Content Formats

The knowledge base uses several distinct content formats, each optimized for different purposes. Understanding when to use each format helps maintain consistency and makes content more useful.

Format Overview

Format	Purpose	Complexity	Example
Wiki Page	Explain concepts, risks, responses	Low-Medium	Bioweapons
Research Report	Deep-dive investigation	Medium	AI Talent Concentration
Cascade Model	Detailed analytical scenarios	High	Cyber Psychosis Cascade
Squiggle Model	Quantitative forecasts	High	Compute Forecast Sketch
Cause-Effect Diagram	Visual causal structure	Medium	Diagrams Index
Estimate Table	Aggregate parameter view	Low	Parameter Table

Choosing the Right Format

If you need…	Use…
Quick reference on a topic	Wiki Page
Comprehensive investigation	Research Report
Visual causal structure	Cause-Effect Diagram
Detailed scenario analysis	Cascade Model
Quantitative forecasts	Squiggle Model
Cross-factor comparison	Estimate Table

1. Wiki Pages

Low-Medium Complexity Most Common

Purpose: The core content unit for explaining concepts, risks, and responses.

When to Use:

Explaining a risk, response, or concept
Providing reference material on a topic
Creating overview pages for categories

Quality Levels: Q1-2 (stub) → Q3 (solid, 5-10 citations) → Q4 (comprehensive, 15+ citations) → Q5 (authoritative, 25+ citations, diagrams)

Reference: Knowledge Base Style Guide

View Structure Templates

Risk Page
Response Page

---
title: "Risk Name"
description: "One-sentence summary with specifics"
quality: 3
importance: 75
---

## Overview
[2-3 paragraphs]

### Risk Assessment
| Dimension | Assessment | Notes |
|-----------|------------|-------|
| Severity | High | ... |
| Likelihood | 20-40% | ... |

### Responses That Address This Risk
| Response | Mechanism | Effectiveness |
|----------|-----------|---------------|
| [Response](/link) | How it helps | Medium |

## Why This Matters
[Detailed explanation]

## Key Uncertainties
[What would change the assessment]

---
title: "Response Name"
description: "One-sentence summary"
quality: 3
importance: 70
---

## Overview
[2-3 paragraphs]

### Quick Assessment
| Dimension | Assessment | Notes |
|-----------|------------|-------|
| Tractability | Medium | ... |
| If alignment hard | High value | ... |
| If alignment easy | Lower value | ... |

### Risks Addressed
| Risk | Mechanism | Effectiveness |
|------|-----------|---------------|
| [Risk](/link) | How it helps | High |

## How It Works
[Detailed explanation]

## Critical Assessment
### Limitations
### Key Uncertainties

2. Research Reports

Medium Complexity 1-5 hours

Purpose: Deep-dive investigations that identify causal factors and inform diagram creation.

When to Use:

Need comprehensive understanding before building models
Investigating specific questions with web research
Preparing groundwork for cause-effect diagrams

Depth Levels: quick (15-30 min, 5-10 sources) → standard (1-2 hours, 15-25 sources) → comprehensive (3-5 hours, 30-50+ sources)

Reference: Research Report Style Guide

View Structure Template

---
title: "Topic: Research Report"
description: "Key finding with specific data (escape \\$ signs)"
topic: "entity-id"
createdAt: 2025-01-07
lastUpdated: 2025-01-07
researchDepth: "standard"  # quick | standard | comprehensive
sources: ["web", "codebase"]
quality: 3
---

## Executive Summary
| Finding | Key Data | Implication |
|---------|----------|-------------|
| **US dominance** | 57% share | ... |

## Background
[Context and safety implications]

## Key Findings
### Theme 1
[Analysis with tables and citations]

## Causal Factors
### Primary Factors (Strong Influence)
| Factor | Direction | Type | Evidence | Confidence |
|--------|-----------|------|----------|------------|
| **Factor A** | ↑ Topic | leaf | Evidence | High |

### Secondary Factors (Medium Influence)
| Factor | Direction | Type | Evidence | Confidence |
|--------|-----------|------|----------|------------|
| **Factor B** | ↓ Topic | cause | Evidence | Medium |

## Open Questions
| Question | Why It Matters | Current State |
|----------|----------------|---------------|
| **Question 1** | Impact | Status |

## Sources
[Organized by type]

Causal Factors → Diagrams Mapping:

Report Element	Diagram Element
Factor name	Node label
Direction (↑/↓)	Edge effect (increases/decreases)
Type (leaf/cause)	Node type
Section (Primary/Secondary)	Edge strength

3. Cascade Models

High Complexity 1000+ lines

Purpose: Detailed analytical models for specific dynamics, scenarios, or risk pathways.

When to Use:

Modeling multi-stage processes with cascading effects
Analyzing specific scenarios with probability estimates
Creating detailed intervention analysis

Characteristics: 1000+ lines, multiple Mermaid diagrams, probability estimates, intervention tables

Example: Cyber Psychosis Cascade

View Structure & Ratings

Required Sections:

Overview (2-3 paragraphs)
Conceptual Framework (cascade stages table + Mermaid diagram)
Mechanisms of Harm (tables: Vector | Mechanism | Vulnerable Population | Severity)
Cascade Pathways (tables: Pathway | Trigger | Progression | Terminal State)
Population Vulnerability Model (factor estimates with correlations)
Specific Scenario Analysis (attack components, timeline, intervention windows)
Scenario Probability Analysis (probability-weighted outcomes table)
Intervention Analysis (tables by level: Individual, Platform, Institutional)
Limitations (key caveats and model boundaries)

Model Ratings (1-5 scale):

Dimension	Description
Novelty	How surprising/original
Rigor	Quality of reasoning and evidence
Actionability	Usefulness for decisions
Completeness	Coverage of relevant factors

Frontmatter:

---
title: "X Cascade Model"
description: "This model analyzes [mechanism]. It identifies [key finding]."
pageTemplate: knowledge-base-model
ratings:
  novelty: 4
  rigor: 3
  actionability: 3
  completeness: 4
---

4. Squiggle Models

High Complexity Requires Squiggle

Purpose: Quantitative probabilistic forecasts with explicit uncertainty.

