
How do you read long articles quickly with Super Intelligence without losing the argument? You do not begin with a blind summary. You first map the article: title, author, publication, date, section structure, central question, thesis, evidence types and conclusion. Then you use SI to compress only after the structure is visible.
This approach is faster because it separates navigation from understanding. A long article can often be mapped in minutes, allowing you to decide which sections deserve close reading. SI can help explain difficult passages, compare claims, build evidence tables and produce concise notes, while the original article remains the source of truth.
This eduKateSG guide teaches a practical long-article reading workflow for students, researchers and professionals. It follows How to Find Reliable Sources With Super Intelligence in Stage 3 of the How to Learn Super Intelligence Quickly curriculum.
Terminology: SI is our editorial term for practical contemporary AI learning. The goal is faster comprehension and evidence handling, not replacing the source with a generated summary.
The First Principle: Map Before You Summarise
A summary made before the reader understands the article’s structure can flatten the argument. Important qualifications disappear, examples are mistaken for central claims and the conclusion is disconnected from the evidence.
Mapping answers: What is this article trying to do? How is it organised? Which sections contain the evidence that matters? Which parts are background?
Once the map exists, SI can help compress the article without treating every paragraph as equally important.
Step 1 — Identify the Article
Record title, author, publication, date and URL. These details matter because source reliability and freshness are part of comprehension.
If the article is undated or authorship is unclear, note that before investing heavily in interpretation.
A reading note without source identity becomes difficult to reuse or cite later.
Step 2 — Read the Title and Standfirst Carefully
Titles frame the reader’s expectations. A headline may be stronger than the article itself. The standfirst or opening summary may reveal the intended thesis.
Do not assume the headline is the article’s precise conclusion. Keep it as a claim to test against the body.
SI can help reformulate the headline as a research question: “What evidence does the article offer for this claim?”
Step 3 — Scan the Headings
Headings reveal the architecture. List them before reading closely. Group them into background, evidence, counterargument, implications and conclusion where appropriate.
If the article has no headings, ask SI to propose a provisional section map from paragraph topics, then check it yourself.
The map becomes the navigation layer for later questions.
Step 4 — Identify the Central Question
Ask what problem the article is trying to answer. Some articles state it directly; others imply it.
Write the question in one sentence. If you cannot, you may still be reading details without understanding the article’s purpose.
The central question becomes the anchor for deciding which sections are relevant.
Step 5 — Identify the Thesis
The thesis is the article’s main answer or argument. It may be explicit or emerge gradually.
Write it in your own words, then locate the passages that justify your interpretation.
Do not ask SI to invent a thesis that the article itself does not support.
Step 6 — Identify Evidence Types
Long articles can contain data, quotations, case studies, historical examples, expert opinion, documents or logical arguments.
Classify the evidence because different types support different claims. One anecdote can illustrate an experience without establishing prevalence. One statistic may establish scale without explaining mechanism.
Evidence-type mapping makes later evaluation faster.
Step 7 — Identify the Article’s Structure
A useful structural outline can contain Background → Problem → Evidence → Counterargument → Conclusion. Not every article follows this pattern, but the idea is to identify functional roles.
This reduces the risk of summarising a counterargument as if it were the author’s own position.
SI can label section roles after you provide the text or relevant excerpts.
Step 8 — Read the Introduction and Conclusion Together
Reading opening and conclusion together reveals whether the article delivered what it promised.
Compare the initial question or claim with the final conclusion. Did the scope narrow? Did the author introduce important qualifications?
This comparison gives you a rapid overview before deep reading.
Step 9 — Choose the High-Value Sections
Not every section deserves equal attention. If your research question concerns methodology, the methods and evidence sections deserve deeper reading than general background.
Mark sections as Must Read, Useful Context or Optional for Current Task.
This is selective reading based on purpose, not careless skimming.
Step 10 — Read One High-Value Section Closely
Close reading means tracking claims, evidence and qualifications. Ask what each paragraph contributes to the article’s argument.
If a sentence is difficult, ask SI to explain the specific concept while preserving the original wording nearby for comparison.
Do not replace close reading with a summary of the section when the section contains the evidence you need.
Step 11 — Build a Claim–Evidence Table
For important articles, use columns Claim, Evidence, Source Within Article, Qualification and Your Note.
