Turn Books, Articles and Notes into Structured Knowledge with Super Intelligence

How to turn books, articles and notes into structured knowledge with Super Intelligence begins by separating collection from understanding. A folder full of PDFs, highlights and summaries is not yet a knowledge system. A useful knowledge system tells you what the important ideas are, how they connect, which claims are supported, which questions remain open and what you can retrieve or apply without reopening the source every time.

SI can accelerate this transformation because it can extract structure, compare sources, normalise terminology and generate retrieval tasks. But if the process loses provenance or replaces reading with generic summaries, the result becomes a polished archive rather than usable knowledge.

In the eduKateSG life series, Super Intelligence, or SI, is our editorial name for practical AI assistance. This article follows How to Learn Through Socratic Questioning with Super Intelligence and connects with active recall, knowledge-gap detection and learning roadmaps.

The central rule is: structure the knowledge around concepts, claims, relationships and questions—not around the order in which the source happened to present them.

The Difference Between Information and Structured Knowledge

Information is content available to you. Structured knowledge is content organised so you can retrieve, compare, update and apply it.

A book chapter may contain twenty pages of examples, explanation and narrative. The knowledge system may need only several core concepts, their relationships, a handful of important claims, two unresolved questions and the examples that make the model memorable.

The purpose of structuring is not to shrink every source aggressively. It is to separate durable structure from presentation.

The Structured Knowledge Spine

Use eight fields:

source;
purpose;
concepts;
claims;
relationships;
evidence;
questions;
applications.

Source

Where did the information come from? Preserve title, author, date, URL or document identity where relevant.

Purpose

Why are you reading it? Research, learning, decision, writing, reference or project execution?

Concepts

Which ideas organise the material?

Claims

What does the source assert?

Relationships

How do the concepts and claims connect? Cause, contrast, dependency, hierarchy, example, condition, sequence?

Evidence

What supports the important claims?

Questions

What remains uncertain, disputed or worth investigating?

Applications

Where could the knowledge change an action, explanation, decision or project?

Start with Reading Intent

The same source should be structured differently depending on why you are reading it.

If you are studying for an exam, extract concepts, definitions, mechanisms and likely transfer tasks.

If you are researching an article, extract claims, evidence, limitations and disagreements.

If you are learning a professional domain, extract vocabulary, workflows, risks, decision criteria and examples.

If you are making a decision, extract facts, assumptions, options, trade-offs and unknowns.

Ask SI to restate the reading purpose before summarising. This prevents generic summaries.

The Source Card

Title: __.
Author/organisation: __.
Date: __.
Source type: book, article, note, report, documentation.
Purpose: why I am using this.
Scope: what the source actually covers.
Authority: why it matters.
Freshness: when this source may need rechecking.

The source card keeps knowledge attached to provenance.

Concept Extraction

A concept is not every important noun.

A useful concept organises several facts or relationships.

In economics, concepts such as scarcity, incentives and equilibrium organise many examples.

In biology, structure-function, energy and regulation connect many topics.

In project management, scope, dependency, risk, ownership and change are reusable concepts.

Ask SI: “Which concepts recur across this source and organise multiple sections? Distinguish core concepts from supporting terminology.”

Concept Definitions Should Preserve the Source

Store the source-defined meaning where terminology matters, then add your own plain-language explanation separately.

This avoids a common problem where repeated paraphrasing changes the definition over time.

For technical, legal, scientific or syllabus material, keep the formal source meaning visible.

Claim Extraction

A claim is a proposition that could be supported, challenged or qualified.

Do not convert every sentence into a claim.

Extract the load-bearing claims: those that support the source’s argument, explain a mechanism or influence your decision.

Ask SI to classify claims as fact, inference, opinion, prediction or recommendation using the framework from How to Separate Facts, Assumptions and Opinions with Super Intelligence.

The Claim Card

Claim: exact proposition.
Type: fact, inference, opinion, prediction, recommendation.
Evidence: what supports it.
Scope: population, time, condition or context.
Confidence: confirmed, supported, plausible, disputed or unresolved.
Use: why this claim matters to your current work.

Relationship Extraction

Knowledge becomes powerful when relationships become visible.

Useful relationship types include:

A causes or contributes to B;
A depends on B;
A is a subtype of B;
A contrasts with B;
A is evidence for B;
A is an example of B;
A occurs before B;
A limits B;
A is a condition for B;
A changes when B changes.

Ask SI to build a small concept graph using only relationships the source supports.

Do Not Let SI Invent Edges

A concept graph can look authoritative even when the model invents relationships not present in the source.

Mark each edge as source-stated, strongly inferred or your own hypothesis.

Important inferred relationships should be checked before becoming durable knowledge.

Evidence Extraction

When a source makes an important claim, extract the kind of evidence supporting it.

Data, experiment, historical record, case study, citation, logical argument, expert judgement or anecdote.

Do not store the conclusion without the evidence type if you later need to evaluate or compare it.

Ask whether the evidence supports the full scope of the claim.

The Evidence Boundary

One study can support a result under its conditions without proving a universal rule.

A case study can illustrate mechanism without establishing frequency.

An expert opinion can inform judgement without becoming direct measurement.

Structure the limitation beside the claim instead of leaving it in the forgotten endnotes.

Question Extraction

A source is valuable partly because of the questions it creates.

Ask SI to identify:

questions the source answers;
questions the source raises;
questions it avoids;
claims that need verification;
conflicts with other sources;
applications worth testing.

