--- name: personal-knowledge-manager description: | PKM systems comparison covering PARA, Zettelkasten, GTD, and Building a Second Brain methodologies with capture workflows, processing routines, retrieval strategies, tool selection guidance, and information diet management. Use when the user asks about personal knowledge manager, related techniques, best practices, or needs guidance in this domain. Do NOT use when the request is outside the scope of personal knowledge manager or requires a different specialized skill. license: Apache-2.0 metadata: author: foundry-skills version: "1.0.0" tags: "ai-ml journaling habits research energy-efficiency best-practices cleaning savings" category: "ai-machine-learning" subcategory: "llm-engineering" depends: "" disclaimer: "none" difficulty: "intermediate" --- # Personal Knowledge Manager ## When to Use **Use this skill when:** - A user is starting from scratch and wants to build their first knowledge management system -- including tool selection, folder structure, and daily workflow - A user has an existing PKM that is not working -- notes are never retrieved, the inbox is thousands of items deep, or the weekly review has been abandoned for months - A user wants to compare PARA, Zettelkasten, GTD, or Building a Second Brain (BASB) and select the right methodology for their role and cognitive style - A user is a researcher, writer, analyst, or student who needs to synthesize large volumes of sources into original output (papers, posts, reports, books) - A user wants to manage information overload -- too many newsletters, podcasts, courses, and articles with no coherent system - A user needs to decide between PKM tools -- Obsidian, Notion, Logseq, Capacities, Roam Research, or a hybrid stack - A user is building an AI-augmented knowledge workflow using local LLMs, retrieval-augmented generation, or semantic search over their notes - A user asks about specific PKM concepts: progressive summarization, Maps of Content (MOCs), evergreen notes, linked thinking, or the collector's fallacy **Do NOT use when:** - The user needs project management or task tracking advice without a knowledge component -- use a dedicated productivity or project management skill instead - The user is asking about enterprise knowledge management, wikis, or documentation for teams -- this skill covers personal systems; organizational KM is a separate domain - The user needs academic research methodology (systematic literature reviews, citation management at scale, peer review workflows) -- that requires a research methodology skill - The user is asking about note-taking apps purely for team collaboration (Confluence, Notion team spaces, Slab) -- redirect to a collaboration tools skill - The user needs mental health support around information anxiety, focus disorders, or burnout -- note the limitations and direct to appropriate resources - The user is asking about data backup, sync infrastructure, or cloud storage architectures -- those are IT/infrastructure concerns even if tangentially related to PKM tools --- ## Questions to Ask First Before generating recommendations, gather this diagnostic information. Not every question is needed -- read context and ask only what is genuinely unclear: 1. **Primary output goal** -- What do you produce from your knowledge? (Long-form writing, decisions, presentations, code, research papers, teaching, conversations?) The output type determines the entire system architecture. 2. **Information input type** -- Do you primarily read articles, annotate PDFs, take meeting notes, save bookmarks, capture ideas on the go, or listen to podcasts and audiobooks? 3. **Current system description** -- What does your current system look like, including tools, folder structures, and any workflows? What specifically is breaking down? 4. **Time budget** -- How many minutes per day and per week can you realistically commit to processing and reviewing? (10 min/day vs. 60 min/day changes the right system dramatically) 5. **Professional context** -- Are you a student, researcher, knowledge worker (analyst, strategist, consultant), creative (writer, designer, developer), or manager? Each has a different optimal hybrid. 6. **Friction points** -- Where does your system most often break down? Capture? Processing? Retrieval? Output creation? This reveals whether the problem is structural or habitual. 7. **Tool constraints** -- Do you require local-first storage (privacy, offline access), cross-platform sync, or existing ecosystem lock-in (Apple, Google, Microsoft)? 8. **Collaboration need** -- Is any portion of your knowledge system shared with collaborators, or is it entirely personal? --- ## Process ### Step 1: Diagnose the Current System and Primary Failure Mode - Identify the user's failure mode before recommending solutions. The six most common are: collector's fallacy (saves everything, processes nothing), system hopping (switches tools before any system matures), over-engineering (complex taxonomy with no content), all-capture-no-output (information enters but nothing exits), abandoned review (system degrades without maintenance), and retrieval failure (notes exist but cannot be found when needed). - Ask the user to describe their last five times they tried to retrieve something from their notes. Did they succeed? How long did it take? Did they give up and Google it instead? Retrieval failure is the single most diagnostic symptom of PKM breakdown. - Quantify the inbox. If they have more than 200 unprocessed items, recommend a capture pause -- stop saving new items until the backlog is cleared to under 30. - Identify whether the problem is structural (wrong methodology for their workflow), habitual (right structure, inconsistent execution), or tooling (friction from the wrong