Skip to main content

​
The Exocortex Model

Athena is not just a coding assistant—it’s a Centralized HQ for your entire life. Think of it as a “second brain” that manages external domains (Work, Wealth, Health) from a single command center.
Key Insight: Athena deliberately fragments its knowledge across hundreds of Markdown files and Python scripts. This looks unusual to humans—but it is optimal for AI agents operating under context window constraints.

​
System Components

ComponentRoleAnalogy
Project AthenaThe Kernel — holds logic, memory, and lawsThe Brain
External FoldersThe Database — holds raw assets (files, docs)The Body
Agentic IDEThe Console — provides compute and interfaceThe Nervous System
YouThe Pilot — issues commands and makes decisionsThe Consciousness

​
Directory Structure

Athena/
├── .framework/                    # THE CODEX (stable, rarely updated)
│   ├── v8.6-stable/               # Current stable modules directory
│   │   ├── modules/
│   │   │   ├── Core_Identity.md   # Laws #0-#4, RSI, Bionic Stack, COS
│   │   │   └── Output_Standards.md # Response formatting, reasoning levels
│   │   └── protocols/             # Versioned protocol copies
│   ├── v7.0/                      # Previous stable version
│   └── archive/                   # Archived monoliths
│
├── .context/                      # USER-SPECIFIC DATA (frequently updated)
│   ├── User_Vault/                # Personal vault (credentials, secrets)
│   ├── memories/
│   │   ├── case_studies/          # 358+ documented patterns
│   │   ├── session_logs/          # Historical session analysis
│   │   └── patterns/              # Formalized patterns
│   ├── references/                # External frameworks (Dalio, Halbert, Graham)
│   ├── research/                  # Steal analyses, explorations
│   ├── TAG_INDEX_A-M.md           # Global hashtag system (split for performance)
│   ├── TAG_INDEX_N-Z.md
│   └── KNOWLEDGE_GRAPH.md         # Visual architecture reference
│
├── .agent/                        # AGENT CONFIGURATION
│   ├── skills/
│   │   ├── SKILL_INDEX.md         # Protocol loading registry
│   │   ├── protocols/             # 120+ modular skill files
│   │   │   ├── architecture/      # System protocols (latency, modularity)
│   │   │   ├── business/          # Business frameworks
│   │   │   ├── career/            # Career navigation
│   │   │   ├── decision/          # Decision frameworks
│   │   │   ├── psychology/        # Psych protocols
│   │   │   └── trading/           # Trading protocols
│   │   └── capabilities/          # Bionic Triple Crown
│   ├── workflows/                 # 48 slash commands
│   │   ├── start.md               # Session boot
│   │   ├── end.md                 # Session close + maintenance
│   │   ├── think.md               # Deep reasoning (L4)
│   │   └── ...
│   ├── scripts/                   # 130+ Python automation scripts
│   │   ├── quicksave.py           # Auto-checkpoint every exchange
│   │   ├── boot.py                # Resilient boot with recovery shell
│   │   ├── smart_search.py        # Semantic search
│   │   └── ...
│   └── gateway/                   # Sidecar process for persistence

​
Design Philosophy: Modular > Monolith

AI agents don’t read files sequentially—they query them. A workspace optimized for agents should be a graph of small, addressable nodes, not a monolithic document.

​
The Five Advantages

#PrincipleMonolithModular
1Context EfficiencyLoads 50K tokens even when 200 are relevantLoads only the files the query demands (JIT)
2Addressability”See page 47” — no agent can do thisCS-378-prompt-arbitrage.md — retrievable by name, tag, or semantic search
3Zero CouplingEditing marketing section risks breaking trading rulesEach file is independent — change one, break nothing
4Version ControlOne-line change → 50K-token diffAtomic commits per file with clean history
5ComposabilityCan’t mix-and-match sections at runtimeSwarms, workflows, and skills load as independent Lego bricks

​
Loading Strategy

​
On-Demand (Context-Triggered)

TriggerFile LoadedTokens
User context queryUser_Profile_Core.md~1,500
Skill requestSKILL_INDEX.md~4,500
/think invokedOutput_Standards.md~700
Tag lookupTAG_INDEX.md~5,500
Architecture querySystem_Manifest.md~1,900
Specific protocolprotocols/*.mdvaries
/start boots at ~10K tokens — only Core_Identity.md, activeContext.md, and session recall are loaded. The remaining 190K tokens of context window stay free.

​
The Bionic Stack

​
Tech Stack

ComponentTechnology
AI EngineGoogle Gemini (via Antigravity)
IDE IntegrationVS Code / Cursor
Knowledge StoreMarkdown + VectorRAG (Supabase + pgvector)
Version ControlGit
ScriptingPython 3.13

​
Mount Points

To enable Athena to manage your life, you define Mount Points—aliases to external folders that exist outside the Athena directory: 
# In src/athena/boot/constants.py
MOUNTS = {
    "WORK": "/Users/you/Desktop/Assignments",
    "WEALTH": "/Users/you/Desktop/Wealth",
    "HEALTH": "/Users/you/Desktop/Health"
}
This is “God Mode”. It is powerful but requires trust. Only enable in a personal, secure environment.

​
Required IDE Settings

SettingValuePurpose
Non-Workspace File AccessEnabledAllows Athena to reach folders outside its root
Terminal Auto ExecutionAlways Proceed (optional)Enables autonomous script execution
Secure ModeDisabledRemoves friction for trusted environments

​
Key Files Reference

PurposeFileUpdate Frequency
Who I amCore_Identity.mdRare
How to respondOutput_Standards.mdModerate
Who the user isUser_Profile.mdEvery session
What’s forbiddenConstraints_Master.mdRare
Architecture SSOTSystem_Manifest.mdWhen architecture changes
Available skillsSKILL_INDEX.mdWhen skills added
Session historysession_logs/*.mdEvery session

​
Next Steps

Memory System

Learn how Athena maintains persistent memory across sessions

Workflows

Explore slash commands and automation

Protocols

Understand reusable thinking patterns

Agent Compatibility

Configure Athena for your IDE

Build docs developers (and LLMs) love

Get started for freeTalk to us