Longterm Wiki
Updated 2026-03-13HistoryData
Page StatusDocumentation
Edited today846 words1 backlinksUpdated quarterlyDue in 13 weeks
20QualityDraft •12.5ImportancePeripheral10.5ResearchMinimal
Summary

Internal style guide for deprecated research report format, specifying table-based layouts, escaped dollar signs, YAML schema, and causal factor documentation for diagram creation. Provides comprehensive formatting rules and validation checklists but addresses an obsolete content structure.

Content3/13
LLM summaryScheduleEntityEdit historyOverview
Tables14/ ~3Diagrams0Int. links2/ ~7Ext. links0/ ~4Footnotes0/ ~3References0/ ~3Quotes0Accuracy0RatingsN:0.5 R:4 A:6 C:7Backlinks1
Issues1
QualityRated 20 but structure suggests 60 (underrated by 40 points)

Research Report Style Guide

Deprecated

This style guide describes the deprecated research-reports format. Research content now lives in the relevant knowledge-base sections. For creating and improving wiki pages, use the Crux content pipeline: pnpm crux content create "Title" --tier=standard and pnpm crux content improve <id> --tier=standard --apply. See the Knowledge Base Style Guide for current formatting guidelines.

Purpose

Research reports are deep-dive investigations into specific topics within the knowledge base. They serve two primary functions:

  1. Comprehensive analysis of a topic with citations from authoritative sources
  2. Input for diagram creation - reports inform the causal models in cause-effect diagrams

Reports are generated with AI assistance (Claude Code) using web research and synthesis of existing content.


Formatting Guidelines

Key Formatting Rules

Follow the Knowledge Base Style Guide for general principles. These additional rules are specific to research reports.

Use Tables Over Bullet Lists

Bad (low density):

**Factor Name**
- Direction: Increases → topic
- Type: leaf
- Evidence: Some evidence
- Confidence: high

Good (table format):

| Factor | Direction | Type | Evidence | Confidence |
|--------|-----------|------|----------|------------|
| **Factor Name** | ↑ Topic | leaf | Some evidence | High |

Use Callouts/Asides Liberally

Add <Aside> components to highlight key insights, caveats, and implications. Aim for at least 3-4 callouts per report.

<Aside type="tip" title="Why This Matters">
Key insight about safety implications.
</Aside>

<Aside type="caution" title="Limitation">
Important caveat about the data.
</Aside>

<Aside type="note" title="Data Source">
Methodological note about how data was collected.
</Aside>

Escape Dollar Signs

Currency values MUST be escaped to avoid LaTeX rendering:

WrongRightRenders As
$100,000\$100,000$100,000
$5B\$5B$5B

Also escape in frontmatter description:

description: "H-1B fees increased to \\$100K"

Use Horizontal Rules

Separate major sections with --- for visual clarity.


Report Structure

Every research report should follow this structure:

---
title: "Topic Name: Research Report"
description: "One-sentence summary with specific data (escape \\$ signs)."
topic: "entity-id"           # Links to knowledge-base or transition-model item
createdAt: 2025-01-07        # Date without quotes (YAML date type)
lastEdited: "2025-01-07"     # Must be a quoted string (unlike createdAt)
researchDepth: "standard"    # quick | standard | comprehensive
sources: ["web", "codebase"] # What sources were consulted
quality: 3                   # 1-5 quality rating
---

Required Sections

SectionPurposeFormat
Executive SummaryKey findings with dataTable (not bullets)
BackgroundContext and why this mattersProse + Aside
Key FindingsMain research resultsProse + tables + Asides
Causal FactorsFactors for diagram creationTables by strength
Open QuestionsKey uncertaintiesTable
SourcesOrganized citationsLists by type

Section Templates

Executive Summary (Table Format)

Use a table, not bullets:

## Executive Summary

| Finding | Key Data | Implication |
|---------|----------|-------------|
| **US dominance** | 57% of top researchers | US remains primary hub |
| **Policy risk** | \$100K H-1B fee | May accelerate brain drain |
| **Key uncertainty** | Immigration policy | Critical inflection point |

Causal Factors (Table Format)

Critical for Diagrams

This section directly feeds into cause-effect diagram creation. Use tables to ensure factors map cleanly to diagram nodes and edges.