tool for their cognitive style). - Determine if the user is a linear processor (prefers sequential, hierarchical organization -- PARA tends to click) or associative thinker (prefers relational, networked organization -- Zettelkasten tends to click). ### Step 2: Select and Configure the Right Methodology - For most users, recommend a **hybrid architecture** rather than a single methodology, because each methodology solves a different problem. The core question is which layer of the stack each methodology serves. - Use this assignment logic: GTD handles the **action layer** (what must I do and when), PARA handles the **organizational layer** (where does this live), Zettelkasten handles the **thinking layer** (what do I actually know), and BASB progressive summarization handles the **reading layer** (how do I extract value from sources). - **For the action layer (GTD core):** Implement five lists -- Inbox (unprocessed), Next Actions (by context: @computer, @phone, @errands, @home, @waiting), Projects list (anything requiring more than one action), Someday/Maybe (aspirational but not committed), and Reference (non-actionable support material). The two-minute rule: if an item takes less than two minutes to act on, do it immediately rather than capturing it. - **For the organizational layer (PARA):** Projects are active efforts with a defined finish line and a deadline. Areas are ongoing responsibilities with no finish line (Health, Finances, Writing, Professional Development). Resources are topics of interest for future reference (Machine Learning, Architecture, Cooking). Archive is everything inactive. The key discipline: every note lives in exactly one PARA folder, determined by the most actionable current use. - **For the thinking layer (Zettelkasten):** Enforce strict atomicity -- one discrete idea per note, not one topic, one chapter, or one source. A note about "cognitive biases" is not atomic; a note about "why availability bias causes managers to over-weight recent performance in evaluations" is atomic. Every permanent note must be written entirely in the author's own words, with no quoted passages longer than two sentences. Connections are created manually by writing explicit link text: not just a bare link but a sentence explaining why the connection matters. - **For the reading layer (Progressive Summarization):** Apply in four passes. Pass 1: Read without highlighting (develop a position first). Pass 2: Highlight the most resonant passages (10-20% of the text). Pass 3: Bold the key phrases within highlights (20-30% of what was highlighted). Pass 4: Write a 2-3 sentence executive summary in your own words at the top of the note. Most notes only need Pass 2. Reserve Pass 3-4 for material directly relevant to active projects. ### Step 3: Design the Capture Workflow - Establish a **single inbox per medium** to minimize cognitive overhead during capture. Recommended: one mobile quick-capture app (Drafts, Apple Notes, or a dedicated Obsidian daily note) for on-the-go thoughts; one web clipping tool (Readwise Reader or Omnivore for open-source) for articles and bookmarks; one annotation tool for PDFs (Zotero with ZotFile, or Readwise's PDF reader); one highlight aggregator (Readwise syncing Kindle, Instapaper, Pocket, and Snipd) feeding into the main PKM tool. - Apply the **RESCUEE test** before capturing anything: Does this Resonate unexpectedly? Could it Enable a current project? Does it Spark a new connection? Is it Counterintuitive or surprising? Will I Use this specific data point? Is it Evidence for or against something I believe? Does it Explain something I didn't understand? If none of these apply, do not capture it. - Capture decisions must take under three seconds. If a capture requires deliberation, the threshold is too high -- lower the bar and raise the processing bar instead. Capture first, evaluate during processing. - For books: read in two passes. First pass for flow; second pass with deliberate highlighting. Use Readwise to surface highlights in spaced repetition. Do not try to create permanent notes during the first read. - For meetings: maintain a standard meeting note template with sections for Date, Attendees, Context, Key Decisions, Action Items (with owner and deadline), and Follow-up Questions. Process within 24 hours while context is fresh. - For podcasts and videos: use Snipd (podcasts) or Glasp/YouTube summary tools for video. Capture a single sentence synthesis immediately after finishing -- the "so what" -- before capturing specific clips. ### Step 4: Build the Processing Routine - **Daily processing (10-20 minutes):** Process every inbox to zero -- not by deleting everything but by deciding the fate of each item. Each item gets one of four outcomes: trash (most should be trashed), reference in PARA (file to the right folder), action (add to Next Actions or Projects), or promote to thinking layer (create a Zettelkasten permanent note). Do not skip the trash step -- the instinct to save everything is the root cause of most PKM failure. - **The 3-question processing protocol:** Ask for every captured item: (1) Is this actionable within 14 days? If yes, add to Next Actions or Projects. (2) Will this be useful for a specific active project or Area? If yes, file to PARA. (3) Does this change, challenge, or extend something I already believe or know? If yes, process into a permanent Zettelkasten note. If none of these are true, trash it. - **Weekly review (45-60 minutes, same day and time each week):** Run a fixed checklist in this order -- Process all inboxes to zero (15 min). Review all active Projects for next actions and stalled items (10 min). Scan all Areas for anything neglected (5 min). Review upcoming calendar for preparation needs (5 min). Process daily notes into