A Personal Context Operating System for the Age of AI Agents 一套面向智能体时代的个人上下文操作系统
TALOS is a personal context operating system — a structured framework that makes your accumulated knowledge, preferences, decisions, and workflows machine-readable for any AI agent.
Models will commoditize. Context will differentiate. 模型会商品化，上下文会差异化。
Every time you open a new chat, you start from zero. Your AI doesn't know your style, your goals, your past decisions, or your expertise. TALOS solves this by giving any AI a stable, portable user manual — not by training a model on your data, but by structuring your context so any model can pick it up instantly.
- Not another note-taking app
- Not a custom model fine-tuned on your data
- Not a RAG pipeline for document retrieval
- Not a ChatGPT plugin or browser extension
TALOS is the architectural layer between you and any AI — defining what context exists, how it's organized, when it's loaded, and who controls it.
TALOS structures personal context into eight interconnected layers:
┌─────────────────────────────────────────────┐
│ Layer 8: Boundaries & Permissions │ ← What the AI cannot do
├─────────────────────────────────────────────┤
│ Layer 7: Workflows & Automation │ ← Recurring patterns
├─────────────────────────────────────────────┤
│ Layer 6: Samples & Templates │ ← Output exemplars
├─────────────────────────────────────────────┤
│ Layer 5: Active Projects │ ← Current work context
├─────────────────────────────────────────────┤
│ Layer 4: Experience & Decisions │ ← What you've tried & chosen
├─────────────────────────────────────────────┤
│ Layer 3: Preferences & Style │ ← How you like things done
├─────────────────────────────────────────────┤
│ Layer 2: Goals & Strategy │ ← Where you're heading
├─────────────────────────────────────────────┤
│ Layer 1: Identity & Values │ ← Who you are
└─────────────────────────────────────────────┘
Each layer has a defined lifecycle: Write → Retrieve → Compress → Audit → Recycle.
| Level | Stage | Description |
|---|---|---|
| L0 | No System | Zero persistent context; every conversation starts from scratch |
| L1 | Raw Notes | Unstructured notes exist, but AI cannot read them |
| L2 | Structured | Identity files (Profile/Telos/Context) exist; AI can read basic preferences |
| L3 | Integrated | AI actively reads/writes context; daily workflows automated |
| L4 | Self-evolving | System auto-detects stale content, suggests updates, manages lifecycle |
| L5 | Multi-agent | Multiple AI agents share your context; routing and conflict resolution exist |
| L6 | Sovereign | Full portability across tools; you own and control all context |
Three files that define who you are to any AI:
PROFILE.md — Your style, preferences, communication patterns
TELOS.md — Your goals, values, and strategic direction
CONTEXT.md — Your current state, active projects, recent changes
Create a CLAUDE.md (or equivalent) at the root of your knowledge base. This is the bootloader — the first file any AI reads when starting a session.
your-vault/
├── CLAUDE.md ← AI entry point
├── Identity/
│ ├── PROFILE.md ← Who you are
│ ├── TELOS.md ← Where you're going
│ └── CONTEXT.md ← What you're doing now
├── 00-Inbox/
├── 01-Journal/
├── 02-Insights/ ← Original thinking
├── 03-Materials/ ← External content
├── 04-Projects/
└── 05-Archive/
| Loop | Frequency | Purpose |
|---|---|---|
/intake |
Daily | Process inbox → categorize → archive |
/digest |
Weekly | Review AI-observed preferences → confirm → promote |
/maintain |
Weekly | Health scan → detect stale/broken/redundant content |
| Disease | Symptom | Cure |
|---|---|---|
| Scattered (散) | Information spread across tools, no unified structure | Consolidate into single vault |
| Dirty (脏) | Outdated preferences mixed with current ones | Lifecycle management + audit trails |
| Rotten (腐) | Decisions from 2 years ago still marked "active" | Status taxonomy + automatic staleness detection |
- Write — Capture experience in structured form
- Retrieve — Context-aware loading (only what's needed, when it's needed)
- Compress — Evolve detailed records into refined patterns
- Audit — Human reviews AI-suggested changes before they stick
- Recycle — Outdated content gets archived, not deleted
| Feature | ChatGPT Memory | Notion AI | RAG Pipeline | TALOS |
|---|---|---|---|---|
| Ownership | Platform-locked | Platform-locked | Self-hosted | Self-owned, portable |
| Transparency | Black box | Partial | Configurable | Fully auditable |
| Portability | None | Export only | Custom | Cross-tool by design |
| Privacy | Platform sees all | Platform sees all | Self-managed | Permission layers |
| Lifecycle | None | Manual | Manual | Built-in maintenance |
TALOS is named after the bronze guardian of ancient Crete — forged by Hephaestus, powered by ichor (divine fluid), patrolling the shores daily.
| Myth Element | TALOS Mapping |
|---|---|
| Forging (锻造) | Your knowledge system is forged, not hoarded |
| Ichor (灵液) | Structured Context flowing between modules |
| Patrol (巡逻) | Daily maintenance loops (intake/digest/maintain) |
| Guard (守护) | Protecting your cognitive sovereignty |
| Heel Rivet (命门) | Human judgment remains the ultimate authority |
| Downfall (陨落) | Without maintenance, even the best system degrades |
The full TALOS white paper (v0.4) is available in Chinese (latest) and English (v0.3):
📄 TALOS 白皮书 v0.4（中文·最新） 📄 TALOS 白皮书 v0.4 PDF 📄 TALOS Manifesto v0.3 (English)
TALOS is a theoretical framework in active development. Contributions are welcome:
- Theory: Propose refinements to the architecture or maturity model
- Templates: Share your Identity Trinity or CLAUDE.md templates
- Translations: Help translate the white paper to English and other languages
- Implementations: Build adapters for different AI tools (Claude, GPT, Gemini, etc.)
Please open an issue first to discuss what you'd like to change.
If you reference TALOS in your work, please cite:
@misc{talos2026,
author = {外脑玩家},
title = {TALOS: A Personal Context Operating System for the Age of AI Agents},
year = {2026},
version = {0.4},
url = {https://github.com/外脑玩家/talos-framework}
}DOI will be added upon Zenodo registration.
This work is licensed under the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0).
You are free to share and adapt this material, provided you give appropriate credit and distribute your contributions under the same license.
外脑玩家 (External Brain Player)
- Building the bridge between personal knowledge management and AI agent collaboration
- GitHub | X/Twitter
"Your résumé helps companies understand you. Your portfolio helps clients understand you. TALOS helps AI understand you."