About This Wiki

About the Longterm 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:

Project Overview — Quick summary of what LongtermWiki is
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

Section	Purpose	Examples
Knowledge Base	Core content on risks, interventions, organizations, people	Deceptive Alignment, AI Safety Institutes
AI Transition Model	Comprehensive factor network with outcomes and scenarios	Factors, scenarios, quantitative estimates
Models	Analytical frameworks for understanding dynamics	Risk models, cascade models, governance models
Project	Public-facing documentation about LongtermWiki itself	Vision, strategy, similar projects
Internal	Contributor documentation, style guides, technical reference	This page, style guides, automation tools

Content Volume

The wiki contains approximately:

~550 MDX pages across all sections
~100 structured data entities (experts, organizations, cruxes, estimates)
~80 analytical model pages with causal diagrams

Page Types and Templates

The wiki uses a two-level classification system.

Page Types (Validation Behavior)

Type	Quality Scored?	Use Case
`content`	Yes	All substantive knowledge base pages (default)
`stub`	No	Redirects, brief profiles, placeholders
`documentation`	No	Style guides, internal docs (like this page)
`overview`	No (auto)	Index pages for navigation

Page Templates (Structure)

Templates determine expected structure and applicable style guide:

Template	Style Guide
`knowledge-base-risk`	Knowledge Base Style Guide
`knowledge-base-response`	Knowledge Base Style Guide
`knowledge-base-model`	Model Style Guide
`ai-transition-model-factor`	ATM Style Guide

For complete details, see Page Type System.

Quality System

Rating Dimensions

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

Dimension	What It Measures
Focus	Does it answer the title’s promise?
Novelty	Value beyond obvious sources
Rigor	Evidence quality and precision
Completeness	Thorough coverage of claimed topic
Concreteness	Specific vs. abstract recommendations
Actionability	Can readers make different decisions?

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

Quality Pipeline

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

# Grade a specific page
npm run crux -- content grade --page scheming --apply

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

For grading criteria and workflows, see Content Quality.

Data Layer

Structured Data Sources

The wiki maintains YAML databases in src/data/:

File	Contents
`experts.yaml`	AI safety researchers and their positions on cruxes
`organizations.yaml`	Labs, research orgs, funders
`cruxes.yaml`	Key uncertainties with expert positions
`estimates.yaml`	Probability distributions for key variables
`publications.yaml`	Research papers and reports
`external-links.yaml`	Curated resource database

Generated Data

Running npm run build:data generates:

Output	Purpose
`database.json`	All entities merged for browser use
`pathRegistry.json`	Entity ID → URL path mapping
`backlinks.json`	Reverse reference indices
`tagIndex.json`	Searchable tag index

Data-Aware Components

Components pull from YAML databases to display structured information. For example, EntityLink provides stable cross-references, while DataInfoBox displays expert or organization profiles from YAML.

For database details and component usage, see Content Database.

Cross-Linking System

Entity Links

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

import {EntityLink} from '@components/wiki';

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

Benefits:

Automatic title lookup from database
Entity type icons
Backlink tracking
Link validity checked during CI

Backlinks

Every page can display incoming links:

import {Backlinks} from '@components/wiki';

<Backlinks entityId="deceptive-alignment" />

Post-Edit Linking Check

After creating or editing a page, verify cross-linking:

npm run crux -- analyze entity-links <entity-id>

This shows inbound links, missing inbound links (pages that mention but don’t link), and outbound links.

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 Diagrams for guidelines.

Cause-Effect Graphs

Interactive causal diagrams using ReactFlow for complex causal models:

import {CauseEffectGraph} from '@components/CauseEffectGraph';

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

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

See Cause-Effect Diagrams for schema and examples.

Automation Tools

Unified CLI

All tools are accessible via the crux CLI:

npm run crux -- --help           # Show all domains
npm run crux -- validate         # Run all validators
npm run crux -- analyze          # Analysis and reporting
npm run crux -- fix              # Auto-fix common issues
npm run crux -- content          # Page management
npm run crux -- generate         # Content generation
npm run crux -- resources        # External resource management

Common Workflows

Task	Command
Validate before commit	`npm run precommit`
Full validation suite	`npm run validate`
Rebuild entity database	`npm run build:data`
Grade a specific page	`npm run crux -- content grade --page <id>`
Find unlinked mentions	`npm run crux -- analyze mentions`
Fix escaping issues	`npm run crux -- fix escaping`

Validation Rules

The validation suite includes 20+ rules:

Validator	What It Checks
`compile`	MDX syntax and compilation
`frontmatter-schema`	YAML frontmatter validity
`dollar-signs`	LaTeX escaping (`\$100` not `$100`)
`comparison-operators`	JSX escaping (`\<100ms` not `<100ms`)
`entitylink-ids`	All EntityLink references exist
`quality-source`	Quality set by pipeline, not manually
`mermaid`	Diagram syntax validation

For complete tool reference, see Automation Tools.

Technical Stack

Core Technologies

Layer	Technology
Framework	Astro 5 with Starlight theme
Components	React 19
Styling	Tailwind CSS 4
Type Safety	TypeScript + Zod schemas
Graphs	ReactFlow (XYFlow) + Dagre/ELK layout
Diagrams	Mermaid 11
Math	KaTeX
Data	YAML sources → JSON build artifacts

Project Structure

apps/longterm/
├── src/
│   ├── content/docs/           # ~550 MDX pages
│   │   ├── knowledge-base/     # Main content (risks, responses, orgs, people)
│   │   ├── ai-transition-model/# Comprehensive factor network
│   │   ├── project/            # Public project documentation
│   │   └── internal/           # Contributor docs and style guides
│   ├── components/
│   │   ├── wiki/               # 50+ content components
│   │   ├── CauseEffectGraph/   # Interactive graph system
│   │   └── ui/                 # shadcn components
│   ├── data/
│   │   ├── *.yaml              # Source data files
│   │   └── *.json              # Generated (build artifacts)
│   └── pages/                  # Astro dynamic routes
├── scripts/
│   ├── crux.mjs                # CLI entry point
│   ├── build-data.mjs          # Data compilation pipeline
│   ├── commands/               # CLI domain handlers
│   └── validate/               # 23 validators
└── astro.config.mjs            # Sidebar and site config

Key Configuration Files

File	Purpose
`astro.config.mjs`	Sidebar structure, Starlight setup
`src/content.config.ts`	MDX frontmatter schema
`src/data/schema.ts`	Entity type definitions (Zod)
`package.json`	Dependencies and npm scripts

Development Workflow

Getting Started

# Install dependencies
npm install

# Start development server (auto-runs build:data)
npm run dev

# Build for production
npm run build

After Editing Content

# After editing YAML data files
npm run build:data

# After editing any content
npm run precommit        # Quick validation
npm run validate         # Full validation

Creating New Content

New pages: Follow the appropriate style guide
New entities: Add to relevant YAML in src/data/, run npm run build:data
New components: Add to src/components/wiki/, use path aliases

For content generation workflows, see Research Reports.

Project & Strategy

Project Overview — What LongtermWiki is, core components
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 Guide — For risk and response pages
Model Style Guide — For analytical model pages
ATM Style Guide — For AI Transition Model pages
Stub Style Guide — For minimal reference pages

Technical Documentation

System Architecture — Data flow, pipelines, design decisions
Page Type System — Complete classification reference
Rating System — Quality scoring system
Content Database — SQLite caching, source fetching, AI summaries
Automation Tools — All scripts and workflows
Cause-Effect Diagrams — Graph schema and examples
Mermaid Diagrams — Diagram guidelines
Documentation Maintenance — Keeping docs updated