.

Cross-node resolution adds a fourth step: KOI-net cross-references link local entities to remote RIDs with confidence scores and relationship types (`same_as`, `broader`, `narrower`, `related`).

### System Stats

| System | Metric | Value |
|---|---|---|
| **RegenAI** (months) | Extracted entities | 88,000+ |
| | RDF triples | 163,000+ |
| | Data sensors | 8 |
| | Entity quality score | 99.7% |
| **Octo** (weeks) | Federated nodes | 3 (Salish Sea, GV, CV) |
| | Total entities | 177 |
| | Entity types in use | 14 of 15 |
| | Cross-ref resolution | Verified (`same_as`, confidence 1.0) |
| | Interop tests | 8/8 passing |