Entity Type Reference

Field-level reference for all data types defined in data/schema.ts. Required fields are marked with bold. Types with ? are optional.

Entity (Base Type)

All entities share these fields. Type-specific fields (documented below) are merged on top.

Field	Type	Required	Description
id	`string`	Yes	Slug identifier (e.g., `deceptive-alignment`)
type	`EntityType`	Yes	One of 24 canonical types + aliases
title	`string`	Yes	Display name
description	`string`	No	1-3 sentence summary
aliases	`string[]`	No	Alternative names for search
content	`EntityContent`	No	Rich YAML-first content (see below)
status	`EntityStatus`	No	`stub` \| `draft` \| `published` \| `verified`
lastUpdated	`string`	No	ISO date, `YYYY-MM` format
tags	`string[]`	No	Standardized tags for filtering
clusters	`Cluster[]`	No	Topic clusters: `ai-safety`, `biorisks`, `cyber`, `epistemics`, `governance`, `community`
severity	`Severity`	No	Risk severity: `low` through `catastrophic`
likelihood	`string \| StructuredLikelihood`	No	Risk likelihood (string or structured)
timeframe	`string \| StructuredTimeframe`	No	Risk timeframe (string or structured)
maturity	`ResearchMaturity`	No	`Neglected` \| `Emerging` \| `Growing` \| `Mature`
website	`string (url)`	No	External website
customFields	`CustomField[]`	No	Arbitrary key-value pairs for InfoBox
relatedTopics	`string[]`	No	Topic tag references
relatedEntries	`RelatedEntry[]`	No	Typed relationships to other entities
sources	`EntitySource[]`	No	Legacy inline citations
resources	`string[]`	No	Resource IDs from `data/resources/`
parameterDistinctions	`ParameterDistinctions`	No	For parameter entities: clarify vs. similar parameters

EntityContent

Structured content stored in YAML. Allows pages to be defined entirely in YAML without MDX files.

Field	Type	Required	Description
intro	`string`	No	Opening paragraph(s) before sections
sections	`ContentSection[]`	No	Main content sections
footer	`string`	No	Closing content after sections

ContentSection

A single section within entity content. Supports multiple content types per section.

Field	Type	Required	Description
heading	`string`	Yes	Section heading
body	`string`	No	Markdown prose content
mermaid	`string`	No	Mermaid diagram source code
table	`ContentTable`	No	Structured table: `{ headers, rows, caption? }`
component	`string`	No	Custom React component name (e.g., `PageCauseEffectGraph`)
componentProps	`Record`	No	Props to pass to the custom component

A typed reference from one entity to another.

Field	Type	Required	Description
id	`string`	Yes	Target entity slug
type	`EntityType`	Yes	Target entity type
relationship	`RelationshipType`	No	One of 45 relationship types
strength	`string`	No	`weak` \| `moderate` \| `strong`

StructuredLikelihood

Rich likelihood assessment for risk entities.

Field	Type	Required	Description
level	`LikelihoodLevel`	Yes	`low` \| `medium` \| `medium-high` \| `high` \| `very-high` \| `near-certain`
status	`LikelihoodStatus`	No	`theoretical` \| `emerging` \| `occurring` \| `established`
confidence	`Confidence`	No	`low` \| `medium` \| `high`
notes	`string`	No	Explanation

StructuredTimeframe

Rich timeframe estimate for risk entities.

Field	Type	Required	Description
median	`number`	Yes	Year number (e.g., 2030)
earliest	`number`	No	Earliest plausible year
latest	`number`	No	Latest plausible year
confidence	`Confidence`	No	`low` \| `medium` \| `high`
notes	`string`	No	Explanation

ParameterDistinctions

Clarifies how a parameter entity differs from similar parameters.

Field	Type	Required	Description
focus	`string`	Yes	Short phrase: what this parameter measures
summary	`string`	No	One sentence summary
distinctFrom	`ParameterDistinction[]`	Yes	Related but distinct parameters

Each ParameterDistinction:

Field	Type	Description
id	`string`	Target parameter ID
theirFocus	`string`	What they measure
relationship	`string`	How they relate

Expert (Person)

Defined in data/experts.yaml. Represents researchers, policymakers, and other notable people.

Field	Type	Required	Description
id	`string`	Yes	Slug (e.g., `paul-christiano`)
name	`string`	Yes	Full name
affiliation	`string`	No	Organization ID reference
role	`string`	No	Job title or role
website	`string (url)`	No	Personal website
twitter	`string`	No	Twitter handle
knownFor	`string[]`	No	Notable contributions
background	`string`	No	Brief bio
positions	`ExpertPosition[]`	No	Stated views on key topics

ExpertPosition

A recorded position or estimate from an expert.