These questions become a research queue or learning queue.

Question Status

Label questions as open, answered, disputed, low-priority or out-of-scope.

The knowledge system should not retain every curiosity forever.

Turning Highlights into Knowledge

Highlights are signals, not finished notes.

For each important highlight, ask one of four questions:

What concept does this illustrate?
What claim does this support?
What relationship does this reveal?
What question does this create?

If a highlight fits none of those roles, it may not deserve permanent storage.

The Highlight Compression Rule

Do not preserve long quoted passages when a short claim and source reference will do.

Keep the minimum quotation necessary when exact wording matters.

This makes the knowledge base easier to navigate and respects source boundaries.

Turning Notes into Atomic Knowledge

Atomic knowledge means each note has one clear purpose.

One concept, one claim, one question, one example or one decision rule.

Atomic does not mean tiny for its own sake. It means the unit can be linked and reused without dragging an entire chapter summary with it.

SI can split long notes into candidate atomic units, but the user should decide which units are worth keeping.

Avoid Note Fragmentation

Over-atomising creates hundreds of tiny notes with no navigational value.

Keep related details together when they are always used together.

Use concept pages or synthesis notes to reconnect the atoms.

The Concept Page

A concept page can contain:

Definition.
Plain-language explanation.
Mechanism.
Examples.
Contrasts.
Boundary cases.
Sources.
Questions.
Applications.

This becomes a durable knowledge object that can be updated as new sources arrive.

The Synthesis Note

A synthesis note combines several sources around one question.

It should separate agreement, disagreement, scope differences, evidence quality and unresolved issues.

SI is especially useful for synthesis when provenance is preserved.

Worked Example: Turning a Book Chapter into Structured Knowledge

Suppose you read a chapter on decision-making under uncertainty.

Instead of a chapter summary, extract:

Concepts: uncertainty, expected value, reversibility, option value.

Claims: reversible decisions require less information than irreversible decisions.

Evidence: examples or cited research supporting the claim.

Relationships: reversibility lowers the value of exhaustive pre-decision analysis.

Questions: when does delay destroy option value?

Application: career experiments, purchases, workflow pilots.

The knowledge is now reusable outside the chapter.

Worked Example: Turning an Article into Structured Knowledge

Article topic: a new educational intervention.

Source card records publication date, population and study type.

Claim card records the measured outcome and scope.

Evidence field records method and sample.

Limitation field records whether transfer, retention or only immediate performance was measured.

Question queue records what would be needed before applying the result to another age group or school system.

This is stronger than saving “study says method works”.

Worked Example: Turning Meeting Notes into Knowledge

Meeting notes contain decisions, proposals, facts, assumptions and unresolved questions.

SI can separate them into status-labelled units:

Decision: approved action.

Proposal: not yet approved.

Fact: verified information used in discussion.

Assumption: working belief.

Question: unresolved dependency.

Owner/deadline: execution information.

This prevents a proposal from later being remembered as an approved decision.

Worked Example: Turning Class Notes into Knowledge

For school or university notes, convert each topic into concept, mechanism, formula/rule, example, misconception and retrieval question.

Then link prerequisite concepts and create transfer questions.

Revision can now operate on the structured knowledge rather than rereading full notes.

Knowledge Compression

Compression should preserve what matters for future retrieval and use.

A useful three-level system is:

One sentence: central idea.
One paragraph: mechanism and main condition.
One page: concepts, evidence, boundaries, examples and applications.

SI can generate all three, but compare them for consistency.

The Compression Loss Test

Ask what was lost when the source became a summary.

Did uncertainty disappear? Did a conditional claim become universal? Did evidence type disappear? Did examples become rules?

Restore load-bearing details.

Knowledge Retrieval

A knowledge base is useful only if you can find and use its contents.

Organise around concepts, questions, projects and decisions—not only source titles.

A source can belong to several concept pages. A concept can appear in several projects.

SI can help generate tags and links, but keep the vocabulary stable enough that retrieval does not fragment across synonyms.

Canonical Terms

Choose one preferred term for important concepts and record aliases.

This makes search and linking more reliable.

Example: “active recall” as the canonical term, with “retrieval practice” recorded as a related or alternative term depending on context.

Knowledge and Active Recall

Turn structured knowledge into retrieval prompts.

Concept page → define, explain mechanism, give example, state boundary.

Claim card → recall claim, evidence and limitation.

Relationship → explain why A connects to B.

Synthesis note → compare sources from memory before rereading.

This connects the knowledge base to active recall and revision.

Knowledge and Projects

The strongest knowledge base is used in real work.

Link concepts and claims to active projects, essays, decisions or systems.

When a project uses a concept, record the application and what reality taught you.

This turns knowledge from static reference into tested understanding.

The Knowledge Update Rule

Knowledge should change when sources change.

Mark facts as current, superseded or stale.

Update claim confidence when evidence changes.

Add new source disagreement instead of silently overwriting the old note when the history matters.

SI can compare new sources with existing structured knowledge and identify material changes.

The Knowledge Deletion Rule

Not everything deserves permanent storage.

Delete or archive duplicate notes, low-value highlights, outdated facts and summaries whose source can be retrieved easily.

The knowledge base should become more useful, not merely larger.

A Copyable Source-to-Knowledge Prompt

“Turn this source into structured knowledge rather than a generic summary. Preserve source identity and scope. Extract core concepts, load-bearing claims, relationships, evidence, limitations, unresolved questions and practical applications. Label claims as fact, inference, opinion, prediction or recommendation where relevant. Separate source-stated relationships from your inferences. Create a concept map, a small set of atomic notes and retrieval questions. Do not add unsupported facts. End by stating what important context would be lost if I kept only your summary.”

