Scope Expansion — A Tier 0 "Knowledge Work Foundations"

S5 · Instance 🔬 Research 2026-04-18

The insight

The scaffold’s “agentic kernel” is about half agent-specific and half knowledge-work-general. The general half has roots in cyberbase’s original SEACOW formulation and the broader knowledge-engineering / information-science literature. Those roots deserve their own home — as a tier 0 upstream project that this agentic-workflow project (and potentially others) inherits from.

Provisional name for the upstream: knowledge-work-foundations.

Decision status: accepted in principle, not yet executed. Execute after Phase 8 extraction of this project’s current kernel/stack.

Evidence

Cyberbase is where SEACOW actually came from

4 VAULTS/cyberbase/📁 54 - Obsidian Vault Organization/Knowledge Platform Organization Meta-Framework/Knowledge Platform Organization Meta-Framework.md (dated March 2025) contains:

The original SEACOW(r) derivation — S(ystem), E(ntity), A(ctivities: C/O/W/r)
An earlier “WORCS” formulation the user iterated toward SEACOW from
The r in SEA(COWr) deliberately marked as debatable — relation may or may not be first-class
The “Isn’t PARA enough?” debate — why PARA misses Output, Relation, and Entity
Naming conventions (MojiDex, folder-level-purpose rules)
Component synonyms — Capture/Input, Work/Processing, Output/Communication/Delivery/Publishing, System/Platform/Technology

Adjacent material in the same vault:

📁 54/Information Organization Systems/ — PARA, Zettelkasten, LYT MOCs, Johnny Decimal comparisons
📁 54/Folders vs Tags vs Links vs Metadata/ — the hierarchy-vs-graph-vs-tag debate
📁 54/Ideas for Knowledge Organization/ — terminology and ontology discussion
CybersaderNotion/04 Cybersader's Arsenal/Building a Knowledgebase.md — terminology map (ontology, taxonomy, epistemology, knowledge graph, symbolic knowledge distillation, semantic embeddings, NLP, philosophical research methods)

None of this material is agent-specific. It’s information architecture and knowledge-engineering thinking applicable to any knowledge platform — Obsidian, Notion, wiki, file share.

Principle audit — which of my 10 kernel principles belong upstream?

#	Principle	Agent-specific?	Tier-0 candidate?
01	Capture → Work → Output	No — GTD/CODE/Zettelkasten predate agents by decades	✅ yes
02	Temperature Gradient	No — Noguchi’s push-left filing predates computers	✅ yes
03	Skills vs Agents	Framed agentic, but dunamis/energeia is general cognitive science	⚠️ partially — the distinction is general; the LLM framing is tier 1
04	Progressive Disclosure	Nielsen UX version is general; LLM cache/context-rot framing is tier 1	⚠️ partially — keep UX-level in tier 0, LLM specifics in tier 1
05	Convention as Compressed Decision	No — DHH/Rails, Chesterton’s fence, convention-over-configuration	✅ yes
06	Single Canonical Addressability	No — Rosenfeld polar-bear, Bergman-Whittaker PIM research	✅ yes
07	Five Strata of Repeatability	No — Alexander’s pattern-language form, software layering	✅ yes
08	Four Channels of Context	Yes — LLM inference mechanics	❌ stays tier 1
09	Meta / Self-Reference	General — compiler/interpreter separation (Rust analogy)	✅ yes
10	Multi-Entity Design	Dual-framed — info science (Rosenfeld) + agents	⚠️ partially — the info-science half belongs in tier 0

Counted: ~6 fully upstreamable + 3 partially = majority of kernel material has a more-general home.

Proposed model

┌──────────────────────────────────────────────────────────────────┐
│  Tier 3: Work (this user's instance)                              │
│  ┌────────────────────────────────────────────────────────────┐   │
│  │  Tier 2: Stack (opinionated agentic toolkit)                │   │
│  │  ┌──────────────────────────────────────────────────────┐  │   │
│  │  │  Tier 1: Agentic Kernel                               │  │   │
│  │  │  (4 channels, skills-vs-agents LLM framing,           │  │   │
│  │  │   progressive disclosure for context windows,         │  │   │
│  │  │   kernel/stack/work meta-self-reference, etc.)        │  │   │
│  │  │  ┌────────────────────────────────────────────────┐  │  │   │
│  │  │  │  Tier 0: Knowledge Work Foundations              │  │  │   │
│  │  │  │  SEACOW framework, capture/work/output,         │  │  │   │
│  │  │  │  temperature gradient, single canonical          │  │  │   │
│  │  │  │  addressability, convention as compressed        │  │  │   │
│  │  │  │  decision, five strata of repeatability,         │  │  │   │
│  │  │  │  knowledge-engineering lexicon                   │  │  │   │
│  │  │  └────────────────────────────────────────────────┘  │  │   │
│  │  └──────────────────────────────────────────────────────┘  │   │
│  └────────────────────────────────────────────────────────────┘   │
└──────────────────────────────────────────────────────────────────┘

Composition rule extends upward

The dependency-direction invariant (see 2026-04-17-dependency-direction.md) applies to the 4-tier model identically:

From → To	Allowed?
Any higher tier → lower tier	✅
Any lower tier → higher tier	❌ (extraction breaks)

Tier 0 is the most universal; it only references itself + external literature (Rosenfeld, Luhmann, DHH, Alexander, etc.). Tier 1 references tier 0 freely. And so on.

