Achhina's Digital Garden

2026-04-25 Vault Restructure Day

3 min read

2026-04-25 Vault Restructure Day

What I set out to do

Triage Brain Dump — the rolling list of 2024 + 2026 items had drifted, and I wanted to clean it up.

What I actually did

The brain dump triage almost immediately surfaced a deeper problem: I had no rules for where things go. References, journals, reflections, and how-tos were all stranded in Inbox/ because nothing told me where they belonged. So we paused the brain dump and rebuilt the vault.

By the end of the session:

  • Adopted a mode-based top-level structure: Atlas/ (thinking), Chronicles/ (reflecting), Praxis/ (acting). Replaced Zettelkasten/, People/, Devices/, Projects/ with these three modes plus a scaffolding bin (x/).
  • Migrated ~250 files into the new structure. Wikilinks survived because Obsidian resolves by filename.
  • Wrote README with zen + spec + decision log all in one file. 16 ADRs document the structural choices.
  • Closed Vault Restructure as completed.
  • Triaged Archive/ (deleted as a concept — status frontmatter replaces it) and the rest of Inbox/.
  • Set up index.md auto-generation from x/quartz-homepage.md so the public Quartz homepage no longer clutters my vault root.

Decisions worth remembering

  • Methodology-named folders are a beginner artifact. Zettelkasten/ described how I take notes, not a content type. Modern PKM voices (Doto, Tietze, Matuschak) all push toward functional naming or flat structure.
  • Capture decisions, not preferences. Personal-domain heuristics (travel, restaurants, learning) should describe how I decide, not what I like — preferences shift, processes are durable.
  • The vault serves both me and agents. This is the constraint that justifies keeping tasks in Obsidian (instead of a separate task app), since splitting breaks unified context for LLM use.
  • ADHD-aligned capture matters more than perfect structure. Considered removing Inbox/ for cleanliness; kept it because friction-free capture is non-negotiable.

What was striking

The convergence across Diataxis, LYT/ACE, ReAct, Kolb’s Learning Cycle, and OODA on mode-based decomposition of cognitive work felt like real signal, not coincidence. Knowledge work decomposes into discrete modes (thinking / reflecting / acting), and good systems give each mode a place. That’s what made the Atlas/Chronicles/Praxis triad feel right rather than arbitrary.

What I got wrong (and corrected)

I (or Claude) initially claimed Quartz v4 hardcoded the static/ folder location and we’d need a carve-out. After pushback to actually check the source, that was wrong — Quartz’s Plugin.Static() reads from the Quartz install dir, not the user’s content. User-content images are handled by Plugin.Assets(), which resolves by wikilink. Fixed and recorded as ADR-012. Worth remembering: agent assertions about external tools should be verified against source before structural decisions.

Top 3 tomorrow

  1. Solidify Thailand itinerary and book reconvene meeting with Ikjyot Gill
  2. Notify Shu about upcoming vacation
  3. Send follow-up to Clare Deck / Shu / Aji to confirm transfer timeline

Graph View