Knowledge-System Failure Modes

Summary archive

Everything becomes a summary. Repair: extract concepts, claims and relationships.

Lost provenance

Notes no longer show where claims came from. Repair: source cards and claim links.

Over-atomisation

Thousands of tiny notes. Repair: concept pages and synthesis.

Tag explosion

Many synonyms fragment retrieval. Repair: canonical vocabulary.

No retrieval

Knowledge is stored but never recalled. Repair: generate active questions.

No application

Notes remain detached from real work. Repair: link to projects.

No pruning

Archive becomes cluttered. Repair: deletion and simplification reviews.

The Structured-Knowledge Quality Audit

Can I trace important claims to their source?
Are core concepts separated from supporting detail?
Are relationships explicit?
Are evidence and limitations visible?
Are unresolved questions tracked?
Can I retrieve by concept or project, not only source title?
Can the notes produce active-recall prompts?
Are stale facts marked?
Are duplicate or low-value notes pruned?
Can I use the knowledge in a real decision or output?

The Knowledge Graph

A structured knowledge system becomes more useful when important concepts are connected through explicit relationships.

A knowledge graph does not need specialised software. It can be a set of concept pages linked by statements such as “depends on”, “contrasts with”, “causes”, “is evidence for”, “is an example of” or “is a subtype of”.

The advantage is that one source can contribute to several concepts, and one concept can collect evidence from several sources.

Ask SI to propose links, then label each as source-stated, strongly inferred or tentative. This protects the graph from becoming a map of fluent guesses.

The Edge Audit

Every important relationship is an “edge” in the graph. Audit the edges that drive conclusions.

Does the source actually say A causes B, or only that they are associated? Does B depend on A, or are they simply often taught together? Is the contrast real or editorial?

When the edge is uncertain, store it as a question rather than a fact.

Source Synthesis: Move Beyond One-Source Summaries

Once several sources exist, the knowledge system should stop thinking in source order and begin thinking in concept and question order.

For each major concept, ask:

Which sources define it?
Which sources provide evidence?
Where do they agree?
Where do they use different populations, methods or assumptions?
Which disagreement is factual and which is interpretive?
Which source is more current or authoritative for this specific claim?

SI can produce the first synthesis matrix quickly, but important discrepancies should be checked against the original sources.

The Synthesis Matrix

Rows: claims or concepts. Columns: Source A, Source B, Source C, agreement, disagreement, scope, evidence quality, unresolved question.

The matrix is useful because it prevents one elegant source summary from dominating simply because it was read first.

When Sources Disagree

Do not force consensus.

Classify the disagreement. The sources may use different definitions, populations, periods, methods, assumptions or value criteria.

Sometimes one source is outdated. Sometimes the evidence is genuinely mixed. Sometimes both are correct within different scopes.

Store the disagreement beside the concept instead of hiding it inside a generic synthesis.

The Contradiction Ledger

For important contradictory claims, keep a short ledger:

Claim A: __.
Source: __.
Claim B: __.
Source: __.
Difference type: factual, scope, method, terminology, interpretation or value.
Resolution: resolved, context-dependent, disputed or open.
Decision impact: what changes if one view is correct.

This turns disagreement into a navigable research object.

The Supersession Rule

When a newer authoritative source replaces an older requirement, mark the old note as superseded rather than leaving both active.

Keep the history only if it helps explain past decisions or version changes.

This is essential for software, law, policy, standards and other time-sensitive knowledge.

The Knowledge State Machine

Knowledge objects can have states.

Captured: source saved but not processed.
Structured: concepts, claims and relationships extracted.
Verified: important claims checked.
Integrated: linked to existing knowledge.
Applied: used in a project, decision or explanation.
Maintained: still current and useful.
Stale: requires source refresh.
Archived: no longer active.

Not every source needs to reach every state. A low-value article may remain captured and later be deleted. A foundational book may become deeply integrated.

The Processing Queue

Do not process every saved item immediately.

Keep a small queue based on relevance to active projects, learning goals or decisions.

SI can help rank the queue, but the system should not create pressure to process everything you collect.

The Capture Rule

Capture should be friction-light and interpretation-light.

When you encounter something potentially useful, save the source plus one sentence explaining why it matters.

Do not force complete synthesis in the moment unless the source is directly relevant to current work.

This prevents note-taking from interrupting reading continuously.

The Processing Rule

Processing should answer: What does this source change in my existing model?

If it adds nothing important, keep only the source reference or discard it.

If it adds a concept, claim, contradiction or useful example, attach that change to the relevant concept page or project.

Turning Notes into Questions

Every durable note can generate a retrieval or research question.

Definition note → “Define this concept and distinguish it from its nearest neighbour.”

Mechanism note → “Explain why A leads to B.”

Claim note → “What evidence supports this and what limits its scope?”

Decision rule → “Under what conditions should this rule be used?”

Question note → “What source or experiment would answer this?”

SI can create these prompts automatically after structuring the source.

The Retrieval Conversion Rule

Do not convert every note into a flashcard.

Use retrieval only for knowledge worth carrying internally. Leave low-value details in reference form if lookup is efficient and safe.

The knowledge system should distinguish memory targets from reference targets.

