Page Type System

Page Type System

LongtermWiki uses a two-level classification system to categorize content:

1. Page Type (pageType): Controls validation behavior and quality scoring eligibility
2. Page Template (pageTemplate): Determines expected structure and applicable style guide

Understanding this distinction is critical: a page can have pageType: content (meaning it receives full validation) while also having pageTemplate: knowledge-base-risk (meaning it should follow risk page structure).

---

Page Types Overview

Type	Set By	Quality Scored?	Validated?	Count
content	Default	Yes	Full	≈450
stub	Explicit frontmatter	No	Minimal	≈10
documentation	Explicit frontmatter	No	No	≈10
overview	Auto (index.mdx)	No	Navigation only	≈50

---

content (Default)

What It Is

The default page type for all substantive knowledge base content. If you don’t specify a pageType in frontmatter, the page is treated as content. This is the workhorse type covering risks, responses, models, organizations, people, and all analytical content.

When to Use

• All knowledge base pages that should be graded, improved, and maintained
• Any page where quality matters and should be tracked
• Pages that will appear in improvement queues and quality reports
• You don’t need to explicitly set pageType: content - it’s the default

Validation Behavior

Content pages receive full validation:

• Quality grading: Must go through grade-content.mjs to get a quality score (0-100)
• Structural checks: Tables, citations, diagrams are counted and affect quality
• Style guide compliance: Must follow the relevant template’s style guide
• Staleness warnings: Old lastEdited dates trigger alerts
• EntityLink validation: All referenced entities must exist
• MDX compilation: Syntax errors fail CI
• Dollar sign escaping: Unescaped $ causes failures

Quality Scoring

Content pages are scored on four dimensions by grade-content.mjs:

Dimension	Scale	What It Measures
Novelty	0-10	Originality beyond summarizing sources
Rigor	0-10	Evidence quality, citations, quantification
Actionability	0-10	Usefulness for decisions
Completeness	0-10	Coverage of the topic

These combine into an overall quality score (0-100) plus automated metrics (wordCount, citations, tables, diagrams).

Important: Quality must ONLY be set by grade-content.mjs, never manually or by LLM self-assessment. The quality-source validation rule enforces this by checking that pages with quality also have ratings.

Relevant Frontmatter Fields

---
title: "Page Title"
description: "Brief description for search and previews"
# pageType: content  # Optional - this is the default
pageTemplate: knowledge-base-risk  # Which structure to follow
quality: 45  # Set ONLY by grade-content.mjs
importance: 60  # How significant for AI safety prioritization
lastEdited: "2026-01-31"
llmSummary: "One-sentence summary with key conclusions"
ratings:
  novelty: 4
  rigor: 5
  actionability: 3
  completeness: 5
metrics:
  wordCount: 1500
  citations: 8
  tables: 2
  diagrams: 1
todos:
  - "Add more quantitative estimates"
  - "Include counterarguments section"
---

Examples

• /knowledge-base/risks/accident/deceptive-alignment.mdx
• /knowledge-base/responses/alignment/interpretability.mdx
• /knowledge-base/organizations/labs/anthropic.mdx
• /knowledge-base/models/threat-models/takeover-scenarios.mdx

---

stub

What It Is

A stub is an intentionally minimal page that exists primarily for navigation or reference purposes. Stubs are explicitly marked and excluded from quality scoring because their brevity is by design, not a deficiency.

When to Use

1. Redirect pages: When content has moved or been consolidated elsewhere

   • Use seeAlso field to point to the primary page
   • Example: Open Philanthropy → Coefficient Giving after rebrand

2. Brief profiles: People or organizations that don’t warrant full pages

   • Just enough info to be useful in search/browse
   • Links to external resources

3. Deprecated concepts: Topics kept for link stability

   • Old terminology that redirects to current naming
   • Historical pages that shouldn’t be deleted but aren’t maintained

4. Intentional placeholders: Topics planned for future expansion

   • Better than a 404, but marked as incomplete by design

Validation Behavior

Stub pages are excluded from most validation:

• ❌ Not graded by grade-content.mjs
• ❌ Not checked by style guide validators
• ❌ Not included in improvement queues
• ❌ Not counted in quality reports
• ✅ Still validated for MDX compilation
• ✅ Still validated for EntityLink references
• ✅ Still validated for dollar sign escaping

Relevant Frontmatter Fields

---
title: "Open Philanthropy"
description: "Redirects to Coefficient Giving"
pageType: stub
seeAlso: "coefficient-giving"  # Points to the primary page
sidebar:
  order: 10
lastEdited: "2026-01-29"
---

The seeAlso field is particularly important for stubs - it tells readers (and automated tools) where the real content lives.

