# Page Anatomy — Compiled Truth vs. Timeline > Every page in the brain uses a strict two-layer structure: an "above the line" compiled-truth section (always accurate, rewritten as evidence changes) and a "below the line" append-only timeline (immutable log). Understanding this split determines when to rewrite vs. append, why the timeline is never edited, and how an AI agent should read vs. write each layer. Also covers frontmatter requirements, page types from gbrain.yml, and file-naming conventions. - Repository: garrytan/anara-brain - GitHub: https://github.com/garrytan/anara-brain - Human wiki: https://grok-wiki.com/public/wiki/garrytan-anara-brain-352efa33f049 - Complete Markdown: https://grok-wiki.com/public/wiki/garrytan-anara-brain-352efa33f049/llms-full.txt ## Source Files - `gbrain.yml` - `README.md` - `clients/template.md` - `clients/README.md` - `people/joseph-gabriel.md` ---
Relevant source files The following files were used as context for generating this wiki page: - [gbrain.yml](gbrain.yml) - [README.md](README.md) - [schema.md](schema.md) - [clients/template.md](clients/template.md) - [clients/README.md](clients/README.md) - [people/joseph-gabriel.md](people/joseph-gabriel.md) - [concepts/progress-is-positivity.md](concepts/progress-is-positivity.md) - [RESOLVER.md](RESOLVER.md)
# Page Anatomy — Compiled Truth vs. Timeline Every page in the Anara Brain follows a strict two-layer structure: an **above-the-line compiled-truth section** that always reflects the current best understanding, and a **below-the-line timeline** that is an immutable, append-only log of evidence. This split is the central design invariant of the GBrain architecture — it determines how an AI agent reads versus writes, and why the two layers must never be confused. Understanding this anatomy matters because it determines the difference between knowledge that compounds and knowledge that decays. The compiled-truth section is a living synthesis that improves with every new session. The timeline is the evidentiary record that makes that synthesis trustworthy and auditable. Neither layer can do its job without the other. --- ## The Two Layers ```text ┌──────────────────────────────────────────────────────────┐ │ COMPILED TRUTH (above the line) │ │ Always current. Rewritten as evidence changes. │ │ What an AI agent reads first. │ │ Answers: Who/what is this? Current state? Key facts? │ ├──────────────────────────────────────────────────────────┤ │ --- │ │ ## Timeline ← the dividing marker │ ├──────────────────────────────────────────────────────────┤ │ TIMELINE (below the line) │ │ Append-only. Reverse-chronological. Never edited. │ │ Raw observations, session notes, quotes, timestamps. │ │ New entries go at the top. │ └──────────────────────────────────────────────────────────┘ ``` Sources: [README.md:19-37]() --- ## Layer 1: Compiled Truth (Above the Line) ### What It Contains The compiled-truth section is the current, always-accurate summary of everything known about the subject. It is not a log — it is a synthesis. It answers the questions an AI agent or human needs to answer *right now* without having to read the entire timeline. The specific sections vary by page type (see [Page Types](#page-types) below), but the common thread is answering: - Who or what is this? - What is the current state? - What are the key facts, frameworks, or takeaways? For a **client page**, this means current programming, goals, key observations, injuries, and personal context — everything a trainer needs to walk into a session prepared. For a **person page** (`people/joseph-gabriel.md`), this means current role, what the person believes, what they're building, and their assessment — structured as prose sections like `## State`, `## What He Believes`, `## What He's Building`. For a **concept page** (`concepts/progress-is-positivity.md`), this means the core idea, evidence from Joseph's own words, how it shows up in practice, its shadow (failure modes), and connections to other concepts. ### When to Rewrite It The compiled-truth section is rewritten **whenever the picture changes** — new evidence contradicts an existing claim, a client's programming shifts, a concept evolves, or a relationship enters a new phase. There is no versioning of the compiled-truth section; the old version is replaced. This is intentional: the compiled truth is meant to be *trusted at face value* without needing to mentally reconcile it against stale prior states. > "This section gets **rewritten** whenever the picture changes. It's always current." Sources: [clients/README.md:30-33]() ### AI Agent Read Behavior When an AI agent queries the brain for current state, it reads the compiled-truth section and *stops there* unless it needs to trace reasoning back to source evidence. The compiled truth is the fast path. The timeline is the audit trail. --- ## Layer 2: The Timeline (Below the Line) ### What It Contains The timeline is a **reverse-chronological, append-only log** of raw evidence. Session notes, dated observations, quotes, events, and updates all live here. Each entry is stamped with a date and — for client pages — a session number. The dividing marker between the two layers is a horizontal rule (`---`) followed by `## Timeline` (or `## Session Timeline` for client-note pages). Everything below this marker is immutable after it is written. ```markdown --- ## Session Timeline ### 2026-05-21 — Session #42 **Focus:** Strength + breathwork **Energy:** 7/10 - Did: Deadlifts 3x5 @ 225, KB swings 3x15, 10min 2x breath meditation - Noticed: Right hip tighter than last week. Mentioned sleeping poorly. - Said: "Work has been intense this week. Big board meeting Thursday." - Next: Address hip tightness with corrective protocol. Check in on sleep. ``` Sources: [clients/README.md:42-54]() ### Why the Timeline Is Never Edited The timeline's immutability is what makes the compiled-truth section trustworthy. If past entries could be revised, there would be no way to distinguish between "the picture changed" (legitimate compiled-truth rewrite) and "the original observation was altered" (loss of evidentiary integrity). The timeline is the ground truth that the compiled truth is synthesized from; editing it would corrupt the chain of evidence. Concretely: if a client's goals shift, the **compiled truth is rewritten** to reflect the new goals, and a **new timeline entry** records when and why the shift happened. The old goals disappear