When to Use:

Making specific numerical predictions
Modeling parameter dependencies quantitatively
When uncertainty ranges matter

Output: Probability distributions, not point estimates

Example: Compute Forecast Sketch

View Example & Comparison

Example Squiggle Code:

// === INPUT PARAMETERS ===

// Base capacity (units/year)
baseCapacity = 50 to 100

// Growth rate (% per year)
growthRate = normal(0.15, 0.05)

// === DERIVED QUANTITIES ===

// Projected capacity in year N
projectedCapacity(year) = baseCapacity * (1 + growthRate)^year

// === SCENARIOS ===

// Optimistic scenario
optimistic = {
  growth: growthRate * 1.3,
  outcome: projectedCapacity(5)
}

// Pessimistic scenario
pessimistic = {
  growth: growthRate * 0.7,
  outcome: projectedCapacity(5)
}

When to Use Squiggle vs. Prose Estimates:

Use Squiggle When	Use Prose When
Need explicit probability distributions	Point estimates suffice
Multiple dependent parameters	Single independent estimate
Want to show uncertainty propagation	Uncertainty is qualitative
Building on other quantitative models	Descriptive analysis

5. Cause-Effect Diagrams

Medium Complexity 10-20 nodes

Purpose: Visual representation of causal relationships between factors.

When to Use:

Showing what drives a particular factor
Visualizing relationships in the AI Transition Model
Summarizing research report findings

Size: 10-20 nodes optimal, max 30 edges

Reference: Cause-Effect Diagram Style Guide

View Schema & Visual Encoding

Node Type Hierarchy:

Layer 1: leaf nodes (root causes, external factors)
    ↓
Layer 2: cause nodes (derived from leaves)
    ↓
Layer 3: intermediate nodes (direct factors)
    ↓
Layer 4: effect nodes (target outcomes)

Visual Encoding:

Element	Meaning
Node color	Type (teal=leaf, gray=cause/intermediate, amber=effect)
Edge thickness	Strength (thin=weak, thick=strong)
Edge color	Direction (blue=increases, red=decreases, gray=mixed)
Edge style	Confidence (dashed=low, solid=high)

YAML Schema:

causeEffectGraph:
  title: "What Drives X?"
  description: "Causal factors affecting X."
  primaryNodeId: main-effect
  nodes:
    - id: upstream-factor
      label: Upstream Factor
      type: leaf
      description: Hover text explanation.
    - id: main-effect
      label: Main Effect
      type: effect
      description: The outcome being modeled.
  edges:
    - source: upstream-factor
      target: main-effect
      strength: strong
      effect: increases

6. Estimate Tables

Low Complexity Auto-generated

Purpose: Aggregate view of parameter estimates across the AI Transition Model.

When to Use:

Comparing estimates across factors
Identifying high-uncertainty or high-impact parameters
Prioritizing research or intervention targets

Location: Parameter Table

View Data Source & Columns

Data Source (entity YAML):

- id: tmc-compute
  type: ai-transition-model-subitem
  ratings:
    changeability: 4
    xriskImpact: 8
    uncertainty: 6
    evidenceQuality: 7

Table Columns:

Column	Source	Description
Parameter	Entity title	The factor being estimated
Changeability	`ratings.changeability`	How tractable/malleable (1-10)
X-Risk Impact	`ratings.xriskImpact`	Effect on existential risk (1-10)
Uncertainty	`ratings.uncertainty`	How well understood (1-10)
Evidence	`ratings.evidenceQuality`	Quality of supporting evidence (1-10)

Table Features:

Sorting by any column
Filtering by category
Identifying intervention priorities (high changeability + high impact)
Finding research gaps (high uncertainty + high impact)

Format Relationships

View Relationship Diagram

                    ┌─────────────────┐
                    │ Research Report │
                    │ (Investigation) │
                    └────────┬────────┘
                             │ informs
            ┌────────────────┼────────────────┐
            ↓                ↓                ↓
┌───────────────────┐ ┌─────────────┐ ┌───────────────┐
│  Cascade Model    │ │ Cause-Effect│ │ Squiggle Model│
│  (Detailed)       │ │   Diagram   │ │ (Quantitative)│
└─────────┬─────────┘ └──────┬──────┘ └───────┬───────┘
          │                  │                │
          └──────────────────┼────────────────┘
                             │ summarized in
                             ↓
                    ┌─────────────────┐
                    │   Wiki Page     │
                    │  (Reference)    │
                    └────────┬────────┘
                             │ aggregated in
                             ↓
                    ┌─────────────────┐
                    │ Estimate Table  │
                    │  (Overview)     │
                    └─────────────────┘

Typical Workflows:

Starting Point	Workflow
New topic	Research Report → Wiki Page + Diagram
Quantitative question	Squiggle Model → Diagram → Wiki Page
Complex dynamics	Cascade Model → Wiki Page + Diagram
Quick reference	Wiki Page (standalone)

Content Formats

Format Overview

Choosing the Right Format

1. Wiki Pages

2. Research Reports

3. Cascade Models

4. Squiggle Models

5. Cause-Effect Diagrams

6. Estimate Tables

Format Relationships

See Also