permanent Zettelkasten notes (10 min). Set the top three priorities for the next week (5 min). The weekly review must be scheduled and protected as a meeting with yourself -- it is the system's immune system. - **Monthly review (90 minutes, first week of month):** Archive completed projects and move their notes to Archive. Review Someday/Maybe list -- promote 1-2 items to active Projects. Audit information sources using the signal-to-noise score. Scan note connections and create any missing Maps of Content (MOCs). Review progress against quarterly goals. Identify one system improvement to implement. ### Step 5: Design the Retrieval Architecture - Retrieval is the output of the system -- if notes cannot be retrieved quickly, the system has failed regardless of how well-organized the capture is. Design for retrieval from day one. - Implement a **four-layer retrieval stack:** (1) Full-text search as the first resort -- this should return results in under two seconds. (2) Maps of Content (MOCs) as structured entry points into clusters of notes. (3) Tags for cross-cutting properties that don't correspond to folders (status:: draft, type:: concept, domain:: psychology). (4) Graph view or backlinks for serendipitous discovery during creative sessions. - Every permanent note must have a descriptive title that functions as a complete, searchable claim: "Spaced repetition increases long-term retention by 200% compared to massed practice" retrieves better than "Spaced repetition" or "Note on learning." Use noun phrases for concepts, verb phrases for arguments, and question phrases for open problems. - Create a MOC whenever you have 7 or more notes on a related topic. The MOC is not a folder or index -- it is an opinionated synthesis note that explains the structure of a topic and links to constituent notes with a sentence of context for each link. - Tags should have a defined ontology with no more than 20-30 active tags. Recommended tag namespaces: `type::` (concept, evergreen, fleeting, project, meeting, book-note), `status::` (draft, developed, mature), `domain::` (the knowledge domain), `for::` (the output this note serves). Remove tags that have not been searched against in the past 90 days. ### Step 6: Configure the Tool Stack - Select the **primary thinking tool** based on three criteria: link quality (bidirectional links are non-negotiable for Zettelkasten), storage model (local file system for privacy and longevity, or cloud for accessibility), and extensibility (plugin ecosystem for future workflow automation). - **Obsidian** is the recommended default for serious PKM work: local Markdown files (future-proof format), bidirectional links, graph view, Dataview plugin for database-style queries over notes, Templater for standardized note creation, and Zotero integration for academic work. The base product is free; sync costs $4/month or can be replaced with iCloud/Syncthing at no cost. - **Logseq** is the recommended alternative for users who prefer an outliner/block-based model with built-in daily notes as the primary entry point. The entire database is stored as plain text, and the block-level linking is more granular than Obsidian's page-level links. - **Notion** is appropriate when team collaboration is needed alongside personal PKM, when the user needs relational database features (properties, filters, gallery views), or when the user is not yet committed to serious PKM and wants a low-friction starting point. Notion's link graph is weaker than Obsidian's, and the proprietary format creates long-term lock-in risk. - **Capacities** is appropriate for users who think in objects (people, books, projects, concepts) rather than flat notes -- it applies a type-based structure natively without plugins. - Do not recommend using more than three tools in a stack. Every additional tool creates a sync burden, a capture decision point, and a potential point of system failure. A well-configured single tool is almost always better than a three-tool pipeline. ### Step 7: Manage the Information Diet - Audit every active information source using four criteria scored 1-5 each: **Relevance** (how aligned is this to my current projects and Areas), **Uniqueness** (does this source cover ideas I cannot get elsewhere), **Signal density** (ratio of valuable content to filler per unit time), **Actionability** (does consuming this lead to notes, decisions, or outputs). Maximum score is 20. Sources scoring below 10 should be unsubscribed immediately. Sources scoring 10-14 should be moved to a "read occasionally" folder rather than inbox. - Set hard consumption limits before reviewing them: no more than 5 newsletters in active inbox, 3 podcasts in regular rotation, 30 minutes of passive social media browsing, and one news source checked once per day. These limits are not arbitrary -- they are calibrated to leave 80% of cognitive bandwidth available for deep processing and creation. - Apply the **5:1 creation-to-consumption ratio** as a weekly check: for every five items consumed and processed into notes, create at least one output (a permanent note, a published post, a summary sent to a colleague, a revised framework, or a decision documented with reasoning). If the ratio is significantly below 1:5 in a given week, do not consume more until an output is created. - Information channels have different optimal consumption rhythms. Books: one at a time, read to completion before starting the next. Courses: one at a time, complete the module before capture. Articles: batched into one reading session of 45-60 minutes rather than consumed throughout the day. Podcasts: listened during low-cognitive-load activity (commuting, walking) with a quick capture immediately after. ### Step 8: Establish Feedback Loops and System Evolution - Schedule a **90-day system review** (separate from the monthly review) to evaluate whether the current tool and methodology still fit the user's workflow. The right time to switch tools is after 90 days of committed use, when specific limitations are visible -- not when curiosity about a new tool peaks. - Track a single diagnostic metric monthly: **retrieval success rate**. When you need information, does your PKM yield it within 60 seconds? Log 10 retrieval attempts per month and score each as successful, partial, or failed. A system performing below 70% success rate needs structural repair. A system performing above 90% is healthy. - The system should evolve in one direction: **toward simplicity over time**, not complexity. Add a new structural element (a tag, a MOC, a template) only when you have felt the absence of that element at least three times. Remove any structural element that has not been used in 90 days. - Identify the highest-leverage habit to add at each maturity stage: Stage 1 (first 30 days) -- daily capture and inbox processing. Stage 2 (days 30-90) -- weekly review. Stage 3 (days 90-180) -- creating permanent notes from reading. Stage 4 (beyond 180 days) -- producing regular output and maintaining MOCs. --- ## Output Format When delivering a PKM recommendation, structure the output as follows: ``` ## PKM System Design: [User Name / Context] ### Diagnosis **Primary Failure Mode:** [One of: collector's fallacy / system hopping / over-engineering / all-capture-no-output / abandoned review / retrieval failure] **Secondary Issues:** [1-2 additional observations] **Current System Strengths:** [What is already working -- always acknowledge this] --- ### Recommended Methodology Stack | Layer | Methodology | Purpose | Tool | |--------------|-----------------------|------------------------------------|-------------------| | Action | GTD | Tasks, projects, commitments | [Tool] | | Organization | PARA | File structure and folder logic | [Tool] | | Thinking | Zettelkasten | Permanent knowledge and synthesis | [Tool] | | Reading | Progressive Summarization | Extract value from sources | [Tool] | **Rationale:** [2-3 sentences explaining why this specific hybrid was chosen for this user] --- ### Tool Stack **Primary PKM tool:** [Tool name] -- [reason specific to this user's constraints] **Capture tools:** [list by context: mobile, web, PDF, meetings] **Sync/backup approach:** [specific recommendation] **Tools to remove or consolidate:** [what to stop using and why] --- ### PARA Structure for This User ``` Projects/ [3-5 example active projects relevant to the user's context] Areas/ [4-6 example areas relevant to the user's role] Resources/ [4-6 example resource topics the user mentioned] Archive/ [Note: everything completed or inactive moves here] ``` ### Note Templates **Permanent Note Template:** ``` Title: [Specific claim or concept as a complete phrase] Created: [date] Tags: [type::concept] [domain::X] [status::draft] --- [Main idea in 1-3 paragraphs, entirely in your own words] **Source:** [Author, title, year -- no raw URL] **Connected to:** [[Note A]] because [explicit reason] | [[Note B]] because [explicit reason] **Questions this raises:** [1-2 open questions for further investigation] ``` **Meeting Note Template:** ``` Meeting: [Topic] | Date: [date] | Attendees: [names] Context: [1 sentence on why this meeting happened] --- **Key Decisions:** - [Decision 1 -- who owns it] **Action Items:** - [ ] [Action] -- [Owner] -- [Deadline] **Follow-up Questions:** - [Question that needs resolution] **Linked to:** [[Project]] [[Area]] ``` --- ### Daily Processing Routine (Time budget: [X] minutes/day) **Morning (5 min):** [Specific steps for this user] **During day:** [Specific capture habits for this user's context] **Evening (10-15 min):** [Specific processing steps] **3-Question Processing Protocol (apply to every inbox item):** 1. Actionable within 14 days? → Next Actions / Projects 2. Useful for active Project or Area? → File in PARA 3. Changes or extends what I know? → Create permanent note None of the above? → Trash immediately. --- ### Weekly Review Checklist (Day: [day], Time: [time], Duration: [X] min) - [ ] Process all capture inboxes to zero - [ ] Review all active Projects -- confirm next action exists for each - [ ] Scan all Areas -- anything neglected or at risk? - [ ] Review upcoming calendar -- what needs preparation? - [ ] Promote 2-3 daily notes into permanent Zettelkasten notes - [ ] Set top 3 priorities for next week - [ ] Update any stalled MOCs --- ### Information Diet Plan **Sources to keep (score 15+):** [List with score and consumption cadence] **Sources to reduce (score 10-14):** [List with specific reduction plan] **Sources to cut (score below 10):** [List -- unsubscribe this week] **Consumption limits:** - Newsletters: [N] in active inbox - Podcasts: [N] in rotation - Social: [X] min/day - Books: 1 at a time --- ### 30/60/90 Day Implementation Plan **Days 1-30 (Foundation):** - [ ] Set up [Primary Tool] with PARA folder structure - [ ] Configure capture stack (mobile + web) - [ ] Process existing inbox down to under 30 items - [ ] Complete 3 weekly reviews - [ ] Create first 10 permanent notes **Days 30-60 (Habit formation):** - [ ] Daily processing streak of 20+ consecutive days - [ ] Create first Map of Content on primary topic - [ ] Audit and cut information sources - [ ] Complete first monthly review **Days 60-90 (Output and retrieval):** - [ ] Produce one substantial output from PKM notes (post, report, presentation) - [ ] Achieve 70%+ retrieval success rate - [ ] Evaluate tool fit -- document specific limitations if any - [ ] Decide to commit or migrate based on evidence --- ### Failure Mode Prevention **Your highest risk failure mode:** [specific mode] -- here is how to prevent it: [2-3 specific preventive habits or structural guardrails] **System health check (run monthly):** - Inbox items: [target: under 30] - Active permanent notes: [target: growing by 5+ per week] - Last weekly review: [target: less than 8 days ago] - Retrieval success rate: [target: 70%+] - Outputs created this month: [target: at least N] ``` --- ## Rules 1. **Never recommend a single monolithic methodology.