## Causal Factors

### Primary Factors (Strong Influence)

| Factor | Direction | Type | Evidence | Confidence |
|--------|-----------|------|----------|------------|
| **Immigration Policy** | ↑↓ Concentration | leaf | 80% retention depends on visas | High |
| **Research Ecosystem** | ↑ Concentration | cause | 60% of top institutions in US | High |

### Secondary Factors (Medium Influence)

| Factor | Direction | Type | Evidence | Confidence |
|--------|-----------|------|----------|------------|
| **Graduate Programs** | ↑ Concentration | cause | 80% of students stay | Medium |

### Minor Factors (Weak Influence)

| Factor | Direction | Type | Evidence | Confidence |
|--------|-----------|------|----------|------------|
| **Nationalist Sentiment** | ↓ Mobility | leaf | Limited evidence | Low |

Column definitions:

  • Factor: Short name (2-5 words)
  • Direction: ↑ (increases), ↓ (decreases), ↑↓ (bidirectional), Mixed
  • Type: leaf (external), cause (derived), intermediate (direct factor)
  • Evidence: Brief summary with source reference
  • Confidence: High, Medium, or Low

Open Questions (Table Format)

## Open Questions

<Aside type="note" title="Key Uncertainties">
These questions represent the highest-value areas for follow-up research.
</Aside>

| Question | Why It Matters | Current State |
|----------|----------------|---------------|
| **Will H-1B fees persist?** | \$100K changes economics | May face legal challenge |
| **Is China's surge sustainable?** | 47% share may be temporary | Depends on retention |

Research Methodology

Source Hierarchy

PrioritySource TypeExamplesTrust Level
1Peer-reviewed researchNature, Science, arXivHigh
2Expert organizationsRAND, Brookings, FHI, CAISHigh
3Government reportsGAO, UK AISI, NISTHigh
4Quality journalismReuters, AP, major newspapersMedium
5Industry analysisMcKinsey, Gartner, OECDMedium
6Blog posts/forumsLessWrong, EA ForumVariable
7Company announcementsOpenAI, Anthropic blogsMedium (verify)

Research Depth Levels

LevelTimeSourcesUse When
quick15-30 min5-10Initial exploration
standard1-2 hours15-25Most topics
comprehensive3-5 hours30-50+Critical topics

Validation Checklist

Before marking a report complete, verify:

CheckWhyHow to Fix
Dollar signs escapedPrevents LaTeX renderingReplace $ with \$
Dates unquotedYAML schema requires date typecreatedAt: 2025-01-07 not "2025-01-07"
Tables usedCausal factors must be in tablesConvert bullet lists to tables
Asides includedHighlight key insightsAdd 3-4 callouts minimum
Sources organizedEasier to assess qualityGroup by type
Links validBroken links hurt credibilityTest all URLs

Run validators:

npm run crux -- validate compile
npm run crux -- validate unified --rules=dollar-signs

Quality Criteria

LevelCriteria
1Basic outline, fewer than 10 sources, major gaps
2Main points covered, 10-15 sources, some gaps
3Solid coverage, 15-25 sources, minor gaps
4Comprehensive, 25-35 sources, well-structured
5Authoritative, 35+ sources, original synthesis

Integration with Diagrams

After a report is complete, the Causal Factors section can be directly converted to a causeEffectGraph:

Report ColumnMaps to Diagram
FactorNode label
TypeNode type (leaf, cause, intermediate)
Direction (↑/↓)Edge effect (increases, decreases)
Strength (section)Edge strength (strong, medium, weak)
ConfidenceEdge confidence

Workflow:

  1. Create report → Research and document findings
  2. Extract factors → Format in tables
  3. Create diagram → Use cause-effect-diagram skill
  4. Link both → Report references diagram

File Location & Naming

Note: The research-reports directory has been removed. Research content should now be integrated into the appropriate knowledge-base sections (risks/, responses/, etc.).

Naming: Use kebab-case matching the topic entity ID:

  • ai-talent-concentration.mdx
  • algorithmic-progress.mdx
  • compute-governance.mdx

Common Mistakes

MistakeProblemFix
Bullet lists for factorsLow density, hard to scanUse tables
Unescaped $Renders as LaTeXUse \$
Quoted datesSchema validation failsRemove quotes
No AsidesMisses key insightsAdd 3-4 callouts
Flat hierarchyHard to navigateUse h2/h3 structure
Missing validationErrors in productionRun npm run validate