Info Architecture Migration Plan

Implements the structure defined in /info-architecture-for-this-site.md.

Jobs to Be Done (JTBD)

Two audiences: self (personal knowledge base + retrieval) and general public (readers finding essays via search or sharing). Both matter; they pull in different directions.

J1: Fast capture — zero friction, zero decisions

"I have an idea / link / fragment / full essay. I want it saved immediately with no overhead."

• Modes: desktop editor, Claude agent, copy-paste from web
• Rule must be simple enough for Claude to follow without asking
• No decision about type, folder, or status at capture time
• Add metadata later, not now

J2: Public reading — polished essays surfaced for strangers

"Someone lands on the site. What should they see?"

• Blog/essays feed = finished, intentional work only
• Distinction from notes: polish and completeness, not topic or length
• Notes are public too but not prominently surfaced — a reader who finds one should be fine, but they're not the shop window

J3: Personal retrieval — find my own stuff fast

"I wrote something about X six months ago. Where is it?"

• Search is primary
• Backlinks and related: frontmatter are secondary
• Recent changes feed serves this too (what was I working on?)

J4: Recent changes — what's new or updated

"What have I added or changed lately?"

• Serves both self (navigation, memory) and public (signal of activity)
• Driven by git commit date or updated: frontmatter field — TBD

These jobs drive the tag/metadata decisions below.

Goal

Collapse false taxonomy folders into root. Result: flat structure with /logs and /projects as the only permanent special-purpose folders.

Metadata / Tagging Design

Folder location no longer signals type. Tags do instead — additive, composable, no forced exclusivity (same reason Wikipedia uses categories not folders).

Key frontmatter fields:

Tag vocabulary

Tags compose freely: a file can be [post, stub] while being revised, or [bookmark, post] if a clipping grew into an essay.

• Blog page = filter tags contains post
• Recent changes = sort by updated: or git commit date
• No tag required at capture — add later

Capture rule (for humans and AI agents):

Create file at root. Filename = topic slug (no date). Add created: in frontmatter. Done. Tags can be added later.

Audit Findings

Folder inventory

How the blog feed currently works

blog/index.md contains <List dir="/post" pageSize={30} /> — Flowershow MDX component that reads from /post/. Moving posts to root requires changing this to filter by type: post frontmatter instead. This is the key technical change.

Nav hardcoded paths (config.json)

/about, /now, /tao, /writings, /projects, /book, /blog, /newsletter, /contact

/writings and /blog are nav links. Both need to keep working after migration (either via redirect or kept as index stubs).

Redirect situation

• post/ URLs are the biggest risk: 620 files with slugs, many with WordPress history. These live at /post/slug currently. Target after migration: /slug.
• Redirects: mix of Flowershow config + Cloudflare. Scale is the constraint, not mechanism.
• Existing redirect log exists — check before adding new ones.

Folders to Kill

Folders to Keep

Redirects

• Flowershow + Cloudflare (mix fine)
• Check existing redirect log before adding
• Add redirects before or simultaneously with file moves — never after
• post/ is the big one: 620 potential redirects /post/slug → /slug

Steps

Filename convention for moved files

Natural language, Wikipedia-style. When moving a file:

• Rename to natural title if currently kebab: superintelligence-bostrom-2014.md → Superintelligence.md
• No type suffix: Anna Karenina.md not Anna Karenina Book.md
• No date/author suffix: goes in frontmatter
• Existing kebab files at root: leave as-is (not worth migrating)

1. Quick wins (low risk, do now)

• Delete empty folder: content/ (was already gone)
• Move ref/ files (34) to root
• Move clippings/ (6) to root, add tags: [bookmark] to frontmatter
• Move works/ (5) to root; renamed superintelligence-bostrom-2014.md → Superintelligence.md
• Move misc/ content files to root — lesson learned: misc/index.md had real content (aphorisms, fragments); restored as misc.md
• Note: always check index.md files for real content before deleting

2. Metadata + blog mechanism redesign

• Replace <List dir="/post"> in blog/index.md with "under construction, use search" notice
• Update List component to filter by tags: [post] — blocked: Flowershow has no clean tag-filter listing yet. Revisit when available via obsidian-bases
• Write capture rule for AGENTS.md
• Recent changes feed — deferred, no good mechanism yet

3. Medium-risk moves

• Move nonfiction/ (8 files) to root — redirects added to config.json
• Create nonfiction.md index page at root so /nonfiction still resolves
• speaking/ — leave as-is. Has proper custom landing page (index.md with video/logos/buttons). Functional site section, not false taxonomy. archive.md references a CSV that doesn't exist yet anyway.

4. Big lift — post/ folder — DEFERRED

620 files, 589 with WordPress provenance, all with public URL history. Cost of bulk redirect generation + verification outweighs benefit right now. Blog listing is already "under construction" so no urgency. Revisit when blog mechanism (tag-filter listing) is solved — at that point a scripted mass-move with redirect generation makes sense.

5. Cleanup

• Update info-architecture-for-this-site.md if decisions changed
• Verify site builds

Out of Scope (for now)

• tao/ — leave as-is
• research/ — deep topic structure, leave
• wp-content/ — separate cleanup task

Risk Summary