Content Expectations

A stub should contain:

1. Brief explanation of what this page is about (1-3 sentences)
2. Clear pointer to where comprehensive coverage exists (EntityLink)
3. Backlinks component to show what references this page
4. Optionally: minimal quick facts if genuinely useful

A stub should NOT contain:

• Detailed analysis or substantial content
• Multiple sections with headers
• Tables, diagrams, or extensive formatting
• Content that duplicates the primary page

Example

---
title: "Open Philanthropy"
description: "Redirects to Coefficient Giving after November 2025 rebrand"
pageType: stub
seeAlso: "coefficient-giving"
---
import {Backlinks, EntityLink} from '@components/wiki';

**Open Philanthropy rebranded to <EntityLink id="coefficient-giving" /> in November 2025.**

For comprehensive information about AI safety funding, grants, and the 2025 Technical AI Safety RFP, see the Coefficient Giving page.

<Backlinks />

---

documentation

What It Is

Documentation pages contain internal reference material about LongtermWiki itself - style guides, process documentation, code examples, and meta-content. They are excluded from content validation because they may contain intentional examples of “bad” formatting, placeholder text, or code snippets.

When to Use

1. Style guides: Documentation of how to write different page types

   • /internal/risk-style-guide/
   • /internal/response-style-guide/

2. Process documentation: How LongtermWiki systems work

   • This page (/internal/page-types/)
   • Rating system documentation

3. Internal planning: Strategy documents, project notes

   • /project/similar-projects/
   • /project/strategy-brainstorm/

4. Technical reference: Code examples, API documentation

   • Content database documentation
   • Script usage guides

Validation Behavior

Documentation pages are excluded from all content validation:

• ❌ Not graded by grade-content.mjs
• ❌ Not checked by style guide validators
• ❌ Not checked for content quality issues
• ❌ Not included in any improvement queues
• ✅ Still validated for MDX compilation (must render)
• ✅ Basic frontmatter schema validation

This exclusion is important because documentation pages often contain:

• Example code with intentional placeholders
• Demonstrations of what NOT to do
• Reference tables that don’t follow content patterns
• Technical details irrelevant to content quality

Relevant Frontmatter Fields

---
title: "Risk Style Guide"
description: "How to write risk analysis pages for LongtermWiki"
pageType: documentation
sidebar:
  order: 3
---

Documentation pages typically have minimal frontmatter - no quality, importance, ratings, or metrics since they’re not graded.

Content Expectations

Documentation pages should be:

• Clear and actionable for their intended audience (usually contributors)
• Well-organized with logical structure
• Up-to-date when processes change
• Self-contained or linking to relevant external resources

Unlike content pages, documentation can freely include:

• Code blocks and technical examples
• Placeholder text for illustration
• Checklists and process steps
• Screenshots and interface mockups

Examples

• /internal/page-types.mdx (this page)
• /internal/risk-style-guide.mdx
• /internal/rating-system.mdx
• /project/similar-projects.mdx

---

overview (Auto-detected)

What It Is

Overview pages are automatically detected from the filename index.mdx. They serve as navigation hubs for sections of the site, providing an entry point and links to child pages. You never set pageType: overview in frontmatter - it’s inferred.

When to Use

Any index.mdx file is automatically treated as an overview page. These exist at:

• Section roots: /knowledge-base/risks/index.mdx
• Category roots: /knowledge-base/organizations/index.mdx
• Topic clusters: /knowledge-base/responses/governance/index.mdx

Validation Behavior

Overview pages receive limited validation:

• ❌ Not graded by grade-content.mjs
• ❌ Not included in quality reports
• ✅ Sidebar structure validation (must be properly configured)
• ✅ Link integrity checking
• ✅ MDX compilation

Content Expectations

Overview pages should provide:

1. Brief introduction to the section (1-2 paragraphs)
2. Navigation structure - links to key child pages
3. Organizational logic - why pages are grouped this way
4. Optionally: summary tables, key takeaways across the section

Overview pages should NOT:

• Duplicate substantial content from child pages
• Require extensive maintenance as children change
• Include detailed analysis (that belongs in child pages)

Example

---
title: "AI Risks"
description: "Analysis of risks from advanced AI systems"
---

This section covers the landscape of risks from AI development and deployment.

