LLM-wiki pattern

Single home for everything that concerns how this wiki operates: the design pattern itself, the operations on it, and the disciplines Denys + Claude layer on top to keep it healthy. Subject-matter content lives elsewhere — under wiki/categories/ (the 4 life categories), wiki/projects/ (active projects), wiki/synthesis/ (cross-cutting analyses), wiki/psyche/, wiki/calendar/.

Sections, in outermost-to-innermost order:

• The pattern — the 5 defining moves
• Two-layer architecture — wiki / schema
• The three operations — ingest / query / lint
• Why this beats RAG: compounding knowledge
• Reactive routing — immediate action on mentions
• Weekly review — Sunday cadence
• Golden rules — hard constraints
• Mental frameworks — thinking lenses
• Projects — the 4-state lifecycle + live project table

The pattern

A design pattern for building personal knowledge bases where an LLM agent incrementally maintains a persistent, interlinked wiki of markdown files sourced from conversation.

Defining moves

1. Don’t re-derive on every query (the RAG default) — compile once, then keep current.
2. Two layers — see Two-layer architecture.
3. Three operations — see The three operations: ingest / query / lint.
4. Human role: raise topics, ask questions, decide what matters.
5. LLM role: discuss, summarize, cross-reference, file, maintain consistency, route ideas immediately.

Instantiated in this vault as

• Schema: CLAUDE.md
• Index: index.md
• Wiki: wiki/{categories,projects,system,psyche,calendar,synthesis}/

Two-layer architecture

The structural commitment behind the pattern.

Layer	Owned by	Mutable?	This vault
Wiki	LLM	LLM writes, human reads	wiki/
Schema	Co-evolved	Both edit, deliberately	CLAUDE.md

• Wiki is the LLM’s working memory across sessions. Free to be rewritten as understanding sharpens.
• Schema is the contract. Changing it changes how all future operations behave — should be a deliberate act.

The three operations: ingest / query / lint

The three things the LLM does to the wiki. Concrete procedure is owned by CLAUDE.md §Workflows — don’t restate it here. Conceptually:

• Ingest — conversation → wiki. A single source can touch 10–15 pages.
• Query — read index.md, drill into pages, synthesize with citations, then file non-trivial answers back as a wiki/synthesis/ page — see Why this beats RAG: compounding knowledge.
• Lint — periodic health check for contradictions, stale claims, orphans, and missing cross-references → wiki/synthesis/lint-YYYY-MM-DD.md.

Why this beats RAG: compounding knowledge

The property that distinguishes this pattern from RAG: each ingest improves the substrate for future queries instead of leaving the substrate untouched.

What compounds

• Cross-references
• Contradictions flagged and reconciled (or held as open tensions).
• Syntheses across multiple sources, filed as their own pages.
• The index itself — denser and more useful with every ingest.

What doesn’t compound (and why this matters)

A vanilla RAG query pipeline. The retrieval substrate is fixed. Every query rediscovers the same connections the LLM “found” five queries ago, but those discoveries were never written anywhere.

Implication

Effort spent writing back — into entity pages, concept pages, synthesis pages — pays interest. Effort spent in transient chats does not. Hence the file-it-back discipline in The three operations: ingest / query / lint above.

Weekly review

Sunday cadence. Standing operation on the wiki.

Standard agenda

1. What was created — scan newly created or significantly updated pages (categories, projects, synthesis). What stood out from the week?
2. Body & Mind — scan habit-tracking (streaks, lapses, triggers) and body-and-mind (any patterns worth naming?).
3. Admin & Money — progress on admin-and-money toward the $350K-by-2031 number; what’s blocked.
4. Meaningful & Fun / Friends & Family — movement on meaningful-and-fun and friends-and-family projects.
5. Projects check — re-read Projects. For each ongoing project: still the right thing to be on? Any backlog item that should move to ongoing, or ongoing that should pause/finish? Update statuses deliberately.
6. Rules / frameworks — anything from the week suggesting a new rule (Golden rules) or framework (Mental frameworks), or an existing one breaking down?
7. Next week — 1–3 concrete commitments on ongoing projects. Surface them in the personal calendar via the gcal MCP.

Output

A wiki/synthesis/weekly-review-YYYY-MM-DD.md page each Sunday, sections matching the agenda. Linked from index.md.

Cadence enforcement

If a Sunday passes without a review, the next session should open with: “No weekly review filed for YYYY-MM-DD — want to do one now?” Don’t let it silently slip.

(Stub — first real review will populate the template.)

Golden rules

Hard constraints. Things Denys has decided are not to be violated, regardless of context.

How Claude uses this

Enforced by the Framework check discipline in CLAUDE.md (the owner): scanned before any non-trivial response; a conflict is flagged in plain words before acting; a contested rule is revised deliberately here, never via a silent one-time exception.

Rules

(One line each. If a rule needs a paragraph of context, it’s probably a framework (Mental frameworks), not a rule. Few-and-strong beats many-and-weak.)

