About This Wiki

The Longterm Wiki is a strategic intelligence platform for AI safety prioritization. It serves as a decision-support tool for funders, researchers, and policymakers asking: "Where should the next marginal dollar or researcher-hour go?"

This page documents the technical side—how the wiki is built, how content is organized, and how to contribute.

For strategic vision and goals, see:

• Vision Document — Full 2-person-year scope, architecture, milestones
• Similar Projects Analysis — Lessons from Arbital, Stampy, MIT Risk Repository, and others

Content Architecture

Major Sections

Content Volume

The wiki contains approximately:

• ~700 MDX pages across all sections
• Hundreds of structured data entities (organizations, people, AI models, legislation, grants, projects)
• Analytical model pages with causal diagrams

Page Types and Templates

The wiki uses a two-level classification system.

Page Types (Validation Behavior)

Page Templates (Structure)

Templates determine expected structure and applicable style guide:

For complete details, see Page Type SystemE739Documents LongtermWiki's four-level page classification system (content, stub, documentation, overview) with explicit validation rules for each type, where content pages receive full quality gradin...Quality: 65/100.

Quality System

Rating Dimensions

Content pages are scored on six dimensions (0-10 scale, harsh—7+ is exceptional):

These combine into an overall quality score (0-100).

Quality Pipeline

Quality must only be set through the grading pipeline, never manually:

# Improve a specific page (includes grading)
pnpm crux w improve <page-id> --tier=standard --apply

For grading criteria and workflows, see Content QualityE741This page documents LongtermWiki's content rating system, which combines LLM-graded subscores (focus, novelty, rigor, completeness, objectivity, concreteness, actionability on 0-10 scales) with aut...Quality: 48/100.

Data Layer

Three Bases

The wiki is organized around three complementary data layers:

YAML and MDX are the source of truth; Postgres (via the wiki-server) keeps read mirrors for querying.

Structured Data Sources

Legacy YAML catalogs live under data/:

Generated Data

Running pnpm build-data:content (or pnpm build, which runs it first) compiles YAML + MDX into:

The Next.js app reads these JSON artifacts directly at build time—content pages make zero runtime API calls.

Data-Aware Components

React components pull from the generated JSON to display structured information. For example, EntityLink provides stable cross-references, <FBF> and <Calc> render FactBase facts, and DataInfoBox displays entity profiles.

For database details and component usage, see Content DatabaseE759Documentation for the wiki's multi-layer data storage architecture. The wiki-server PostgreSQL database stores citation content, audit results, claims, facts, and structured data accessed via Hono ...Quality: 44/100.

Cross-Linking System

Entity Links

The wiki uses stable ID-based linking that survives path reorganization:

The <EntityLink id="E274" name="scheming">scheming</EntityLink> risk relates to
<EntityLink id="E93" name="deceptive-alignment">deceptive alignment</EntityLink>.

Benefits:

• Automatic title lookup from database
• Entity type icons
• Backlink tracking
• Link validity checked by the gate

Post-Edit Linking Check

After creating or editing a page, verify cross-linking with the validation gate:

pnpm crux w validate gate --scope=content --fix

This runs link integrity, EntityLink resolution, and related checks in ~15 seconds.

Visualizations

Mermaid Diagrams

Flowcharts, sequences, and graphs for static illustrations:

flowchart LR
  A[Training] --> B[Alignment Techniques]
  B --> C{Robust?}
  C -->|Yes| D[Safe AI]
  C -->|No| E[Misalignment Risk]

See Mermaid DiagramsE735Internal style guide for creating Mermaid diagrams in the wiki, recommending vertical layouts over horizontal (max 3-4 parallel nodes), providing semantic color palette, and advocating tables over ...Quality: 37/100 for guidelines.

Cause-Effect Graphs

Interactive causal diagrams using ReactFlow for complex causal models:

<CauseEffectGraph
  initialNodes={graphNodes}
  initialEdges={graphEdges}
  selectedNodeId="current-factor"
/>

Features: zoom, pan, minimap, node highlighting, path tracing, entity linking.

See Cause-Effect DiagramsE758Technical documentation for creating cause-effect diagrams in YAML format, covering node types (leaf/cause/intermediate/effect), edge properties (strength/confidence/effect), semantic color coding,...Quality: 44/100 for schema and examples.

Automation Tools

Unified CLI

All tools are accessible via the crux CLI, organized into groups by data layer:

pnpm crux --help            # Full CLI reference
pnpm crux w --help          # Wiki (content) commands
pnpm crux fb --help         # FactBase commands
pnpm crux tb --help         # TableBase commands
pnpm crux linear --help     # Linear issue tracker
pnpm crux gh --help         # GitHub (PRs, CI, deploy tasks)
pnpm crux sys --help        # System / agent tools

Common Workflows

Validation Rules

The gate check runs 50+ validators in parallel. CI-blocking rules include:

For complete tool reference, see Automation ToolsE757Comprehensive technical documentation for wiki maintenance automation, covering page improvement workflows (Q5 standards requiring 10+ citations, 800+ words), content grading via Claude API (~$0.02...Quality: 41/100.