The Memory–Reference Split

Some knowledge should be memorised because it supports judgement, fluency or prerequisite reasoning.

Other knowledge can remain external because it is detailed, infrequently used or safer to verify fresh.

Memorise core concepts, relationships, frequent procedures and high-leverage vocabulary.

Reference changing prices, long tables, detailed regulations, rare syntax or exact current specifications.

SI can help classify which knowledge belongs in memory versus reference.

The Freshness Advantage of Reference Knowledge

Some knowledge becomes safer when it is looked up rather than memorised indefinitely.

Current requirements, software behaviour and time-sensitive rules are examples.

Store where and how to verify them, not only the current value.

Structured Knowledge for Writing

Writers can use the knowledge system to separate evidence from prose.

Build a claim set, source set, counterargument set and concept glossary before drafting.

Then ask SI to help compose from the structured evidence rather than asking it to invent an article from a topic.

This reduces unsupported claims and makes revisions easier because the underlying knowledge remains separate from the current wording.

The Article Evidence Pack

Primary claim.
Supporting claims.
Sources.
Definitions.
Counterevidence.
Examples.
Open questions.
Internal links.

SI can then draft while the evidence pack remains the source of truth.

Structured Knowledge for Decisions

Decision knowledge should be organised around the decision rather than the sources read.

Keep verified facts, assumptions, options, constraints, stakeholder input, risks and unknowns in separate sections.

Attach sources to the facts they support.

When a fact changes, the affected decision model can be updated without rereading every source.

Structured Knowledge for Projects

Project knowledge should connect concept to action.

For each active project, link relevant decisions, definitions, procedures, risks, source documents and lessons learned.

When the project ends, extract durable lessons into concept or workflow knowledge and archive the rest.

Structured Knowledge from Meetings

Meeting transcripts and notes can create enormous archives with little practical value.

Extract only durable categories:

decisions;
owners and deadlines;
verified facts;
assumptions;
open questions;
changed definitions or policy;
lessons worth carrying forward.

Leave casual discussion in the transcript unless it has a reason to become structured knowledge.

The Meeting-to-Knowledge Rule

Do not turn every conversation into institutional memory.

Promote only information that should influence future action, interpretation or recovery.

Structured Knowledge from PDFs and Reports

Long reports often contain executive claims, methods, data, caveats and appendices.

SI can separate these layers so the summary does not collapse caveats into certainty.

Extract report purpose, population, method, headline findings, limitations, key tables, definitions and decision implications.

Keep page or section references for important findings so the source can be reopened efficiently.

The Table and Figure Rule

If a conclusion depends on a table or figure, store what the visual actually shows, the unit and the relevant source location.

Do not rely only on a prose summary of the visual.

Worked Example: Building a Research Topic Knowledge Base

Suppose the topic is active recall in education.

Concept pages: retrieval practice, spacing, feedback, transfer, metacognition.

Claim cards: specific claims from studies, each with population, method and outcome.

Synthesis note: which findings recur and where methods differ.

Question queue: effects by age, subject, delay and output type.

Application note: how the evidence changes the design of an SI revision system.

The knowledge base can now support writing, tutoring and system design.

Worked Example: Building Product Knowledge

Instead of saving reviews, create fields for requirements, verified specifications, price/date, warranty, constraints, competing products and unresolved questions.

Separate manufacturer facts from reviewer opinions and user anecdotes.

When the purchase is complete, keep only durable lessons or warranty/reference information. Archive the comparison clutter.

Worked Example: Building Professional Knowledge

A new manager studying incident response reads procedures, post-mortems and handbooks.

Concepts: severity, owner, escalation, containment, communication, recovery, post-mortem.

Decision rules: what triggers escalation, who owns communication, when a workaround becomes unsafe.

Cases: real incidents linked to failure patterns.

Retrieval: scenario prompts rather than memorising handbook prose.

The knowledge base becomes operational, not bibliographic.

The Knowledge Review Cadence

Different knowledge needs different review.

Active project knowledge: review weekly or when decisions change.

Learning knowledge: review through active recall and transfer.

Current regulations or software: review when source changes or before consequential use.

Foundational concepts: review mainly through application and occasional retrieval.

Do not schedule universal reviews for every note.

The Knowledge Pruning Review

Every few months, ask:

Which notes are duplicated?
Which facts are stale?
Which sources no longer support current work?
Which concept pages have become too broad?
Which unresolved questions no longer matter?
Which lessons belong in a reusable workflow?

Pruning creates information quality by subtraction.

The Knowledge Continuity Card

For any active domain, maintain a compact continuity card.

Current purpose.
Core concepts.
Trusted sources.
Open questions.
Active project links.
Freshness risks.
Next review.

This gives SI enough context to continue the knowledge work without loading the entire archive.

The Knowledge Ownership Test

Ask whether you can explain the structure without opening the knowledge base.

Can you name the core concepts? State the main claim? Explain the central relationship? Identify the strongest source and the main uncertainty?

If not, the knowledge may be well stored but not yet learned.

A personal knowledge system should support memory and judgement, not become a substitute for having a mental model at all.

Atomic Note Types

Atomic notes become easier to manage when each note has a type.

Concept note: definition, mechanism, examples and boundaries.
Claim note: one proposition plus evidence and scope.
Question note: unresolved issue plus what could answer it.
Decision note: a choice, rationale and review trigger.
Procedure note: repeatable sequence and failure conditions.
Example note: a case illustrating a reusable concept.
Source note: bibliographic or document context for later reference.