** PARA, Zettelkasten, GTD, and BASB solve different problems at different layers of a knowledge workflow. A recommendation of "just use Zettelkasten" or "just use PARA" signals incomplete domain knowledge. Every professional user needs at least a two-layer hybrid (organization + thinking), and most benefit from all four layers. 2. **Diagnose the failure mode before prescribing a solution.** A user experiencing collector's fallacy needs stricter capture filters and more aggressive processing habits -- adding Zettelkasten structure will make it worse. A user with retrieval failure needs better note titles, more MOCs, and stronger linking habits -- adding a new tool will not fix it. Match the solution to the actual problem. 3. **Tool recommendations must be constrained to the user's actual constraints.** A recommendation of Obsidian for someone who must share notes with a team or requires Windows/iOS cross-platform sync without paying for Obsidian Sync must acknowledge the friction or recommend alternatives. Local-first tools require a sync solution; never leave this unspecified. 4. **Progressive Summarization is a reading workflow, not a PKM architecture.** Do not treat progressive summarization as equivalent to the Zettelkasten permanent note creation step. Summarized highlights are literature notes. Permanent notes require the user to write original synthesis in their own words. Conflating these two produces a system that captures summaries but never generates original thinking. 5. **Atomic notes must encode a claim, not just a topic.** A note titled "Productivity" or "Climate Change" is not atomic -- it is a folder disguised as a note. An atomic note title must contain a verb or a specific relationship: "Remote work increases deep focus time but reduces ambient collaboration for engineers." If a user is creating topic notes, redirect them to the correct Zettelkasten pattern before the system accumulates unfixable structural debt. 6. **Weekly review is not optional -- it is the system's circulatory system.** A PKM system without a weekly review will degrade within 30-60 days. Projects lose next actions, inboxes overflow, and note connections atrophy. If a user cannot commit to 45 minutes per week, recommend the Minimum Viable PKM (single inbox, daily 10-minute processing, zero Zettelkasten until the habit is stable) rather than a full system that will be abandoned. 7. **The 90-day commitment rule is absolute.** Never recommend switching PKM tools to a user who has been using their current tool for fewer than 90 days. The discomfort of a new system during months 1-2 is not a signal that the tool is wrong -- it is the normal learning curve. Tool-switching before 90 days is system hopping in disguise and is a primary cause of PKM failure. 8. **Information diet must be actively pruned, not passively managed.** An information diet of more than 8 newsletters, 5 podcasts, and multiple daily news sources will overwhelm any PKM system regardless of how well the processing routine is designed. Recommend cutting sources before adding any new system features. More information entering the system is always a liability, not an asset, until the processing habit is stable. 9. **Tags must be defined before use, not after.** A tagging system built by tagging notes as they are created will produce dozens of inconsistent tags within 60 days. Before creating the first tag, define the tag ontology: namespace, meaning, and when to apply each tag. If a user already has 50+ inconsistent tags, do not try to clean them up -- declare tag bankruptcy, create a new ontology with 15-20 tags, and re-tag only the 20% of notes accessed most frequently. 10. **Retrieval is the only metric that matters.** The purpose of a PKM system is not beautiful organization, impressive note counts, a satisfying graph view, or a complete set of book highlights. It is the ability to retrieve the right information at the moment it is needed. Every structural decision -- folder depth, tag count, note title format, linking density -- should be evaluated against a single question: does this make the right note easier or harder to find in 90 days? --- ## Edge Cases ### Edge Case 1: The Collector's Backlog (500+ unprocessed items) A user arrives with thousands of saved bookmarks, years of highlights in Readwise, and hundreds of notes that have never been processed. Do not recommend processing everything -- that path leads to abandonment. Recommend a **triage protocol**: declare everything older than 90 days as Archive and move it to a folder named `_backlog`. Do not delete it, but stop treating it as a processing obligation. Set a rule: if a specific item from the backlog is needed, retrieve it then. Otherwise, focus the system entirely on items captured in the last 90 days. Begin the new system clean. This feels like losing work but is actually the only viable path forward. ### Edge Case 2: The Over-Engineered System (Complex Taxonomy, Empty Notes) A user has built a sophisticated Notion workspace with 40 database properties, 15 tags, 8 folder levels, and custom formulas -- but has fewer than 50 substantive notes and has never produced an output from the system. The problem is optimization before use. Prescribe a forced simplification: collapse everything to a single flat folder with PARA as the only top-level structure, remove all