Excellent. This is the precision layer of KDOS. If domains are the skeleton, tags are the nervous system.

Below is KDOS Tag Ontology + Tagging Rules v1.0, designed to:

  • Stay flexible without becoming chaotic
  • Be AI-assignable with high accuracy
  • Support search, cross-linking, synthesis, and future embeddings
  • Scale to hundreds of thousands of tags without rot

KDOS Tag Ontology & Tagging Rules v1.0

0. Mental model (lock this in)

  • Domain = discipline
  • Tag = concept, object, method, school, or context

If a domain answers “what academic field?” a tag answers “what specifically?”

1. Tag categories (implicit, not stored)

Tags themselves are flat strings. But conceptually, they fall into categories — this helps AI choose well.

A. Conceptual tags

Abstract ideas and constructs.

Examples:

  • probability
  • causation
  • entropy
  • rationality
  • emergence

B. Method / technique tags

How something is done.

Examples:

  • bayesian-inference
  • monte-carlo
  • regression
  • optimization
  • meditation

C. Object / entity tags

Concrete or named things.

Examples:

  • fourier-transform
  • black-scholes
  • newtonian-mechanics
  • dna

D. School / framework tags

Named traditions or paradigms.

Examples:

  • stoicism
  • keynesianism
  • logical-positivism
  • behaviorism

E. Context / application tags

Where or why something is used.

Examples:

  • trading
  • clinical
  • education
  • policy
  • engineering-design

F. Meta tags

About knowledge itself.

Examples:

  • methodology
  • ontology
  • epistemic-risk
  • taxonomy

⚠️ Do NOT encode category into the tag string. The category is conceptual, not structural.

2. Tag formatting rules (strict)

These rules are non-negotiable.

2.1 Canonical format

  • lowercase
  • kebab-case
  • ASCII only
  • no punctuation except -

✅ Good:

bayesian-inference
signal-processing
epistemic-uncertainty

❌ Bad:

Bayesian Inference
signal_processing
epistemic/uncertainty

2.2 Tag length

  • Prefer 1–3 words
  • Avoid sentence-like tags

✅ decision-theory ❌ how-to-make-better-decisions

2.3 Plurals

  • Use singular unless the plural is standard

✅ derivative ❌ derivatives (unless referring specifically to plural instruments)

3. Tag count per entry (important)

Each KDOS entry must have:

  • Minimum: 5 tags
  • Maximum: 12 tags
  • Ideal: 7–10 tags

This ensures:

  • Enough cross-linking
  • No tag spam
  • Consistent AI behavior

4. Tag scope rules (very important)

4.1 Avoid domain leakage

Do not restate domains as tags.

If domain = statistics, do NOT tag:

  • statistics
  • statistical-methods

Tags should be below the domain layer.

4.2 Prefer specific over general

Always choose the most specific stable concept.

Prefer:

  • bayesian-updating over:
  • probability

Use both only if:

  • Both are genuinely central

4.3 Timelessness test

Ask: Will this tag still make sense in 20 years?

Good:

  • bayesian-inference
  • stoicism

Bad:

  • chatgpt
  • latest-research

5. Special tag classes (allowed)

These are explicitly allowed and encouraged.

5.1 Person tags

Format:

person-firstname-lastname

Example:

  • person-thomas-bayes
  • person-aristotle

Use only when the person is central.

5.2 Text / work tags

Format:

work-title-kebab

Example:

  • work-nicomachean-ethics
  • work-principia-mathematica

5.3 Formal object tags

For equations, laws, or theorems:

Examples:

  • bayes-theorem
  • noether-theorem
  • shannon-entropy

6. Disallowed tags (hard rules)

Never use:

  • Domain names (mathematics, philosophy)
  • Content type (article, post, note)
  • Quality judgements (important, basic)
  • Time-relative tags (modern, recent)
  • Platform names (youtube, twitter)

7. AI tagging rules (drop-in block)

Include this verbatim in generation prompts:

TAGGING RULES:
- Assign 5–12 tags.
- Use lowercase kebab-case.
- Prefer specific, timeless concepts.
- Do not repeat domains as tags.
- Do not invent meta-structure in tag names.
- Tags should represent concepts, methods, objects, schools, or applications.

8. Example: correct tagging

Topic: Bayesian Updating Domains: statistics, epistemology

tags = [
  "bayesian-inference",
  "bayes-theorem",
  "prior-belief",
  "posterior-probability",
  "epistemic-uncertainty",
  "belief-revision",
  "decision-theory",
  "evidence"
]

This is excellent KDOS tagging.

9. Optional: Tag governance (future-proofing)

Later, you may add:

  • tags/_index.md explaining tag rules

  • A script that:

    • Normalizes synonyms
    • Flags near-duplicates (bayesian-update vs bayesian-updating)

  • A “preferred tag list” (soft guidance, not enforcement)

But do not over-engineer now.