The type makes reuse easier because the system knows what the note is for.

The Note Purpose Test

Before keeping a note, ask what future task it supports.

Will it help recall? Support a claim? Guide a decision? Explain a concept? Rebuild a workflow? Preserve a source?

If the answer is unclear, the note may be a temporary capture rather than durable knowledge.

Canonical Vocabulary and Aliases

Knowledge fragments when the same concept appears under many names.

Choose one canonical term for the concept and record aliases, abbreviations or older terminology underneath it.

This is especially useful in technical and interdisciplinary work where different sources use different language for similar ideas.

SI can identify likely duplicates, but merging concepts should happen only after checking that the meanings really align.

The False-Merge Risk

Two sources can use similar words for different concepts or different words for similar concepts.

Do not let automatic semantic similarity collapse distinctions that matter.

Ask SI to compare definitions, scope and usage before proposing a merge.

Claim Confidence and Knowledge Confidence

Not all knowledge objects deserve the same confidence.

A directly verified current requirement may be confirmed. A mechanism supported by several strong sources may be strongly supported. A model connecting several ideas may be plausible but still inferential. A disputed historical interpretation may remain contested.

Use confidence labels sparingly and connect them to evidence rather than intuition.

Confidence Should Be Revisable

New evidence can strengthen or weaken a knowledge object.

Do not preserve an old “confirmed” label when the source has changed, and do not leave a once-speculative claim weak after repeated verification.

A structured knowledge system should make updating confidence easy.

Literature and Source Synthesis

When many sources cover the same question, avoid one-summary-per-source as the final structure.

Instead, build one synthesis organised around the question.

For each claim, record which sources support it, which qualify it and which contradict it.

Then compare differences in method, sample, date, geography and definition.

SI can create the first cross-source map rapidly, but the analyst should inspect the claims that carry the conclusion.

The Evidence Convergence Test

Multiple sources increase confidence only when they provide genuinely different evidence or analysis.

Three articles repeating the same press release do not create three independent confirmations.

Ask SI to trace likely source ancestry where possible and distinguish independent evidence from repeated citation.

The Scope-Conflict Test

Two sources may appear to disagree because one studies a different population, period or condition.

Before creating a “controversy” note, ask whether the claims truly occupy the same scope.

Turning Structured Knowledge into Explanations

A strong explanation can be generated from the concept graph rather than from raw source text.

Pull the definition, mechanism, example, contrast and boundary from the concept page, then adapt the vocabulary to the reader.

This makes explanations more consistent across articles, lessons and conversations.

It also reduces drift because the formal core remains attached to the same structured knowledge object.

Turning Structured Knowledge into Practice

Concepts can generate retrieval questions. Claims can generate evidence questions. Relationships can generate explanation questions. Boundaries can generate counterexamples. Procedures can generate recovery scenarios.

SI can therefore use the same knowledge structure to produce learning material without inventing an unrelated curriculum.

Turning Structured Knowledge into Decisions

When a decision depends on the knowledge base, extract only the relevant current facts, assumptions, uncertainties and decision rules.

Do not load the entire archive into the decision brief.

Structured knowledge is valuable partly because it lets you retrieve the minimum relevant context.

Knowledge Versioning

Some concept pages and decision rules change over time.

Use versioning when the history matters.

Example: v1 based on the old syllabus, v2 after a curriculum revision. v1 project workflow, v2 after a failure revealed a missing verification step.

Do not version every typo. Version material changes to meaning, scope or operational use.

The Change Log

Record what changed, why, which source or evidence triggered it and which dependent notes may need review.

SI can help find downstream references that may now be stale.

Knowledge Decay and Staleness

Some knowledge decays because memory fades. Other knowledge becomes stale because reality changes.

The system should distinguish these.

Memory decay → retrieval practice.

Source decay → re-verification.

Do not spend time memorising a value that should instead be checked fresh before use.

Privacy and Sensitive Knowledge

A personal knowledge system does not need every sensitive detail to be useful.

Use minimum necessary context when structuring personal, family, health, financial, educational or workplace records.

Where possible, store the durable concept or action separately from unnecessary identifying information.

Do not upload confidential documents merely because automation would be convenient.

The Sensitivity Label

For personal systems, label some knowledge as public, private, confidential or restricted according to your real context.

Then use that label to control what may be shared with external tools or collaborators.

Knowledge structure and access control should be designed together when the information is sensitive.

The Knowledge Inbox

Create one place for unprocessed captures.

The inbox can contain book notes, article links, meeting extracts, observations or questions.

Review it on a reasonable cadence and decide: process, archive, delete or defer.

Do not allow the inbox to become a second archive that is never processed.

The Capture-to-Knowledge Funnel

Capture: save source and why it matters.
Triage: decide whether it deserves processing.
Structure: extract concepts, claims, evidence and questions.
Integrate: connect to existing knowledge.
Apply: use in learning, writing, decision or project.
Review: update or prune.

This funnel keeps collection subordinate to use.

The Knowledge Garden Versus the Knowledge Warehouse

A warehouse optimises storage. A knowledge garden optimises growth, connection and pruning.

Use SI to surface related concepts, old questions and stale claims when they become relevant to current work.

Do not browse the archive aimlessly simply because it exists.

The knowledge system should become active when a real question, project or learning goal calls for it.

The Serendipity Rule

Some useful connections are unexpected.

SI can periodically surface one or two concept links across domains, but those suggestions should remain hypotheses until the user sees a real mechanism or application.

