A brain isn't a folder of files. It's closer to a research desk: documents
on top, marginal notes pulled from them, a stack of people the documents
keep mentioning, and a legal pad on the side where the dates of every
development get jotted down. The workspace surfaces each of these as its
own section so you can read the brain the way you'd actually look up
something in it, not the way it happens to be stored.

## The four views

**Lately** is the landing page. It answers "what did this brain learn
recently." Every statement the brain extracted from a document shows up
here, newest first, grouped by day. Read top-down, you get a running
record of the brain's recent intake.

**People** is the cast of characters. Every person the brain has noticed
across all its documents, with the most notable thing on record about
them. Read this like a publication's contributor index.

**Memories** are the source documents themselves. The agendas, the
transcripts, the articles, the notes. Anything that got ingested as a
page lives here. If you want to see the original prose a fact came from,
this is where you go.

**Timeline** is everything on record in chronological order, by year.
The single vertical rule with dots that runs down the page is exactly
what it looks like: a timeline. Click an entity name on any fact and
Timeline narrows to "everything on record about this entity."

## What's a fact

A **fact** is a single statement the brain pulled out of a document. They
look like one-line summaries because they are: an LLM reads each
document as it gets ingested, identifies the substantive claims, and
writes one short sentence per claim. Each fact carries its source, the
date it refers to, and the entity it's about.

Facts are not a destination. They're never shown as a bare list. You see
them on Lately, on a person's lead-quote, on Timeline, inline on the
Record they came from.

The brain extracts facts only from documents with substantive prose.
Short stubs, concept pages, and very brief notes produce nothing.
Meeting transcripts, briefings, and articles produce the most.

## How the brain marks what matters

Every fact carries one of three notability marks. The brain assigns
these as it extracts, based on how load-bearing the statement is to the
document it came from.

**Central** facts are the load-bearing ones. The vote totals, the
structural deficit, the major policy decisions. These are what someone
researching this brain would put in their notes.

**Noted** facts are the default. Things worth recording but not headline
material. Most facts land here.

**Passing** facts are aside-level. Side comments during public comment,
introductions, procedural mentions. Still on record so you can audit
the brain's intake, but you'd skip them when skimming.

In Lately, the column on the left of each fact carries this label as a
word. Click it to scope the feed.

## The five kinds of fact

Alongside notability, each fact is one of five kinds.

**Event** is something that happened on a specific date. A vote, a
meeting decision, a public statement. Most facts in a meeting-heavy
brain are events.

**Fact** is a standing claim about the world. "The city faces a $35
million deficit." Not bound to a specific moment.

**Commitment** is a promise or obligation. "The City Manager committed
to modeling a 5% reduction scenario."

**Belief** is what someone holds to be true. "Councilmember X believes
the proposal is fiscally irresponsible."

**Preference** is what someone wants. "Mayor Y prefers a phased rollout."

The kind appears in the Lately label column when it carries signal that
notability alone doesn't already convey. A central event is rendered as
"central"; a noted commitment is rendered as "commitment."

## What an entity is

An **entity** is a thing the brain has noticed across multiple facts.
People, organizations, places, concepts. Each entity has a slug:
`people-matt-mahan`, `city-of-san-jose`, `organizations-tenants-union`.
The slug is what links the entity across all the facts about it.

Click an entity anywhere in the workspace and you land on Timeline
scoped to that entity. That's the closest thing to a profile page until
each entity gets its own surface.

## What you don't see (and why)

Some things the brain stores aren't surfaced as their own destinations,
on purpose:

- **Pages by directory** — the storage layout (`sjc/2025-05-14-7102`,
  `wiki/people/matt-mahan`) is how rows are organized in the database.
  It's not how anyone thinks about the brain. Records, People,
  Timeline, and Lately are the views; the directory is incidental.
- **Tags and frontmatter raw** — both are visible on a Record's detail
  page, where they belong. They're not a top-level filter axis because
  in practice they're a derivative of type + entity.
- **The graph as a graph** — facts link to entities, entities link to
  documents, documents link to documents. The graph exists, but reading
  one is editorial work, not a node-and-edge diagram. The graph
  shows itself in how the People page leads with each person's most
  notable fact, and in how Timeline narrows on click.