Skip to main content

Education Research Wiki

LLM-maintained wiki for education startup development - Systematic knowledge organization supporting product decisions with research evidence.

Core Purpose​

This is an LLM-maintained wiki where Claude actively:

  • Processes research sources (papers, articles, videos, case studies)
  • Extracts key information and integrates into existing knowledge
  • Maintains cross-references between concepts
  • Detects contradictions and updates claims
  • Builds a structured, queryable knowledge base

Human: Curate sources, direct analysis, ask questions. LLM: Bookkeeping, cross-referencing, consistency, synthesis.

Content Catalog​

See readme.md for full content catalog (30 files: product concepts, learning science, market research, competitors, strategic context).

Why LLM-Maintained Wiki?​

"The tedious part of maintaining a knowledge base is not the reading or the thinking — it's the bookkeeping." — Andrej Karpathy

Humans struggle with: Updating 10-15 related pages, maintaining cross-refs, detecting contradictions.

LLMs excel at: Simultaneous multi-file updates, systematic cross-referencing, pattern recognition.

Not traditional RAG: This wiki sits between you and raw sources as a curated, interconnected knowledge layer.

  • ❌ RAG: Query → Search raw docs → Generate answer
  • ✅ LLM Wiki: Query → Search maintained wiki → Synthesize → Update wiki

Three-Layer Architecture​

Layer 1: Raw Sources (immutable)

  • Research papers (PDFs, arXiv links)
  • Blog posts and articles
  • Video lectures and talks
  • Case studies and reports
  • Data files and datasets
  • Stored in /docs/education/sources/ (if needed) or referenced via URLs

Layer 2: The Wiki (LLM-maintained)

  • Topic pages (pedagogical theories, learning science, edtech tools)
  • Entity pages (researchers, institutions, frameworks, methodologies)
  • Concept summaries with cross-references
  • Synthesis documents connecting research findings
  • All markdown files in /docs/education/

Layer 3: The Schema (this file)

  • Defines structure, conventions, workflows
  • Guides how Claude processes and organizes information
  • Sets taxonomy for education topics

Core Operations​

Ingest: Processing New Research​

When provided with new sources:

  1. Extract: Main findings, methodology, evidence quality, key concepts, contradictions
  2. Update 10-15 related pages: Topic pages, entity pages, cross-references
  3. Log: Add entry to log.md with date, action, source
  4. Maintain: Update readme.md content catalog, note contradictions, link concepts

Query: Synthesizing Knowledge​

  1. Search wiki pages (Obsidian CLI or grep)
  2. Synthesize answer with citations
  3. If valuable → save as new wiki page
  4. Update cross-references and log

Lint: Health Checks​

Periodically check: contradictions, stale claims, orphans, broken links, dead-ends.

Commands: See parent /docs/CLAUDE.md for Obsidian CLI reference (orphans, deadends, unresolved).

Activity Log​

log.md - Chronological append-only record:

Format: [YYYY-MM-DD] <action> - <description> - <source>

[2026-05-04] CATALOG - Streamlined CLAUDE.md, added content catalog
[2026-04-30] INGEST - Processed Bjork paper on desirable difficulties
[2026-04-30] UPDATE - Updated spaced-repetition.md with new findings

Target Taxonomy​

Folder structure to organize research:

  • pedagogy/ - Constructivism, Montessori, Waldorf, project-based learning
  • competitors/ - Platform analyses

Obsidian CLI Usage​

Prefer Obsidian CLI over direct file operations for link graph updates and backlink maintenance.

Full command reference: See /docs/CLAUDE.md for complete Obsidian CLI examples (search, read, append, backlinks, orphans, etc.)

VAULT path: /Users/deepaksood/Library/CloudStorage/GoogleDrive-dsood@confluent.io/.shortcut-targets-by-id/1a1SooxwlvVEf843YfQKUVj04gJnPyMGO/deepaksood619.github.io/docs

Content Standards​

CRITICAL: Use standard markdown links, NOT Obsidian wikilinks.

❌ Wrong: [[active-learning]] (breaks Docusaurus) ✅ Correct: [Active Learning](education/pedagogical-theories/active-learning.md)

Link Path Requirements:

  • Use full paths starting with education/ (not relative ../ paths)
  • Example: [Market Analysis](education/competitors/technical-hiring-assessment-market.md)
  • NOT: [Market Analysis](../competitors/technical-hiring-assessment-market.md)
  • Reason: Docusaurus builds work better with full paths from /docs/ directory

Research Paper Template​

---
title: Paper Title
authors: Author Names
year: 2024
source: https://arxiv.org/abs/...
type: research-paper
tags: [learning-science, meta-analysis]
---

# Paper Title

## Key Findings
- Finding 1
- Finding 2

## Evidence Quality
[Strong/Medium/Weak] - Explanation

## Connections
- Related to: [Spaced Repetition](../learning-science/spaced-repetition.md)
- Contradicts: [Previous claim](../learning-science/cognitive-load.md#section)

## Startup Implications
How this research informs product decisions

Key Workflows​

Ingest Research:

  1. Create note with frontmatter
  2. Search & update 10-15 related pages
  3. Add cross-references
  4. Log entry in log.md
  5. Update content catalog in readme.md

Query & Synthesize:

  1. Search wiki (CLI or grep)
  2. Check backlinks/outgoing links
  3. Synthesize with citations
  4. Save valuable analyses as new pages
  5. Log query

Link Maintenance:

  • Search for concepts mentioned but not linked
  • Check backlinks to find connections
  • Find orphans, add cross-references
  • Note contradictions explicitly with ⚠️

Periodic Lint:

  • Check orphans, dead-ends, broken links
  • Review recent changes, tag consistency
  • Update stale research, consolidate duplicates

Markdown & Formatting​

See /docs/CLAUDE.md for full Docusaurus markdown guidelines (blank lines, code blocks, MDX compatibility).

Key rules:

  • Blank lines between all content blocks
  • Standard markdown links (no wikilinks)
  • Language specified for code blocks
  • Relative paths for links/images

MDX compatibility (CRITICAL):

  • Escape < and > symbols with backticks: <50 employees, >100 users
  • Don't use [] square brackets without escape (used for links/anchor tags)
  • These symbols are parsed as JSX/HTML tags in MDX and will break the build

Research Quality Standards​

Evidence Quality: Always note [Strong/Medium/Weak/Anecdotal], sample size, replication status.

Citations: Include source URL/DOI in frontmatter, link to originals, track affiliations.

Cross-Referencing: Link all mentioned concepts, create pages for important new concepts, update bidirectional links.

Synthesis Quality: Distinguish consensus vs emerging research, note context specificity, track temporal aspects.

Startup-Relevant Focus​

Always consider: Product implications, scalability, measurability, evidence strength, implementation cost, target audience.

File Naming​

  • Semantic, lowercase, hyphens: spaced-repetition.md
  • Frontmatter: type, tags, source, year
  • Images: /docs/media/ with relative paths

Expected Behaviors​

New source: Extract findings → Update 10-15 pages → Log → Update readme.md catalog → Cross-link → Note contradictions

Question: Search wiki → Synthesize with citations → Save if valuable → Log query

Proactive: Monthly orphan checks, verify links, update stale research, consolidate duplicates

Security​

Do not access office/* folder - contains sensitive data requiring explicit approval.