Serendipity is useful when bounded; unrestricted association becomes noise.

A 30-Day Structured Knowledge Build

Week 1: choose one active domain, define canonical concepts and create source cards.

Week 2: convert the most important sources into concept and claim notes; create a small knowledge graph.

Week 3: build synthesis notes, contradictions and retrieval prompts.

Week 4: use the knowledge in one real output, prune low-value captures and create a continuity card.

After 30 days, the goal is not a giant vault. It is one domain whose knowledge can be retrieved, updated and applied more easily.

The 90-Day Knowledge Review

At 90 days, inspect the system itself.

Which sources were actually reused? Which concept pages became central? Which tags fragmented retrieval? Which facts became stale? Which questions were resolved? Which notes were never touched?

Archive or delete aggressively where the system has proven a note low-value.

Promote frequently used knowledge into clearer canonical pages or procedures.

Worked Example: Building a Mathematics Knowledge Base

Concept pages: variable, equality, distributive property, factorisation, function, gradient.

Relationship edges: factorisation reverses expansion; equations preserve equality when equivalent operations are applied; gradient connects change in y to change in x.

Misconception notes: sign distribution, cancelling across addition, confusing gradient with intercept.

Retrieval prompts: define, explain, choose method, give counterexample, solve changed application.

Project link: exam revision uses the concept pages only to generate targeted mixed questions, not to reread long notes.

Worked Example: Building a Vocabulary Knowledge Base

Canonical word page: meaning, connotation, register, collocations, near-synonyms, contrast, example sentence and active-use status.

Links connect related words by meaning field and contrast.

Active recall pulls words from recognition into sentence and paragraph production.

Stale or low-value entries can be archived if they are no longer relevant to the learner’s writing goals.

Worked Example: Building a Career-Domain Knowledge Base

A professional entering cybersecurity creates concept pages for threat, vulnerability, control, incident, access and risk.

Source cards distinguish official standards, company policy, current product documentation and general articles.

Case notes capture incidents and lessons.

Decision notes record why one control or process was chosen.

Revision uses scenario questions rather than memorising every policy sentence.

The Knowledge Operating Loop

Source → Structure → Connect → Retrieve → Apply → Review → Update.

If a source never gets structured, it remains reference material. If structured knowledge is never retrieved or applied, it remains an archive. If applied knowledge is never reviewed, lessons and stale assumptions accumulate.

The loop keeps the system alive without requiring constant maintenance of everything.

The Knowledge Minimalism Rule

Keep the smallest knowledge system that reliably helps you learn, think, decide and create.

Do not build taxonomies because they are intellectually satisfying if they do not improve retrieval or use.

Do not keep ten summaries when one synthesis note and the original sources are enough.

Structured knowledge should reduce the cost of thinking later. If maintaining the structure costs more than it returns, simplify it.

Searchability Is Part of Knowledge Quality

Knowledge that cannot be found at the moment of need behaves almost like knowledge that was never stored.

Design retrieval paths around how you are likely to search later.

Search by concept, question, project, decision, person, source or date where relevant.

Do not depend on remembering the exact title of the book or file.

SI can help generate aliases and related terms, but keep one canonical name for each important concept so search results do not fragment unnecessarily.

The Retrieval Path Test

Imagine you need the knowledge six months later.

What would you remember first: the concept, the project, the question or the source?

Make at least one of those paths obvious.

Then test the system occasionally by asking SI to find the note from a natural-language description rather than a perfect keyword.

When Your Own Notes Conflict

Personal notes can conflict because your understanding changed, the source changed or two notes describe different contexts.

Do not resolve the conflict by choosing the newer note automatically.

Ask: Which source did each note rely on? Did the definition change? Did the scope change? Was one note your own inference? Was one written before a later correction?

SI can compare the two notes and propose a reconciliation, but the final update should preserve the history when it explains a meaningful change.

The Personal Supersession Rule

When a newer understanding clearly replaces an older one, mark the old note as superseded and link to the current version.

This is better than leaving two contradictory notes active or deleting the history blindly.

The Reading Pipeline

A sustainable reading system separates discovery, reading, processing and application.

Discovery: decide whether the source is worth attention.
Reading: understand the source in its own structure.
Processing: extract only the durable knowledge.
Integration: connect it to existing concepts and questions.
Application: use it in an output, decision or learning task.

Trying to do all five at once can make reading slow and mechanical.

Progressive Processing

Process sources in proportion to likely value.

Low-value source: title, link and one-line relevance note.

Medium-value source: key concepts, claims and one question.

High-value source: full source card, claim cards, concept links, evidence and applications.

This keeps the knowledge system from treating every article like a textbook.

The Stop-Processing Rule

Stop processing when additional detail will not change recall, reasoning, decision or future use.

SI can always extract more. The user must decide when enough structure exists.

Knowledge from Conversations

Conversations can contain valuable knowledge, but they are noisy sources.

Separate what the person directly stated from your interpretation of what they meant.

For important professional or research conversations, capture the speaker, date, context, direct claim and any uncertainty.

Do not turn a casual remark into an authoritative source simply because SI summarised it elegantly.

Interview-to-Knowledge Workflow

Extract speaker-attributed claims, examples, decision rules, terminology and unresolved questions.

Then compare those claims with documents or other interviews when the information is consequential.

Preserve attribution where perspective matters.

Knowledge from Experience

Your own experience can create knowledge, but experience needs structure too.

Record observation, interpretation, lesson and boundary separately.