This table reveals whether the argument is evidence-heavy or assertion-heavy. It also makes later synthesis across articles much easier.
Claims without evidence can be flagged for further checking.
Step 12 — Mark Definitions
Long analytical articles often depend on key definitions. Record how the author uses important terms.
Two articles can appear to disagree because they define “success”, “AI literacy” or “productivity” differently.
Definition notes improve both comprehension and later comparison.
Step 13 — Mark Assumptions
Ask what must be true for the argument to work. Some assumptions are stated; others are implicit.
Examples include assumptions about causality, representative samples, rational behaviour or stable technology.
Use SI to propose assumptions, but check whether they actually matter to the argument.
Step 14 — Mark Qualifications
Words such as may, suggests, under these conditions and for this population are often crucial. Summaries frequently erase them.
Highlight qualifications in your reading notes. They define the operating envelope of the claim.
A fast reader who preserves qualifications can be more accurate than a slow reader who remembers only the headline.
Step 15 — Identify Counterarguments
Strong articles may present opposing views. Distinguish the author’s summary of a counterargument from endorsement of it.
Record how the author responds and whether the response is evidence-based.
This prevents simplified summaries from turning a contested argument into false consensus.
Step 16 — Identify What the Article Does Not Establish
Ask what conclusion would be too strong. If the article reports correlation, it may not establish causation. If it studies one region, it may not generalise globally.
This negative boundary is one of the best checks against over-summarisation.
Write one sentence beginning: “This article does not by itself establish…”
Step 17 — Create a One-Paragraph Compression
Only after mapping the article should you write a concise summary. Include question, thesis, main evidence and key limitation.
A strong one-paragraph compression allows a future reader to decide whether the article deserves reopening.
Keep source identity attached to the note.
Step 18 — Create a One-Line Research Note
For a source register, compress further: “Useful for X because it provides Y evidence; limited by Z.”
This is not a replacement for the article. It is an index into your research library.
The one-line note saves enormous time when revisiting dozens of sources.
The Three-Pass Reading Method
Pass 1 — Orientation
Read title, metadata, opening, conclusion and headings. Produce a map and decide relevance.
Pass 2 — Evidence
Read high-value sections closely. Extract claims, evidence, definitions and qualifications.
Pass 3 — Integration
Create the summary, research note and links to your question or other sources.
Three passes prevent the reader from spending equal effort on every word.
The Five-Minute Article Triage
When deciding whether to read deeply, spend a few minutes on metadata, headings, opening, conclusion and source links.
Ask: Is this relevant? Is the source credible enough for my purpose? Does it contain evidence I cannot get more directly elsewhere?
If no, archive or reject it. Fast reading includes fast rejection.
The Ten-Minute Evidence Scan
For a moderately important article, locate numbers, named studies, documents, quotations and methodological descriptions.
Use browser find or SI-assisted search within the supplied text to locate key terms.
The goal is to identify where evidence lives before reading every transition and example.
Close Reading Still Matters
SI does not eliminate the need for close reading when wording, evidence or interpretation matters.
Legal language, technical methods, definitions and disputed claims often require reading the original passage carefully.
The efficient reader uses SI to decide where close reading earns the most value.
Ask SI for a Structural Map, Not Just a Summary
A useful instruction is: “Map this article by section. For each section, state its role in the argument, key claim and evidence type. Do not merge sections.”
This produces a navigational aid. You can then open the most important section directly.
Check the map against the article headings and text.
Ask SI for an Argument Map
An argument map can show Main Claim, Supporting Claims, Evidence, Counterarguments and Qualifications.
This structure is especially useful for essays and opinion-analysis pieces.
The map should distinguish the author’s argument from quoted or opposing positions.
Ask SI for an Evidence Inventory
Instead of “summarise”, ask: “List every empirical statistic, named study, primary document and direct quotation used as evidence.”
The inventory helps you evaluate whether the article’s thesis rests on strong evidence or mostly interpretation.
Open the relevant source links when important.
Ask SI for Definitions
Ask for the key terms the article defines or relies on. Then compare those definitions with the source text.
This is useful when reading technical or theoretical material.
Definitions can become glossary entries in a larger research project.
Ask SI for Qualifications
Prompt: “List language that limits the scope of the article’s claims: may, suggests, in this sample, under these conditions and similar qualifiers.”