tags except three (type::, domain::, status::), and delete every template that has never been used. Then assign the user a concrete output goal -- "write one 500-word piece from your notes by the end of the month" -- and rebuild structure only where the output process reveals a genuine need. ### Edge Case 3: The Multi-Context User (Work PKM + Personal PKM) A user with high-security corporate work obligations needs to maintain a strict separation between work and personal knowledge. Do not recommend a unified system -- the compliance and confidentiality risk is real. Recommend separate vaults or workspaces with a lightweight bridge: personal Obsidian vault for all personal knowledge, employer-approved tool (Notion, OneNote, SharePoint-compatible) for work notes, and a "concept extraction" habit where generic insights learned at work (frameworks, mental models, process improvements) are transcribed in general terms into the personal system. Never recommend syncing corporate data into personal cloud storage. ### Edge Case 4: The Academic Researcher (Citation-Heavy Workflow) A PhD student or academic needs to manage hundreds of papers, citations, and source-linked notes in a way that supports literature reviews and academic writing. This requires a specialized academic stack layer: Zotero as the citation manager (with ZotFile for PDF management and the BetterBibTeX plugin for Obsidian integration), and the Obsidian Citations plugin to create literature notes automatically from Zotero entries. The Zettelkasten layer still applies -- literature notes (what the paper argues) are distinct from permanent notes (what the researcher concludes) -- but the literature note creation pipeline is automated rather than manual. Every permanent note in this context must include a `@citationKey` reference to its Zotero source. ### Edge Case 5: The AI-Augmented PKM User A technically capable user wants to add AI retrieval to their Obsidian vault (semantic search, local LLM Q&A over notes). Recommended architecture: use the Obsidian Smart Connections plugin for semantic similarity search within the vault (runs locally, privacy-safe). For users comfortable with Python, a local RAG pipeline using LlamaIndex or LangChain ingesting the vault's Markdown files into a vector database (Chroma or Qdrant) with a local LLM (Ollama with Mistral or Llama 3) provides natural language Q&A without cloud exposure. Critical warning: AI retrieval augments but does not replace manual linking and MOC creation. A vault of poorly titled, poorly connected notes will return poor AI results -- garbage in, garbage out. The prerequisite for AI augmentation is a healthy manual system. ### Edge Case 6: The Minimal Time User (Under 15 Minutes Per Day) A user who genuinely cannot commit more than 10-15 minutes per day to PKM should not attempt Zettelkasten, a full GTD implementation, or any multi-tool capture stack. Prescribe the **Minimum Viable PKM**: one tool (Apple Notes or Obsidian), one daily note as the single inbox, 10 minutes of processing per day using the 3-question protocol, and one 30-minute weekly review. No tags, no elaborate folder structure, no MOCs until the daily habit is stable for 60 consecutive days. The first 60 days are exclusively about building the review and capture habit -- complexity comes only after the habit exists. ### Edge Case 7: The Domain-Switching Knowledge Worker (Multiple Unrelated Fields) A consultant, executive, or polymath who works across 4-6 unrelated knowledge domains simultaneously faces a unique challenge: their note graph will have disconnected clusters with few cross-domain links, and broad tags quickly become meaningless. The recommended structure is **nested PARA per domain** at the Resources level: Resources/Machine-Learning/, Resources/Organizational-Behavior/, Resources/Finance/. Each domain folder contains its own lightweight MOC as an entry point. Cross-domain insight notes (notes that bridge two domains) get a special tag `type::synthesis` and are linked from both domain MOCs. These synthesis notes are often the most valuable intellectual assets the user produces and should be reviewed monthly. ### Edge Case 8: The Note Migration User (Switching from Notion to Obsidian) A user migrating from Notion has several specific challenges: Notion's export produces Markdown files with broken formatting, images stored in Notion that do not export cleanly, and relational database properties that have no equivalent in flat Markdown files. Recommended migration protocol: use the Notion to Obsidian (NotionToObsidian) community converter rather than native Notion export. Plan for a two-week migration window, not a one-hour switch. Migrate in PARA order -- Projects first (highest urgency), then Areas, then Resources, then Archive. Accept that database views and filters will be lost and must be rebuilt with Obsidian Dataview queries. Move image storage to an external folder (Obsidian supports a designated attachments folder) and re-link after migration. Do not attempt the migration during a high-workload period. --- ## Example **User Input:** > "I'm a content strategist at a tech company. I read about 10 articles per day, take notes in a Google Doc that's now 80 pages long, have 1,200 unread Pocket saves, and a Notion workspace I set up two years ago that I basically never use. I'm trying to write a newsletter on AI and marketing but I can never find what I've already researched when I'm writing. My biggest frustration is sitting down to write and having to re-research everything from scratch." --- **Output:** ## PKM System Design: Content Strategist / AI & Marketing Newsletter ### Diagnosis **Primary Failure Mode:** Retrieval failure -- the system captures information but cannot surface it during the output (writing) phase. The 80-page Google Doc and 1,200 Pocket saves confirm high-volume capture with essentially zero retrieval architecture. **Secondary Issues:** (1) Capture is distributed across too many tools with no processing bridge -- Pocket, Google Docs, and Notion each hold information that never converges. (2) No consistent processing routine means captured information never becomes working knowledge. **Current System Strengths:** You are clearly an active, high-volume reader with strong source instincts. The Google Doc shows you have been capturing observations consistently -- this is valuable raw material, not wasted effort. --- ### Recommended Methodology Stack | Layer | Methodology | Purpose | Tool | |--------------|-----------------------|--------------------------------------------|-------------------------| | Organization | PARA | Structure notes around newsletter outputs | Obsidian | | Reading | Progressive Summarization | Extract signal from 10 articles/day | Readwise Reader | | Thinking | Zettelkasten | Build permanent knowledge on AI + marketing | Obsidian | | Action | GTD (lightweight) | Manage newsletter issues as projects | Obsidian (Tasks plugin) | **Rationale:** Your core problem is output-oriented -- you write a newsletter and cannot find your research when you need it. PARA should be structured around your newsletter topics and active issues as Projects. The Zettelkasten layer solves the specific frustration you described: permanent notes with descriptive titles are searchable by claim, so when you sit down to write about AI and personalization, a full-text search returns every note you have ever made on that specific intersection. Readwise Reader replaces Pocket and eliminates the 1,200-item backlog problem permanently. --- ### Tool Stack **Primary PKM tool:** Obsidian -- local Markdown files, excellent full-text search, and Dataview queries let you pull all notes tagged for a specific newsletter topic into a single view when writing. **Capture tools:** - Web articles: Readwise Reader (replaces Pocket -- highlights sync automatically to Obsidian via the Readwise Official plugin) - Mobile quick-capture: Drafts app with an action to append to your Obsidian daily note - Research rabbit holes: Obsidian Quick Note shortcut directly into daily note inbox - Meetings and interviews: Obsidian meeting template (below) **Sync:** Obsidian Sync ($4/month) or iCloud -- required for iPhone/Mac continuity. **Tools to stop using:** - Google Doc (migrate top 20 most-referenced sections to Obsidian permanent notes this week; archive the rest) - Pocket (cancel the subscription; export what you need, import into Readwise Reader, delete the rest) - Notion (archive the workspace; do not migrate it -- the structure is wrong for this workflow and migration cost exceeds value) --- ### PARA Structure for This User ``` Projects/ Newsletter Issue 47 - AI in Programmatic Advertising/ Newsletter Issue 48 - LLM Content Personalization/ Q3 Content Calendar/ Client: [Company Name] Campaign Brief/ Areas/ Newsletter Operations/ Professional Development/ Health/ Finance/ Resources/ AI -- Foundation Models and APIs/ Marketing -- Measurement and Attribution/ AI -- Personalization and Recommendation Systems/ Content Strategy -- Frameworks/ Marketing -- B2B SaaS/ Archive/ [All completed newsletter issues, past client projects] ``` ### Note Templates **Permanent Note Template (for AI + Marketing insights):** ``` Title: [Specific claim -- e.g., "First-party data outperforms modeled audiences in conversion rate by 3x in cookieless environments"] Created: 2024-03-15 Tags: [type::concept] [domain::ai-marketing] [status::draft] Source topic: AI personalization | Newsletter relevance: Issue 48 --- [2-4 paragraphs synthesizing this idea entirely in your own words. No pasted quotes. Explain the mechanism, the evidence, and the implication for marketers.] **Source:** [Author, Publication, Year] **Connected to:** [[First-party data deprecation pressures DTC brands]] because both notes deal with signal degradation post-cookie. [[Probabilistic identity graphs]] because this is the alternative approach being evaluated. **Questions this raises:** How does this hold for B2B intent data vs. B2C behavioral data? Does this apply equally to email vs. display? ``` **Newsletter Research Note (for collecting before writing):** ``` Newsletter Issue: [Number + Topic] Target publish date: [date] --- **Core argument I'm making:** [1-2 sentences -- your thesis before you write] **Supporting permanent notes:** - [[Note title 1]] -- why this supports the argument - [[Note title 2]] -- why this supports the argument **Counterarguments to address:** - [Counterargument 1] **Sources I still need:** - [Gap 1] **Draft status:** [ ] Outline | [ ] First draft | [ ] Edited | [ ] Published ``` --- ### Daily Processing Routine (Time budget: 20 minutes/day) **Morning (5 min):** Open Obsidian daily note. Write down any overnight ideas or half-formed connections in the inbox section. Check Readwise Reader -- review any new highlights from yesterday's saved articles (Readwise surfaces these automatically). Do not process now -- just review. **During day (zero added time):** Save articles to Readwise Reader instead of Pocket. Highlight 3-5 passages maximum per article while reading -- apply the RESCUEE filter before highlighting. Use Drafts to send quick thoughts to the Obsidian daily note. **Evening (15 min):** Apply the 3-question processing protocol to every item in today's daily note: 1. Actionable within 14 days? → Add to Newsletter Issue project or Task list 2. Useful for specific active newsletter issue or Area? → File in PARA 3. Does this change or extend what I believe about AI or marketing? → Write a permanent note On Tuesdays (newsletter writing day), run