Name choice

Working name: knowledge-work-foundations.

Alternatives considered:

Name	Pros	Cons
`knowledge-work-foundations`	Clear, broad, self-explanatory	Maybe too generic — “foundations” of what?
`kb-engineering-foundations`	Tighter tie to knowledge engineering	”KB” is jargon; less inviting
`information-architecture-foundations`	Directly cites the Rosenfeld polar-bear book lineage	Too academic; narrows the scope
`second-brain-foundations`	Forte/PKM resonance	Tiago Forte owns the branding; confusing
`seacow`	Brand consistency with cyberbase origin	Obscure; requires lookup; loses the opportunity to be explicit
`knowledge-platform-foundations`	Matches the cyberbase doc’s phrasing	Still a bit jargony

Final pick holds unless a better name surfaces in review. Provisional — can rename before the actual extraction happens.

What stays in this project (agentic-workflow) after tier 0 extraction

Strictly agent-specific material:

Principle 08: Four Channels of Context — LLM inference mechanics (weights, principal, environment, self)
Principle 03: Skills vs Agents (LLM-framed version) — with a pointer to tier 0 for the general cognitive-science foundation
Principle 04: Progressive Disclosure (LLM-framed version) — context rot, cache economics, lost-in-the-middle
Principle 10: Multi-Entity Design (agent half) — humans + AI agents as dual first-class consumers
Meta-agents, meta-skills, hooks, templates — all still tier 1/2/3
The 3-zone kernel/stack/work model (still load-bearing for extraction; only the name of the upstream tier changes)

Reduced from ~10 principles in tier 1 → maybe 4-5, each tighter and more focused.

Other potential specializations of tier 0

Once knowledge-work-foundations exists, it’s a parent to arbitrary sibling specializations:

agentic-workflow-and-tech-stack ← this project (agentic specialization)
research-methodology-foundations ← for grad students / academics doing literature synthesis
corporate-kb-foundations ← for enterprise knowledge teams
personal-productivity-foundations ← for non-agent PKM (Tiago Forte / Milo audiences)
writing-foundations ← for book-writing / long-form essay workflows

The key property: tier 0 is general enough that it pays rent for all of these. That’s the test of whether a piece belongs in tier 0 vs tier 1+: does a non-agentic audience benefit from it? If yes, tier 0.

Execution plan (rough)

Now (this research entry, done)

Capture the decision. Don’t execute.

During KB review

Add to each 01-kernel/principles/*.md frontmatter a hint:

would-live-in-tier-0: true  # for 01, 02, 05, 06, 07, 09; partial for 03, 04, 10

Makes the later extraction a grep.

After Phase 8 (this project’s kernel/stack/work extraction)

Scaffold cybersader/knowledge-work-foundations as a new private repo
Site structure: Starlight, similar to current, but NO 02-stack/ or 03-work/ tiers — just principles/, patterns/, and a lexicon (the cyberbase-derived terminology map)
Source material to pull in:
- Cyberbase’s Knowledge Platform Organization Meta-Framework → principles/01-seacow-framework.md
- Cyberbase’s Information Organization Systems → patterns/knowledge-systems-comparison.md (PARA, Zettelkasten, LYT, Johnny Decimal)
- Cyberbase’s Folders vs Tags vs Links vs Metadata → principles/02-hierarchy-vs-graph-vs-tag.md
- Cyberbase’s Building a Knowledgebase terminology → reference/knowledge-engineering-lexicon.md
- This project’s tier-0-candidate principles (from the audit above) → rewritten in neutral voice, with agentic framing stripped
Rework this project’s kernel to reference tier 0 + hold only truly agent-specific content
Update ROADMAP.md, CONTRIBUTING.md, README.md to describe the relationship
Same stratum framework, same extraction approach, just one level up

Eventual consequences

agentic-workflow-and-tech-stack becomes a specialization, not a generalist scaffold
Non-agentic audiences (researchers, writers, PKM enthusiasts) benefit from knowledge-work-foundations
Tier 1 becomes smaller + more focused on agentic concerns
The 4-level stack is explicit: tier 0 → tier 1 → tier 2 → tier 3

What this doesn’t change

Current monorepo organization (still 01-kernel/, 02-stack/, 03-work/, 00-meta/)
Phase 8 extraction plan for this project’s current tiers
The stratum frontmatter or dependency-direction invariant
Anyone’s current reading of the principles — they’re still valid; they just find better homes later

Candidate promotion

This research entry itself is candidate for promotion to the future knowledge-work-foundations project as an “architecture decision record” (ADR) explaining why that project exists.

When tier 0 is extracted, move this file to knowledge-work-foundations/adr/ or similar and update its frontmatter.