from the compiled truth but remain implicit in earlier timeline entries. ### New Entries Go at the Top Entries are reverse-chronological: newest first. This means the most recent evidence is always immediately visible without scrolling. An AI agent scanning the timeline for context on recent sessions reads the top entries first. Sources: [clients/README.md:41](), [clients/template.md:48]() --- ## Frontmatter Requirements Every page — regardless of type — must include four required frontmatter fields, as declared in `gbrain.yml`: ```yaml type: # one of 12 valid types title: "" # human-readable title created: YYYY-MM-DD # creation date tags: # at least one tag - tag-name ``` Sources: [gbrain.yml:18-24]() Additional fields are type-specific. For example, `client-note` pages add `client`, `updated`, `session_count`, `cadence`, and `focus_areas`. Person pages add `role`, `email`, `gender`, `aliases`, `company`, and `location`. See `schema.md` for the full template per type. ### Frontmatter Does Not Replace Compiled Truth Frontmatter is machine-readable metadata for indexing, querying, and routing. It is not the compiled truth. The compiled truth lives in the page body — it is the prose sections that a human or AI reads to understand the subject. --- ## Page Types Twelve page types are defined in `gbrain.yml`. Each type corresponds to a directory and a frontmatter template in `schema.md`: | `type` value | Directory | What it represents | |---|---|---| | `person` | `people/` | Significant relationships — clients, collaborators, mentors | | `methodology` | `methodology/` | Training protocols, breathwork techniques, frameworks | | `concept` | `concepts/` | Philosophical ideas and recurring worldview building blocks | | `content` | `content/` | Published or planned content pieces | | `client-note` | `clients/` | Client-specific session tracking and progress | | `reading` | `reading/` | Books, articles, and intellectual influences | | `event` | *(any)* | Workshops, talks, community events | | `project` | `projects/` | Business initiatives and active work | | `analysis` | *(any)* | Research and analytical documents | | `reference` | *(any)* | Meta-documents like `schema.md`, `clients/README.md` | | `transcript` | `transcripts/` | Raw transcripts of videos, talks, sessions | | `brain` | *(any)* | Brain-level meta documents like `index.md` | Sources: [gbrain.yml:25-38]() ### Which Types Have Timelines Not all page types use a timeline. Types where evidence accumulates over time have an explicit `## Timeline` or `## Session Timeline` section in their template: - **`person`** — `## Timeline` with dated bullet entries - **`client-note`** — `## Session Timeline` with structured session entries - **`event`** — `## Event Timeline` for recurring event instances - **`project`** — `## Timeline` with dated update entries Types like `concept`, `methodology`, `content`, and `reading` do not follow the append-only timeline pattern; they are maintained as living documents with no explicit timeline section, because their content is synthesized knowledge rather than a log of interactions. Sources: [schema.md:66-71](), [schema.md:229-234](), [schema.md:304-308](), [schema.md:346-350]() --- ## File-Naming Conventions All filenames are lowercase and hyphenated. No spaces, no camelCase, no uppercase. | Content type | Convention | Example | |---|---|---| | People | `firstname-lastname.md` | `joseph-gabriel.md` | | Concepts | Descriptive slug | `progress-is-positivity.md` | | Methodologies | Descriptive slug | `breathwork-protocols.md` | | Clients | `firstname-lastname.md` | *(matches person page slug)* | Sources: [README.md:81-85]() Client pages are explicitly linked to person pages via the `client` frontmatter field (set to the person's filename slug) and by convention include a `See also:` link to `../people/firstname-lastname.md`. The person page holds the relationship overview; the client-note page holds the training detail and session log. Sources: [schema.md:196-198](), [clients/README.md:60-64]() --- ## Read vs. Write Decision Table The two-layer model creates a precise decision rule for any agent or human writing to the brain: | Situation | Action | |---|---| | New evidence confirms what's already in compiled truth | No action needed | | New evidence **changes** an existing fact | Rewrite the relevant section of compiled truth | | New evidence **adds** a session, event, or observation | Append a new entry to the timeline (top) | | Current programming changes | Rewrite `## Current Programming` in compiled truth; add timeline entry noting the change | | Client makes a notable quote | Add to timeline; consider surfacing in compiled truth if it reveals a pattern | | A past timeline entry is wrong | **Do not edit it.** Add a corrective note as a new timeline entry with today's date | --- ## How the Layers Interact: An Example In `people/joseph-gabriel.md`, the compiled truth opens with a one-line summary, then expands into structured sections (`## State`, `## What He Believes`, `## What He's Building`, `## Assessment`, `## Trajectory`, `## Contact`, `## Network`). These sections are continuously maintainable prose. The `## Timeline` at the bottom of the same page records dated milestones in reverse order — career events, founding dates, notable moments — each as a bullet under a bold date. These entries were written at a point in time and are never revised; new milestones get prepended. An AI agent asked "What is Joseph currently building?" reads `## What He's Building` from compiled truth and gets an immediate answer. An AI agent asked "When did Joseph found The Art of Life?" scans the timeline and finds `**2024-01** | Founded The Art of Life corporate wellness company`. Sources: [people/joseph-gabriel.md:19-80]() --- ## Summary The compiled-truth / timeline split is not a formatting convention — it is the architectural principle that makes the brain both **trustworthy** and **compoundable**. The compiled truth is always safe to read as current state; the timeline is always safe to append to without fear of corrupting past records. An AI agent writing to a page has exactly two valid operations: rewrite compiled truth when the picture changes, or prepend a new entry to the timeline. Everything else — editing past entries, embedding current state into the timeline, or treating frontmatter as compiled truth — breaks the invariant that makes the brain useful over time. Sources: [README.md:37]()