Observation: the workflow failed when two inputs arrived late.

Interpretation: the dependency chain is fragile.

Lesson: add an explicit missing-input state and fallback.

Boundary: one incident does not prove all similar workflows will fail the same way.

SI can help convert experience into reusable lessons without overgeneralising it.

The Lesson Card

Situation: what happened.
Observation: what was directly seen.
Interpretation: why you think it happened.
Change: what you will do differently.
Boundary: where the lesson may not apply.
Evidence to watch: what will confirm or reject the lesson later.

From Knowledge Base to Briefing

A structured knowledge base can produce concise briefings because the important material is already separated from source noise.

Ask SI to assemble only the concepts, current facts, open questions and decisions relevant to a particular meeting or task.

The briefing should link back to the underlying source or concept pages rather than becoming another detached copy.

The Briefing Compression Rule

A briefing is a temporary view of the knowledge base, not a replacement for it.

After the event, update the durable knowledge objects instead of storing endless briefings as separate truth sources.

Knowledge Handoffs

When another person needs the knowledge, do not hand them the entire archive.

Give them the current purpose, core concepts, relevant sources, decisions, open questions and operational procedures.

Ask what they need to do, not what you want to show them.

The Handoff Completeness Check

Can the receiver act without asking what the terms mean, which version is current, who owns the decision or where to verify the claim?

If not, the knowledge object may be complete for you but incomplete as a handoff.

Knowledge and Decision Traceability

When a decision uses structured knowledge, record which claims and sources were load-bearing.

If a source is later corrected, you can identify which decisions may need review.

This is much stronger than storing the final decision alone.

Knowledge and Project Traceability

Link project changes to the knowledge that caused them.

Example: a new source changes the risk threshold, so the workflow is updated. The project record should show the source and changed rule.

This preserves organisational learning over time.

Knowledge Security and Least Exposure

Structured knowledge can make sensitive information easier to find, which also makes access design more important.

Store only what is needed for the purpose. Separate confidential detail from general procedure when possible. Use access controls appropriate to the system you are using.

Do not copy sensitive source material into multiple summaries simply for convenience.

The Minimal Disclosure Rule

When asking SI for help, provide the minimum context required to structure the knowledge.

An anonymised extract, abstracted workflow or relevant paragraph may be enough.

More context should be purposeful, not automatic.

The Structured-Knowledge Decision Tree

Is the source relevant? If no, archive or ignore.

Does it add a new concept or claim? If no, link as supporting source or discard.

Does it change an existing claim? If yes, update confidence or contradiction ledger.

Does it matter for active work? If yes, link to project, learning or decision.

Should it be remembered? If yes, create retrieval prompts. If no, preserve verification path.

Is it still current? If uncertain, mark freshness risk.

The Structured-Knowledge Ownership Sentence

For an active domain, complete: “The core concepts are ___; the strongest current sources are ___; the main unresolved question is ___; the knowledge I must remember is ___; the knowledge I should verify fresh is ___.”

If you cannot complete that sentence, the archive may be structured for storage but not yet for thinking.

The Final Knowledge Rule

Do not ask SI to remember everything for you.

Ask it to help you separate what deserves memory, what deserves reference, what deserves a link and what deserves deletion.

The purpose of structured knowledge is not maximum retention of information. It is reliable access to the concepts and evidence that improve future learning, reasoning and action.

The Knowledge Contradiction Resolution Workflow

When two durable notes conflict, use a fixed sequence rather than editing whichever one looks older.

1. Compare the exact claims. Are they actually contradictory?
2. Compare scope. Same population, time period and conditions?
3. Compare source authority. Which source is primary or current?
4. Compare evidence. What supports each statement?
5. Compare terminology. Are they using the same words differently?
6. Decide status. resolved, context-dependent, disputed or superseded.
7. Update dependent notes. Which explanations, decisions or workflows relied on the old claim?

SI can perform the comparison quickly, but important contradictions should still be checked against the original source material.

The Dependent-Knowledge Audit

When a foundational concept or current requirement changes, inspect what depends on it.

A changed definition can affect explanations. A changed software API can affect procedures. A changed policy can affect checklists. A corrected data claim can affect an article or decision.

Structured knowledge becomes powerful because the links make those dependencies visible.

The Memory-versus-Reference Audit

Every few months, reconsider what deserves internal memory.

A concept that once seemed like reference material may become core because your role changed. A detail you once memorised may now be safer to verify fresh because the underlying standard changes frequently.

Ask SI to classify knowledge by frequency of use, decision consequence, stability and lookup cost.

High-frequency, stable, judgement-relevant knowledge usually deserves stronger internalisation. Low-frequency, unstable detail often belongs in reference form.

The Lookup-Cost Rule

Externalising knowledge is safest when lookup is quick, authoritative and available at the moment of need.

If a critical fact is difficult to retrieve during the task, more internal memory or a better local reference may be justified.

The Knowledge Lifecycle

Useful knowledge often moves through a lifecycle.

Discover: encounter the source.
Capture: save it with relevance.
Structure: extract concepts, claims and relationships.
Integrate: connect to existing knowledge.
Internalise: retrieve what matters.
Apply: use it in work, learning or decisions.
Review: update from reality.
Retire: archive, supersede or delete when no longer useful.

SI can assist at every stage, but the user decides which knowledge deserves to continue through the lifecycle.

The Knowledge Exit Rule

A knowledge object can leave the active system when it is obsolete, low-value, fully replaced, safely externalised or no longer relevant to the user’s goals.