Field	Type	Required	Description
topic	`string`	Yes	e.g., `p-doom`, `alignment-difficulty`, `timelines`
view	`string`	Yes	Their position in words
estimate	`string`	No	Quantified value (e.g., `10-20%`, `2030s`)
confidence	`Confidence`	No	How confident they are
date	`string`	No	When they stated this
source	`string`	No	Source title
sourceUrl	`string (url)`	No	Source link
notes	`string`	No	Additional context

Organization

Defined in data/organizations.yaml. Labs, research orgs, funders, government bodies.

Field	Type	Required	Description
id	`string`	Yes	Slug (e.g., `anthropic`)
name	`string`	Yes	Display name
type	`OrgType`	Yes	`frontier-lab` \| `safety-org` \| `academic` \| `government` \| `funder` \| `policy`
founded	`string`	No	Year founded
headquarters	`string`	No	Location
website	`string (url)`	No	Official website
description	`string`	No	Brief description
keyPeople	`string[]`	No	Expert ID references
funding	`string`	No	e.g., `$7B+`
employees	`string`	No	e.g., `≈1000`
safetyFocus	`string`	No	Description of safety work
parentOrg	`string`	No	Org ID for subsidiaries

Estimate

Defined in data/estimates.yaml. Aggregated expert estimates on key variables.

Field	Type	Required	Description
id	`string`	Yes	e.g., `p-tai-2030`
variable	`string`	Yes	What's being estimated
description	`string`	No	Context for the estimate
unit	`string`	No	e.g., `%`, `years`
aggregateRange	`string`	No	Summary range across sources
category	`string`	No	e.g., `timelines`, `risk`, `alignment`
estimates	`EstimateSource[]`	Yes	Individual source estimates
lastUpdated	`string`	No	ISO date

EstimateSource

Field	Type	Required	Description
source	`string`	Yes	Who made the estimate (expert ID or org name)
value	`string`	Yes	The estimate value
date	`string`	No	When stated
url	`string (url)`	No	Source link
notes	`string`	No	Context
confidence	`Confidence`	No	Confidence level

Crux (Key Uncertainty)

Defined in data/cruxes.yaml. Disagreements and uncertainties that drive different conclusions.

Field	Type	Required	Description
id	`string`	Yes	e.g., `alignment-difficulty`
question	`string`	Yes	The key question
domain	`string`	No	`Technical`, `Governance`, etc.
description	`string`	No	Context
importance	`Importance`	Yes	`low` \| `medium` \| `high` \| `critical`
resolvability	`string`	No	`soon` \| `years` \| `decades` \| `never`
currentState	`string`	No	Current state of evidence
positions	`CruxPosition[]`	Yes	Different views on this crux
wouldUpdateOn	`string[]`	No	Evidence that would change positions
relatedCruxes	`string[]`	No	Crux ID references
relevantResearch	`{ title, url? }[]`	No	Key papers

CruxPosition

Field	Type	Required	Description
view	`string`	Yes	The position
probability	`string`	No	e.g., `40-60%`
holders	`string[]`	No	Expert IDs or org names
implications	`string`	No	What follows from this view

Risk

Defined in data/ entity files. Failure modes and hazards.

Field	Type	Required	Description
id	`string`	Yes	e.g., `deceptive-alignment`
name	`string`	Yes	Display name
category	`RiskCategory`	Yes	`accident` \| `misuse` \| `structural` \| `epistemic`
description	`string`	No	What the risk is
severity	`Importance`	No	How bad if it happens
likelihood	`string`	No	How likely
mechanisms	`string[]`	No	How it could happen
prerequisites	`string[]`	No	Risk IDs that must occur first
mitigatedBy	`string[]`	No	Intervention IDs
relatedRisks	`string[]`	No	Related risk IDs
keyPapers	`string[]`	No	Source IDs

Intervention

Responses to risks: technical, governance, or institutional.

Field	Type	Required	Description
id	`string`	Yes	e.g., `interpretability-research`
name	`string`	Yes	Display name
category	`string`	Yes	`technical` \| `governance` \| `institutional` \| `field-building` \| `resilience`
subcategory	`string`	No	More specific category
description	`string`	No	What this intervention does
addressesRisks	`string[]`	No	Risk IDs this addresses
organizations	`string[]`	No	Org IDs working on this
tractability	`Importance`	No	How tractable
neglectedness	`Importance`	No	How neglected
importance	`Importance`	No	How important
timeHorizon	`string`	No	e.g., `1-3 years`, `5+ years`

Resource

Defined in data/resources/*.yaml. External references with AI-generated summaries.