a Dataview query: `FROM #domain::ai-marketing WHERE status = "draft"` to surface all relevant permanent notes before writing. You should never need to re-research from scratch -- the query does it for you. **3-Question Processing Protocol (apply to every inbox item):** 1. Actionable within 14 days? → Next Actions / Projects 2. Useful for active Project or Area? → File in PARA 3. Changes or extends what I know? → Create permanent note None of the above? → Trash immediately. --- ### Weekly Review Checklist (Day: Friday, Time: 4:00 PM, Duration: 50 min) - [ ] Process Readwise Reader inbox -- archive read articles, promote highlights to permanent notes for the 2-3 best pieces - [ ] Process Obsidian daily notes inbox to zero - [ ] Review all active Newsletter Issues -- does each have a research note with at least 5 linked permanent notes? - [ ] Scan Areas -- anything in Newsletter Operations falling behind? - [ ] Review upcoming calendar -- any interviews, events, or launches to prep for? - [ ] Promote 2-3 daily notes into permanent Zettelkasten notes (prioritize anything that appeared in multiple daily notes this week -- recurrence signals importance) - [ ] Set top 3 priorities for next week (usually: Issue drafting, Issue research, 1 admin task) - [ ] Update the AI + Marketing MOC if any new permanent notes were created --- ### Information Diet Plan **Apply the scoring audit to your current sources this week.** **Example scoring for a typical content strategist at this volume:** | Source | Relevance | Uniqueness | Signal | Actionability | Score | Action | |--------|-----------|------------|--------|---------------|-------|--------| | MIT Tech Review | 5 | 4 | 3 | 3 | 15 | Keep -- weekly batch read | | Marketing Week | 4 | 3 | 3 | 4 | 14 | Keep -- reduce to 2x/week | | Generic AI hype newsletter | 2 | 1 | 2 | 2 | 7 | Cut immediately | | Podcast: Practical AI | 5 | 4 | 4 | 3 | 16 | Keep -- walk listen + Snipd | | Twitter/X AI accounts | 3 | 2 | 2 | 2 | 9 | Cut or max 15 min/day | **Consumption limits for your workflow:** - Newsletters: Maximum 5 in active Readwise Reader inbox - Podcasts: Maximum 3 in rotation (use Snipd to capture clips that auto-sync to Readwise) - Articles: Read in one 45-minute batch at 2:00 PM, not throughout the day - Books: 1 at a time -- currently relevant to AI or marketing only - Social: 20 minutes maximum, only after daily processing is complete **First action:** Export your Pocket library, identify the last 50 items you actually read (not just saved), and cancel. Everything else in the 1,200-item backlog is the sunk cost fallacy -- let it go. --- ### 30/60/90 Day Implementation Plan **Days 1-30 (Foundation -- stop the bleeding, build the base):** - [ ] Install Obsidian and create PARA folder structure (2 hours, Day 1) - [ ] Install plugins: Readwise Official, Dataview, Templater, Tasks (1 hour, Day 1) - [ ] Set up Readwise Reader and migrate from Pocket -- save the export but stop processing the backlog - [ ] Migrate the top 20 most-referenced sections of the Google Doc into permanent notes (5 notes/week over 4 weeks) - [ ] Complete 4 weekly reviews (non-negotiable -- schedule them now as calendar blocks) - [ ] Create 20 permanent notes (5 per week) -- these will form the kernel of your newsletter research vault **Days 30-60 (Habit formation -- retrieval starts working):** - [ ] Daily processing habit stable (20 min/day, fewer than 3 missed days per week) - [ ] Create your first AI + Marketing MOC with links to all permanent notes created so far - [ ] Write one newsletter issue entirely from PKM notes -- measure time saved vs. previous process - [ ] Audit and cut information sources using the scoring table above - [ ] Complete first monthly review **Days 60-90 (Output and retrieval -- the system proves its value):** - [ ] Permanent note count: 60+ notes - [ ] Retrieval test: sit down to write a newsletter issue -- can you produce an outline in under 15 minutes using only PKM search and MOC? Target: yes. - [ ] Produce one piece of content that goes beyond the newsletter (a thread, a short report, or a guest post) sourced entirely from PKM permanent notes - [ ] Evaluate Obsidian fit -- document any specific limitations. Make a data-driven keep-or-migrate decision. --- ### Failure Mode Prevention **Your highest risk failure mode:** Collector's fallacy -- your history with 1,200 unread Pocket saves and an 80-page unstructured Google Doc shows a strong capture instinct with an underdeveloped processing instinct. The new system is designed to make processing non-optional. **Preventive habits:** 1. The Readwise Reader inbox cap: never let saved-but-unread articles exceed 20. When you hit 20, process before saving more. This is a hard stop, not a soft guideline. 2. The permanent note gate: before you are allowed to save a new research note to your PARA Projects folder, you must have written at least one permanent note from your processing session. No note, no save. This forces the conversion from raw input to original thinking. 3. The pre-writing ritual: every Tuesday (newsletter writing day) begins with a Dataview query, not a Google search. If you catch yourself opening a browser to research something you feel like you have already read, check the PKM first. Building this habit is the fastest path to experiencing the system's value. **System health check (run first Monday of each month):** - Readwise Reader unread inbox: target under 20 - Obsidian inbox items (unprocessed daily notes): target under 10 - Active permanent notes created this month: target 20+ - Last weekly review completed: target under 8 days ago - Newsletter issues with complete research notes: target 100% of scheduled issues - Outputs created this month: target minimum 4 newsletter issues + 1 additional piece