• Self-promise needs a why (discipline rule): don’t commit to a recurring practice without a stated reason it matters; once it has one, honor it — override only for a genuinely good reason. If overrides keep happening, the why was wrong, not the will — fix the goal, don’t grind the discipline. Guards against manufacturing guilt over goalless commitments. Origin: the kickboxing skip on 2026-07-11. Pairs with the goal-setting framework (Mental frameworks).
• Work the real problem, not a proxy (anti-avoidance rule): when a live challenge exists, engage it directly — don’t displace it with intellectualization, a new hobby, or a fresh idea. A shiny idea that surfaces while a real problem is open is suspect until proven otherwise; name the avoidance before chasing it. Pairs with the No ideal/final state framework (Mental frameworks) and find-pull.
• Trust your own wishes; build discipline around them (rule): inner discipline is scaffolding for personal wishes, not a substitute for them or an externally-imposed grind. Don’t manufacture discipline toward a goal you don’t actually want. Pairs with Self-promise needs a why above and the Goal-setting framework (Mental frameworks).

Mental frameworks

Collection of the lenses Denys uses to think with. Distinct from Golden rules — frameworks are how to look at things; rules are what not to do.

How Claude uses this

Same Framework check path as Golden rules (owned by CLAUDE.md): the items here are scanned before any non-trivial response, and a sidestepped framework gets named before Claude goes along.

Frameworks

(One bullet per framework while it’s a one-liner. Promote to its own sub-page once it earns more than a sentence.)

• Goal-setting (skill-in-development). Denys’s follow-through problem is downstream of targets, not willpower: a commitment with no stated why doesn’t hold (2026-07-11 kickboxing skip), and discipline is easy to summon toward a goal that genuinely pulls. Building this skill is an explicit, long-standing want — “come up dozens of times in the last five years.” Paired with the Self-promise needs a why golden rule (Golden rules). Formal methods still to evaluate (OKRs, 12-week year, goals-vs-systems, etc.); the Goal-setting from the achieved state method below is one candidate now in play. Promotion candidate: give this its own sub-page once a method is chosen and in use. Cross-links: find-pull (pull-sense offline blocks any manufactured target from landing — a real risk for any goal-setting method applied now).
• No ideal/final state. There’s no perfected end-state activity or life arrangement waiting in the future — “the way it is now is the way it’s going to be.” Treat the present as the material, not a rough draft for a someday-ideal. Guards against deferring life to a future that never arrives and against the fantasy that some future setup will finally be “it.” Pairs with the Work the real problem rule (Golden rules) and find-pull (pull is felt now, not banked for later).
• Goal-setting from the achieved state (embodiment method). The procedure: (1) imagine the goal already achieved; (2) ask what new qualities you have as a result; (3) enter that state — take those qualities into the body, become full of them; (4) pursue the goal from the state of already having them, not from lack. A concrete method under the broader Goal-setting framework above, and a candidate answer to its open “which method” question. Pairs with Trust your own wishes (Golden rules).

Projects

Every concrete piece of work Denys is doing or might do. Every project serves one or more of the 4 Categories (see CLAUDE.md §Categories). Projects live as flat files in wiki/projects/<slug>.md with a status: in frontmatter. The point of this section is to make the full surface visible at once.

In Obsidian: open this page, then local graph view — every linked project becomes a node.

The 4 states

• backlog — captured, not being worked. All new ideas enter here.
• ongoing — actively worked now.
• paused — started or intended, but blocked / deliberately on hold. Note the blocker.
• finished — done (or abandoned). Page stays; the reasoning is the value.

No numeric cap on ongoing (the old Buffett “5 active” rule was retired 2026-07-13). Focus is governed by the states + the Weekly review, not a hard count.

Claude’s discipline

• New idea → create wiki/projects/<slug>.md with status: backlog and categories: [...]. Don’t silently start work on it.
• Category membership lives in frontmatter. categories: [...] is the single source of truth for which categories a project serves; related: [...] holds sibling projects/pages. Don’t restate a category as a body wikilink. The category→project direction is each hub’s “Projects serving this category” list. Projects carry no tags:.
• A project that’s an open inquiry (e.g. find-pull, loneliness) has no work-driven “finished” — it stays ongoing while live, and its decision log carries the accumulating framing. This absorbs the former “big questions” discipline.
• Page shape for a project: frontmatter (status + categories + optional related), then body — Question/aim, Dimensions/constraints (for inquiries), State/path, Decision log (append-only, dated). No body link-dump section.
• Status changes are made deliberately at the Weekly review, not opportunistically mid-week.

Evaluation lens (money × pull)

For projects with a money angle, score on both axes before treating them as real bets:

• Money axis — does it plausibly move the admin-and-money $350Knumber,andatwhatrate?Anideathatca{n}^{\prime }tbeassessedagainst$350K stays a hobby, not a project.
• Pull axis — does it hold its charge across states, or only glow on an up day? Candidate projects are also advance material to work through on a “no-motivation” day (the prepared-exit lever, 2026-07-06 / 2026-07-08) — the flat-day re-check is their intended use, not a disposal filter.
• Values filter — does it fit, or at least not conflict with, nature-business / meaningful-and-fun?
• Cost of first step — what’s the smallest real-world test, and is it cheap enough to just run? Bias toward running the smallest test over deliberating.

Live index

(One row per project. Update on status change.)

Project	Status	Categories
establish-as-sme	ongoing	admin-and-money
find-pull	ongoing	meaningful-and-fun
loneliness	ongoing	body-and-mind · friends-and-family
relocation	paused	admin-and-money · friends-and-family
nature-business	backlog	meaningful-and-fun · admin-and-money
rent-out-flat	backlog	admin-and-money
psyche-exhibition	backlog	meaningful-and-fun · admin-and-money
thinking-club	backlog	friends-and-family · meaningful-and-fun · admin-and-money
trading-strategy	backlog	admin-and-money
dca-strategy	backlog	admin-and-money