Field	Type	Required	Description
id	`string`	Yes	Human-readable slug or hash
url	`string (url)`	Yes	Source URL
title	`string`	Yes	Display title
type	`ResourceType`	Yes	`paper` \| `blog` \| `report` \| `book` \| `talk` \| `podcast` \| `government` \| `reference` \| `web`
authors	`string[]`	No	Author names
published_date	`string`	No	`YYYY` or `YYYY-MM-DD`
local_filename	`string`	No	Path in `.cache/sources/`
importance	`number (0-100)`	No	Determines review depth
abstract	`string`	No	Original abstract
summary	`string`	No	AI-generated 1-2 sentence summary
review	`string`	No	AI-generated 1-4 paragraph review
key_points	`string[]`	No	AI-generated key points
cited_by	`string[]`	No	Entity IDs that cite this
fetched_at	`string`	No	When content was fetched
publication_id	`string`	No	Reference to `publications.yaml`
tags	`string[]`	No	Topic tags
credibility_override	`number (1-5)`	No	Override publication default

Publication

Defined in data/publications.yaml. Maps domains to credibility ratings.

Field	Type	Required	Description
id	`string`	Yes	e.g., `nature`, `arxiv`
domains	`string[]`	Yes	Domains that map to this publication
name	`string`	Yes	Display name
type	`PublicationType`	Yes	See enum table below
credibility	`number (1-5)`	Yes	1 = unvetted, 5 = gold standard
peer_reviewed	`boolean`	No	Whether peer-reviewed
description	`string`	No	Description

GlossaryTerm

Defined in data/glossary.yaml. Definitions of key terms.

Field	Type	Required	Description
id	`string`	Yes	e.g., `agi`
term	`string`	Yes	Display name
definition	`string`	Yes	The definition
aliases	`string[]`	No	Other names
related	`string[]`	No	Other term IDs
seeAlso	`string[]`	No	Page slugs

TimelineEvent

Defined in data/timeline.yaml. Key events in AI safety history.

Field	Type	Required	Description
id	`string`	Yes	Event identifier
date	`string`	Yes	ISO date, `YYYY`, or `YYYY-MM`
title	`string`	Yes	Event title
description	`string`	No	What happened
category	`string`	No	`capability` \| `safety` \| `governance` \| `incident` \| `organization`
importance	`Importance`	No	How significant
actors	`string[]`	No	Expert or org IDs
url	`string (url)`	No	Reference link

Source (Publication/Paper)

Legacy citation format (being migrated to Resources).

Field	Type	Required	Description
id	`string`	Yes	e.g., `superintelligence-2014`
title	`string`	Yes	Paper/book title
authors	`string[]`	No	Expert IDs or names
year	`string`	No	Publication year
type	`string`	No	`book` \| `paper` \| `blog` \| `report` \| `talk` \| `podcast`
url	`string (url)`	No	Link
abstract	`string`	No	Summary
significance	`string`	No	Why it matters

Fact

Defined in data/facts/*.yaml. Canonical referenceable values per entity.

Field	Type	Required	Description
value	`string`	No	Display value (auto-generated for computed facts)
numeric	`number`	No	Machine-readable numeric value
asOf	`string`	No	Date the value was current
source	`string`	No	Citation
note	`string`	No	Additional context
noCompute	`boolean`	No	If true, numeric value cannot be referenced in compute expressions
compute	`string`	No	Expression: `{entity.factId} * {other.factId}`
format	`string`	No	Display format: `$%.1f billion`
formatDivisor	`number`	No	Divide result before formatting (e.g., `1e9` for billions)

Database (Top-level)

The complete database schema aggregating all types. Generated as database.json at build time.

{
  experts: Expert[];
  organizations: Organization[];
  estimates: Estimate[];
  cruxes: Crux[];
  risks: Risk[];
  interventions: Intervention[];
  glossary: GlossaryTerm[];
  timeline: TimelineEvent[];
  sources: Source[];
  graphs: Graph[];
  entities: Entity[];
  resources: Resource[];
  publications: Publication[];
}

Plus build-time additions:

typedEntities[] — entities with resolved types and numeric IDs
backlinks — reverse reference index
tagIndex — tag → entity mapping
pathRegistry — entity ID → URL path mapping
facts — resolved canonical facts
insights[] — extracted cross-page insights
pages[] — page-level metadata (quality scores, importance, last updated)