---
concept: Human Name
aliases: [alt1, alt2]           # 1-2 natural, not keyword stuffing
tags: [domain, subdomain]       # first tag MUST be domain tag
created: YYYY-MM-DD
updated: YYYY-MM-DD
# internal provenance (hidden from Quartz via .wiki-meta.json, NOT in body):
# sources_count, last_source, content_hash managed in wiki/.wiki-meta.json
---

Body sections in order:

1. ## The Problem — easy language, why you care (2-3 lines, teach a friend)
2. ## Formal Definition — textbook citation inline Per Tanenbaum / MDN / Wikipedia: "..." — no Sources section
3. ## Explanation — plain human story, 2-4 sentences
4. ## How It Works — 4-8 numbered steps, mechanism
5. ## Visual Explanation — ONE Graphviz DOT ```dot digraph, rankdir=LR/TB, 4-8 nodes, labeled edges, concept vocab, node [shape=box style=filled fillcolor="#f0f4ff" fontname="Helvetica"]
6. ## Semantic Network — ONE Graphviz DOT ```dot graph semantic_* { layout=neato; THIS [fillcolor="#ffd700"] } — 5-10 real pages, gold center, blue prereq, green builds-into, orange contrasts, gray related
7. ## Key Properties or ## Objectives or ## Functions or ## Pros & Cons — pick ONE compatible with concept, 4-6 bullets
8. ## Real-World Example — one minimal ```cpp|python|bash or story, human runnable
9. ## Connections — 4+ [[page|Name]] — one-line why (no <!-- TODO --> comments in body)
10. ## Edge Cases & Gotchas — 2-3 pitfalls

DO NOT include in public body: Sources section, Active Recall callouts, <!-- TODO: add backlink -->, status: stub markers, Aliases: heavy lists. Provenance lives in hidden summary + .wiki-meta.json.

• YAML frontmatter with: title, type (comparison|deep-dive|analysis), tags, created, updated
• Clear framing: what is being compared/analyzed
• Structured comparison (table or dimensions)
• Insight that goes beyond individual pages
• Connections back to ALL compared concept pages
• Same diagram hygiene: optional single DOT if helpful, no Sources/Active Recall

If page A links to page B, page B must mention page A in its Connections section.

• Built from: prerequisite concepts this one depends on
• Builds into: concepts that use this one
• Contrasts with: similar but different concepts
• Related: adjacent concepts, same domain, neither prerequisite nor application

• Concept pages: 4+ connections
• Synthesis pages: links to ALL concept pages it compares
• Source summaries: links to all concept pages it created/updated

1. Read the source from sources/[filename], compute sha256sum, check wiki/.wiki-meta.json — if hash exists, SKIP unless --force is passed
2. Identify all atomic concepts — 25-30 minimum, no upper limit. Extract every distinct idea, mechanism, pattern, principle. You MUST create at least 25 concept/synthesis pages per source (excluding the summary). If thin, decompose broader concepts into finer atomic pieces. Human teach tone, easy words.
3. Determine topic folder name (source filename, kebab-case), create wiki/[topic]/ if needed
4. For each concept: check if wiki/**/[concept].md exists anywhere — if YES update+merge, if NO create in current topic folder. Every page follows Page Rules: The Problem → Formal Definition (with citation) → 2 DOT diagrams → Properties/Objectives/Functions → Real-World Example → no Sources/TODO in body
5. Create source summary wiki/[topic]/[topic]-summary.md with draft: true (hidden from Quartz) + append ledger entry to wiki/.wiki-meta.json
6. Create/update 1-3 synthesis pages in wiki/[topic]/ if natural tensions exist
7. Create/update wiki/[topic]/[topic]-moc.md (Map of Content) via generate-wiki-mocs.sh logic
8. Run consolidate_images.py tail if source had images, then wiki-lint-det.sh --topic=[topic] deterministic fix (broken stubs, orphans)
9. Update wiki/index.md — one table row per topic, wiki link + one-line human description only. No Stats/agent-module refs, no updated: bump.
10. Append to wiki/log.md + wiki/open-questions.md (2-4 open questions). Verify count: ls wiki/[topic]/*.md | grep -v summary | wc -l must be 25-30, else decompose further. Fail ingest if still <25.

1. Deterministic pre-pass (scripts/wiki-lint-det.sh — full wiki, no LLM, <5s): grep all [[links]], verify file exists → broken list; stub auto-create if link appears ≥2 times (heuristic), else flag; orphan zero-inbound → inject Related link from same topic MOC; tag domain check; missing The Problem/Formal Definition/2×DOT/missing Connections<4 → flag.
2. Sharded LLM pass (2-3 topics/week round-robin, fits free tier): run opencode --agent lint-deep ONLY on shard wiki/[topic]/ for contradictions (Pass 5) + thin fleshing (Pass 6). Full coverage monthly.
3. Report findings and fix what you can; append to wiki/log.md Format: ## [YYYY-MM-DD] — Lint: [scope] with counts: Resolved: N broken (M stubs), N orphans, N tags + Flagged: ...