Do not preserve everything simply because storing digital information is cheap.

The Structured Knowledge Stress Test

Test the system with four scenarios.

New source arrives: can you see what it changes?
Old fact becomes stale: can you find dependent notes?
New project begins: can you retrieve the relevant concepts quickly?
Another person needs a handoff: can you give them current knowledge without the entire archive?

If the answer is no, the structure may be optimised for storage rather than use.

The Final Structured-Knowledge Checklist

Every important claim has provenance.
Core concepts have canonical names.
Important relationships are explicit.
Disagreements are preserved rather than flattened.
Memory targets are separated from reference targets.
Current projects can retrieve relevant knowledge quickly.
Stale material has a refresh path.
Low-value material is pruned.
Structured notes can generate retrieval and application.
The user still understands the mental model without reading the whole archive.

The Final Knowledge Ownership Rule

When SI structures a source for you, review the resulting model in your own words. State the core concepts, main claim, strongest evidence, important limitation and one application.

If you cannot do that, the source may have been processed by the system without being processed by you.

The best knowledge system makes external information easier to think with while still leaving the human capable of understanding what the system contains and why it matters.

The Knowledge Maintenance Budget

Every knowledge system has a maintenance cost: reviewing sources, resolving duplicates, refreshing stale facts, pruning notes and reconnecting old material to current work.

Set a budget for that maintenance. If the system needs constant housekeeping to remain usable, simplify the structure, reduce capture volume or archive inactive domains.

SI should lower maintenance cost by detecting duplicates, surfacing stale claims and proposing merges. It should not justify collecting more merely because organisation can be automated.

The Knowledge Use Ratio

Periodically compare what you stored with what you actually used.

If hundreds of processed notes rarely support learning, writing, decisions or projects, the system may be optimising capture rather than value.

Increase the use ratio by processing fewer sources more deeply, connecting notes to active outputs and deleting low-value material sooner.

The Final Knowledge Maintenance Rule

Knowledge deserves ongoing maintenance only when it continues to improve understanding, retrieval, decision quality or action.

Store less than you can store, structure only what matters, and maintain only what remains alive in your learning and work.

The Knowledge Reuse Test

A structured note earns its place when it can be reused in more than the moment that created it. Test whether the concept page, claim card or procedure can support a new explanation, question, project, decision or revision session without reopening the entire original source.

If the note works only as a reminder that you once read something, it may need more structure—or it may not deserve permanent storage.

The Recontextualisation Rule

When reusing knowledge in a new domain, preserve its original scope. A concept learned in one context may transfer, but the evidence supporting it may not transfer automatically.

Ask SI to separate the reusable principle from the context-specific evidence. This lets the system create useful cross-domain connections without pretending that one source proves more than it does.

The Knowledge-to-Action Test

For every active domain, identify at least one way the structured knowledge changes behaviour. It may improve a lesson, a decision, a workflow, an article, a project, a study plan or a conversation with an expert.

If the knowledge never changes understanding or action, decide whether it belongs in reference storage rather than in the active system.

The final proof of structured knowledge is not that it looks organised. It is that the organisation reduces the effort required to think, learn and act well the next time the subject matters.

The Knowledge Friction Test

Watch where using the knowledge system still feels slow. Do you struggle to find the source, understand the note, see which version is current, connect a concept to a project or decide what deserves revision? That friction reveals the next structural improvement.

Do not solve every friction by adding metadata. Sometimes the better repair is deleting duplicate notes, choosing one canonical concept page, reducing source count or writing a clearer synthesis.

SI can help diagnose the friction by tracing the path from question to source to note to action. The strongest repair is the smallest change that makes that path reliable.

A structured knowledge system should feel easier to use as it matures. If it becomes harder to navigate than the original sources, the structure needs simplification.

The Final Reuse Rule: when a note is used again, improve the underlying durable knowledge rather than creating another detached summary. Repeated use should make the canonical concept, claim or procedure clearer, better sourced and easier to retrieve for the next task.

Structured knowledge should compound: each reuse should make future retrieval, explanation and action slightly easier than before.

Frequently Asked Questions

Should I summarise every book I read?

No. Structure the sources connected to goals, projects, learning or decisions. Many books can remain lightly noted.

What is the difference between a summary and structured knowledge?

A summary compresses one source. Structured knowledge separates reusable concepts, claims, relationships, evidence and questions so they can connect across sources.

Can SI read my notes and build the structure?

Yes, when the notes or source are available. Review important claims and preserve provenance.

How many tags should I use?

Use a small stable vocabulary. Prefer concept links and project links over uncontrolled tag growth.

Should I keep highlights?

Keep only highlights whose wording, evidence or example has future value. Convert the rest into claims or concepts with source references.

How do I know what to delete?

Delete or archive items that are duplicated, stale, low-value, easily recoverable or unrelated to current and likely future use.

What comes next?

The next article turns these practices into a lifelong learning system that can continue across subjects, careers and changing goals.

Helpful Reading

Turn Sources into a System You Can Think With

Books, articles and notes are inputs.

Structured knowledge is the reusable layer that survives after the source is closed.

Preserve provenance. Extract concepts. Separate claims. Map relationships. Keep evidence and uncertainty visible. Turn the structure into retrieval and application.

Super Intelligence becomes most valuable when it helps convert information abundance into a smaller, clearer system you can actually think with.

How to Turn Books, Articles and Notes into Structured Knowledge with Super Intelligence