AIX Zettelkasten
Architecture & Workflow

The Problem

Insights produced in AI conversations have two fates: manual copy-paste with degradation and friction, or evaporation when the tab closes. This system closes that gap.

Every significant idea generated in conversation becomes a durable, linked node in a living knowledge network — with no manual filing and full provenance preserved.

The Architecture

Layer 1 — Conversation

Where thinking happens. Prompt sequences are the cognitive work. The transcript is the primary record. RTW (Read-Think-Write) cycles produce ideas, corrections, and refinements. The /ark command closes each conversation, capturing metadata and exporting to SharePoint.

Layer 2 — Archive (SharePoint)

Where full conversation transcripts live as markdown exports. This is the MICA corpus. MICA (Monadic Inference Cycle Analysis) treats each prompt-response exchange as a discrete unit with its own context and timestamp — the archive is what MICA runs against.

Layer 3 — Atomic Knowledge (TiddlyWiki)

Where individual ideas live as nodes. Each tiddler is one concept, insight, definition, method, or question — tagged, linked, traversable. The wiki runs as a Node.js server in a GitHub repo. Forward and backlinks are automatic: TiddlyWiki parses [[wikilinks]] and surfaces both directions without any manual field maintenance.

Layer 4 — Generation (Node.js / Claude Code)

Where tiddlers are produced programmatically. A synthesis document goes in; correctly-formatted .tid files come out, dropped directly into TiddlyWiki. No copy-paste. No reformatting.

Tiddler Schema

Domain tags — subject area:

Semantic tags — kind of thing:

The source-conversation field is the MICA bridge — every tiddler points back to the conversation that produced it. The wiki is not just a knowledge base but a trace of how the knowledge was made.

MICA & Zettelkasten Together

They operate at different scales and are complementary — not redundant.

The source-conversation field is the bridge. Analysis can move in either direction — from corpus to atom, or from atom back to its originating conversation.

End-of-Conversation Workflow

1. Conversation produces insight — through RTW cycles, directional correction, and synthesis
2. Synthesis document written — in the same session, while the thinking is live. This is the markdown that captures the idea in full.
3. /ark command — captures conversation metadata, exports transcript to SharePoint archive
4. .tid generator runs — synthesis document fed to script; one or more .tid files produced with correct fields
5. Tiddlers land in wiki — knowledge is now a linked node, not a lost conversation

Build Order

1. GitHub repo — initialized with Node.js TiddlyWiki server configuration
2. Seed tiddlers — generate .tid files from the founding synthesis document (PKMAtThePrompt)
3. Stub tiddlers — for each linked concept: MICA RTW InterCycleAnnotation RTWRubric ArkWorkflow LearnerPermit TranscriptAssessment
4. Generator script — markdown synthesis → .tid files, with correct field population
5. Workflow integration — script runs at conversation close, alongside /ark