🧠 Cortex — Context Refinery Pipeline
Source → Content Type → Processing → Classification → Routing → Destination · Hover any node
to
trace its path
Master Overview
The complete pipeline in one view. Hover any node to trace its full path.
Sources
📓 Notion ×2
journals, notes, discussions
🍎 Apple Notes
quick notes, journals
🤖 ChatGPT
long discussions
🔔 Portal
feedback w/ metadata
🎙️ Voice
self-talk, transcribed
📚 External
articles, books
→
Content Type
Fragments
🧩 Charted
🌫️ Uncharted
💭 Stream
Documents
📝 Narrative
💬 Dialogue
📚 3rd-party
→
Process
/ingest
parse · match · route
/distill
read · extract · propose
/learn
extract · store · scan · propagate
→
Classify
/ingest output
Todo
Info
Thinking 🚫
/distill output
Principle
Fact
Action
Insight
Observation
/learn output
About Adrian
Universal
Behavioral
Tool-specific
Project-only
Business KB
Hub / Arch KB
Engineering KB
Personal KB
fog
pattern detect → fog-clear?
→
Route
/ingest
match task → comment
match ongoing → update
thinking → never route
/distill
🟣🟠 always review
🔵 high→auto · low→review
🟢 match→comment · new→create
🟡 review → pattern
/learn + /learn scan
Adrian → profile
Universal → cross-session
Behavioral → agent rules
Tool → module docs
Project → task folder
Business → kb/mipos/
Hub → kb/hub/
Eng → kb/dev/
Personal → kb/adrian/
Propagate → task docs
Propagate → arch docs
fog — 光粒光团
光粒 accumulate → 光团 forms → Nexus?
→
Destinations (21)
🪞 identity
adrian.md
memory.md
Principles.md
.agent/rules/
kb/adrian/
📚 learning (kb)
kb/hub/
kb/mipos/
kb/dev/
⚡ action
D1 Tasks
ongoing.md
task folders
project docs
⚙️ operations
architecture.md
workflow docs
skill docs
☀️ capture
stream.mipos.my
parked (光团)
🧠 thinking
research reports
blog posts
kb/patterns
Hover any node to trace its path through the pipeline
Source → Content Type Mapping
Many-to-many: one source can produce multiple content types. Hover to highlight connections.
📓 Notion ×2
journals, SOPs, notes, discussions
🍎 Apple Notes
quick notes, journals
📋 Smart Todo
daily dumps · todo.adrian.my
💬 WhatsApp
random text, snippets
🤖 ChatGPT
long discussions
⚡ Agentic Session
Gemini / Claude work
🔔 Portal Feedback
feedback with metadata
🎙️ Voice Memo
self-talk, transcribed
📚 External
articles, books, research
🧩 Fragment (charted)
🌫️ Fragment (uncharted)
💭 Fragment (stream)
📝 Document / Narrative
💬 Document / Dialogue
📚 Document / 3rd-party
📄 Document Pipeline → /distill
Long-form content: journals, voice transcripts, AI sessions, articles. Multi-extract if
hybrid.
Documents
Narrative · Dialogue · 3rd-party
→
/distill
Read full doc
Extract items
Classify into 5 types
→
Principle
Fact
Action
Insight
Observation
→
🟣🟠 Always review
🔵 High→auto · Low→review
🟢 Match→comment · New→create
🟡 Review → pattern
→
Principles.md
kb/facts.md
D1 Tasks
kb/patterns
adrian.md
🧩 Fragment Pipeline → /ingest
Short-form content: todos, notes, ideas, portal feedback. Quick classify and route.
Fragments (charted)
Todo, note, info, idea
Portal feedback
→
/ingest
Parse items
Fuzzy match
Hub entities
→
→
Match task → D1 comment
Match ongoing → update
Thinking → never route
No match → leave
→
D1 comment
ongoing.md
KB files
🌫️ 光粒光团 — Light Particle Clustering (Fog-Clearing)
Fog doesn't "clear" — individual 光粒 (light particles) accumulate into a 光团 (light cluster)
until the shape is visible.
光粒 (Sparks)
Initiative seeds
New domain signals
Portal feedback #tags
→Park (feedback)
→光粒 accumulate
→光团 forms (luminance ↑)
→Shape visible → Research-ready → Action-ready
💡 Idea Lifecycle
Ideas cross content types as they mature. They start as fragments but evolve into documents
and
tasks.
💡 Spark (fragment)→
🔍 Exploration (session)→
⚖️ Validation→
🎯 Decision→
✅ Task / Project
Ideas and parking are the same problem — content
without a
home that needs capture → consolidate → graduate (R2 scope)
⚡ Agentic Sessions → /learn (exception path)
Session-level extraction. Runs at end of each agent session via /wrapup.
Agentic Sessions
Mistakes, corrections
Learnings + preferences
→
/learn
Scan session
Extract corrections
+ preferences
→
About Adrian
Universal
Behavioral
Tool-specific
Project-only
→
Adrian → profile
Universal → cross-session
Behavioral → agent rules
Tool → module docs
Project → task folder
→
adrian.md
memory.md
.agent/rules/
docs/
tasks/
📚 KB Pipeline → /learn scan
Domain knowledge capture. Scans sessions for business, architecture, engineering, or
personal
knowledge worth documenting. Merged from /kb into /learn as a scan mode (R1B).
Session Content
Business processes
Architecture decisions
Domain knowledge
→
/learn scan
Scan session
Check existing KB
Propose & merge
→
Business
Hub / Arch
Engineering
Personal
→
Business → kb/mipos/
Hub → kb/hub/
Eng → kb/dev/
Personal → kb/adrian/
→
kb/mipos/
kb/hub/
kb/dev/
kb/adrian/
💭 Stream → Sacred Space
Passing thoughts. No processing.
Browse only — no processing pipeline. Sacred space. → stream.mipos.my
Why →
Capture signals→
Extract knowledge→
Think & propose→
Decide & act→
Better outcomes for Adrian