Technical Stack

Core Technologies

Project Structure

longterm-wiki/
├── apps/
│   ├── web/                    # Next.js 15 frontend (longterm-next)
│   │   ├── src/
│   │   │   ├── app/            # App Router pages and API routes
│   │   │   ├── components/     # React components (wiki/, ui/, ...)
│   │   │   ├── data/           # Data-layer modules (tablebase, factbase)
│   │   │   └── lib/            # Shared utilities
│   │   └── scripts/
│   │       ├── build-data.mjs  # Data compilation pipeline
│   │       └── assign-ids.mjs  # Wiki ID allocation
│   ├── wiki-server/            # Hono API server (Postgres, Drizzle ORM)
│   ├── discord-bot/            # Discord integration
│   └── groundskeeper/          # Background maintenance jobs
├── packages/
│   ├── factbase/               # FactBase types, YAML loading, serialization
│   │   └── data/fb-entities/   # FactBase entity YAML (authoritative)
│   └── id-utils/               # Stable ID helpers (sid_ prefix)
├── content/
│   └── docs/                   # ~700 MDX pages
│       ├── knowledge-base/     # Risks, responses, organizations, people
│       ├── project/            # Public project documentation
│       └── internal/           # Contributor docs and style guides
├── data/
│   ├── entities/               # Lightweight entity YAML catalog
│   ├── *.yaml                  # Legacy YAML data sources
│   └── schema.ts               # Shared data schema
└── crux/                       # Crux CLI + 50+ validators
    ├── commands/               # CLI command handlers (w, fb, tb, linear, ...)
    ├── validate/               # Validators (gate, data, daily)
    └── crux.mjs                # CLI entry point

Key Configuration Files

Development Workflow

Getting Started

# Install dependencies (pnpm is required)
pnpm install

# First-time setup: install + build data
pnpm setup:quick

# Start development server on port 3001
pnpm dev

# Build for production
pnpm build

After Editing Content

# After editing any page, clean up formatting
pnpm crux w fix escaping
pnpm crux w fix markdown

# Run the pre-push gate (CI-blocking checks)
pnpm crux w validate gate --fix

# For content-only edits, use the faster content scope (~15s)
pnpm crux w validate gate --scope=content --fix

Creating New Content

1. New pages: use the Crux content pipeline — pnpm crux w create "Title" --tier=standard — rather than writing MDX by hand. It handles frontmatter, IDs, citations, and EntityLink validation.
2. New entities: add a YAML file to data/entities/ (lightweight) or allocate a wiki entity ID with pnpm crux tb ids allocate <slug> for entities that deserve their own page.
3. New components: add to apps/web/src/components/wiki/ and reference via the @components/wiki path alias.

For content generation workflows, see Research ReportsE748Internal style guide for deprecated research report format, specifying table-based layouts, escaped dollar signs, YAML schema, and causal factor documentation for diagram creation. Provides compreh...Quality: 20/100.

Related Documentation

Project & Strategy

• Vision Document — 2-person-year scope, architecture, milestones
• Similar Projects — Lessons from Arbital, Stampy, MIT Risk Repository
• Strategy Brainstorm — Failure modes and strategic options

Style Guides

• Knowledge Base Style GuideE763Internal style guide for wiki content creation, emphasizing flexible hierarchical structure over rigid templates, integrated arguments over sparse sections, and selective use of visualizations. Pro...Quality: 34/100 — For risk and response pages
• Model Style GuideE736Internal style guide prescribing dense, quantified content structure for model pages: minimum 800 words, 2+ tables (4x3+), 1+ Mermaid diagram, mathematical formulations, strategic prioritization co...Quality: 38/100 — For analytical model pages
• Stub Style GuideE753Internal documentation providing guidelines for creating minimal placeholder pages (stubs) in the knowledge base, including when to use them, required formatting, and when to convert them to full p...Quality: 19/100 — For minimal reference pages

Technical Documentation

• System ArchitectureE734Technical architecture of the Longterm Wiki — data pipeline, clever design patterns, and the ideas that make it workQuality: 73/100 — Data flow, pipelines, design decisions
• Page Type SystemE739Documents LongtermWiki's four-level page classification system (content, stub, documentation, overview) with explicit validation rules for each type, where content pages receive full quality gradin...Quality: 65/100 — Complete classification reference
• Rating SystemE741This page documents LongtermWiki's content rating system, which combines LLM-graded subscores (focus, novelty, rigor, completeness, objectivity, concreteness, actionability on 0-10 scales) with aut...Quality: 48/100 — Quality scoring system
• Content DatabaseE759Documentation for the wiki's multi-layer data storage architecture. The wiki-server PostgreSQL database stores citation content, audit results, claims, facts, and structured data accessed via Hono ...Quality: 44/100 — Storage architecture (PostgreSQL, caching, YAML)
• Automation ToolsE757Comprehensive technical documentation for wiki maintenance automation, covering page improvement workflows (Q5 standards requiring 10+ citations, 800+ words), content grading via Claude API (~$0.02...Quality: 41/100 — All scripts and workflows
• Cause-Effect DiagramsE758Technical documentation for creating cause-effect diagrams in YAML format, covering node types (leaf/cause/intermediate/effect), edge properties (strength/confidence/effect), semantic color coding,...Quality: 44/100 — Graph schema and examples
• Mermaid DiagramsE735Internal style guide for creating Mermaid diagrams in the wiki, recommending vertical layouts over horizontal (max 3-4 parallel nodes), providing semantic color palette, and advocating tables over ...Quality: 37/100 — Diagram guidelines
• Documentation MaintenanceE760Guidelines for keeping internal documentation accurate and up-to-dateQuality: 53/100 — Keeping docs updated