## Categories

- **Accident Risks**: Unintended harmful behaviors from AI systems
- **Misuse Risks**: Deliberate harmful applications of AI
- **Structural Risks**: Systemic issues from AI development dynamics

## Key Pages

- [Deceptive Alignment](/knowledge-base/risks/accident/deceptive-alignment/)
- [Power Concentration](/knowledge-base/risks/structural/power-concentration/)

---

Page Templates

While pageType controls validation behavior, pageTemplate controls the expected structure and which style guide applies. A page can be pageType: content (fully validated) with different templates.

Available Templates

Template	Used For	Style Guide
knowledge-base-risk	Risk analysis pages	/internal/risk-style-guide/
knowledge-base-response	Intervention/response pages	/internal/response-style-guide/
knowledge-base-model	Analytical models	/internal/models-style-guide/
ai-transition-model-factor	ATM top-level factors	/internal/ai-transition-model-style-guide/
ai-transition-model-sub-item	ATM sub-factors	/internal/ai-transition-model-style-guide/
ai-transition-model-scenario	ATM scenarios	/internal/ai-transition-model-style-guide/

Template Detection

Templates can be:

1. Explicitly set in frontmatter: pageTemplate: knowledge-base-risk
2. Inferred from URL for some paths (e.g., /knowledge-base/risks/**)

Example with Both Fields

---
title: "Deceptive Alignment"
description: "Risk of AI systems appearing aligned during training but pursuing different goals"
pageType: content  # Optional - default
pageTemplate: knowledge-base-risk  # Required structure
quality: 67
importance: 85
---

---

Interaction with Quality Pipeline

What Gets Graded

Page Type	Graded by grade-content.mjs?	In improvement queues?
content	Yes	Yes
stub	No	No
documentation	No	No
overview	No	No

The quality-source Validation Rule

To enforce that quality ratings come from the grading script (not manual entry or LLM self-assessment), the quality-source rule checks:

• If a page has quality set
• But does NOT have ratings (novelty, rigor, etc.)
• Then it was set outside the proper pipeline → warning

This catches pages where someone manually typed quality: 75 or an LLM suggested a quality score during content generation.

Fixing Pages with Manual Quality

If you see: Quality (X) set without ratings - use 'npm run regrade -- page-id'

Run:

npm run regrade -- page-id

This will use Claude to properly grade the page and set both quality and ratings.

---

Quick Reference

Setting Page Type

# Content page (default - can omit)
pageType: content

# Stub page
pageType: stub
seeAlso: "primary-page-id"

# Documentation page
pageType: documentation

Checking Page Types

# Count pages by type
grep -r "pageType:" src/content/docs/ | cut -d: -f3 | sort | uniq -c

# Find pages without explicit type (default to content)
find src/content/docs -name "*.mdx" -not -name "index.mdx" | \
  xargs grep -L "pageType:" | wc -l

# List all stubs
grep -l "pageType: stub" src/content/docs/**/*.mdx

Validation Commands

# Run all validation
npm run validate

# Run quality-source check specifically
npm run crux -- validate unified --rules=quality-source

# Grade a specific page
npm run regrade -- page-id

# Grade all ungraded pages
node scripts/content/grade-content.mjs --skip-graded --apply

---

Adding a New Page Type

If you need a new page type (rare), follow these steps:

1. Update schema in src/content.config.ts:

   pageType: z.enum(['content', 'stub', 'documentation', 'new-type']).optional()

2. Update validation rules in scripts/lib/rules/quality-source.mjs and other rules that check pageType

3. Update PageStatus component in src/components/wiki/PageStatus.tsx

4. Create style guide at /internal/new-type-style-guide/

5. Update this documentation

Most new content needs should be handled by adding a new pageTemplate, not a new pageType. Only add a page type if you need fundamentally different validation behavior.