This directly protects against summary inflation.
The output should link or quote only short relevant fragments, not reproduce large portions of the article.
Ask SI for Contradictions Within the Article
Long articles sometimes contain tension between sections. A strong conclusion may exceed cautious evidence earlier.
Ask whether any claims appear inconsistent and then verify the passages yourself.
Internal consistency is part of reading comprehension.
Ask SI What You Should Read Closely
Provide your research question and article structure. Ask which sections most directly answer the question and why.
This makes reading purpose-driven.
Do not accept the recommendation blindly; skim the surrounding sections to preserve context.
Reading News Features
News features often combine reporting, interviews, data and narrative. Separate direct reporting from interpretation and quoted claims.
Check dates and whether later updates exist.
Narrative detail can be useful context without being central evidence.
Reading Opinion Articles
Identify thesis, premises, evidence and rhetorical strategy. Opinion status does not make the article useless; it changes what the source can support.
Use opinion for arguments and perspectives, not as sole evidence for contested factual claims.
Distinguish assertion from sourced fact.
Reading Academic Articles
For academic work, begin with abstract, research question, methods, results, limitations and conclusion.
The methods and results deserve close reading when the study matters to your research. Do not rely only on the abstract or discussion.
Article 25 later in the series focuses on books; Article 26 will cover PDFs as files and document structures.
Reading Technical Documentation
Technical docs are task-oriented. Identify product/version, prerequisites, supported behaviour, examples and limitations.
Read the exact section matching your implementation. Old snippets from search can become obsolete.
Test documentation against the current environment where appropriate.
Reading Policy Articles
Separate current rule, rationale, implementation guidance and commentary. Identify jurisdiction and effective date.
If the article summarises official policy, trace the rule to the official source.
Policy commentary can be useful without being the authority itself.
Reading Investigative Journalism
Track primary documents, interviews, anonymous-source claims and reporter interpretation.
Strong investigations may contain evidence unavailable elsewhere. Note which claims are document-backed and which rely on confidential sourcing.
Later corrections or legal responses may matter for high-consequence use.
Reading Vendor Articles
Vendor articles can provide deep product knowledge and useful methods. They also have commercial incentives.
Use them for product explanations and first-party claims. Seek independent evidence for comparative superiority or broad impact.
Label the source role in your notes.
Reading Community Posts
Community posts often contain one person’s experience. Extract the situation, conditions and outcome rather than generalising immediately.
Search for similar independent experiences before identifying a recurring pattern.
The article or post may be valuable precisely because it reveals edge cases official documentation misses.
Reading Long Articles for Exams or Study
Students should not let SI replace comprehension. Use it to preview structure and clarify difficult paragraphs, then explain the argument independently.
Create retrieval questions from the article and answer without looking at the summary.
This turns fast reading into learning rather than temporary recognition.
Reading Long Articles for Work
Professionals often need decisions, risks and action items rather than complete literary understanding.
Map the article, read the sections tied to the business question and produce a source-backed brief.
Preserve any limitation that would change the decision.
Reading Long Articles for Research
Research reading should produce reusable notes: research question, source role, main finding, evidence, limitation and relevance.
These notes make multi-source synthesis dramatically faster.
Do not save only generic summaries; save why the article matters.
The Article Reading Card
- Source.
- Date.
- Central question.
- Thesis.
- Structure.
- Key evidence.
- Definitions.
- Qualifications.
- Counterarguments.
- Limitations.
- Relevance to my task.
- What I must verify.
- One-line research note.
The Article Relevance Scorecard Without a Score
Ask whether the article directly addresses your question, contains useful evidence, comes from an appropriate source, is current enough and adds something not already covered by stronger sources.
Do not total these mechanically. One decisive mismatch can make an article background-only.
The point is faster selection, not ranking for its own sake.
Summaries Can Hallucinate Structure
When an article is only partially available, SI may infer sections or claims from limited text. Confirm what content the system actually received.
Do not ask for a complete summary of an article that the system cannot access fully unless you provide the text appropriately.
Access should be verified before comprehension is claimed.
Summary Compression Can Remove Caveats
Short summaries naturally drop detail. Protect the caveats that change interpretation.
Ask for “main conclusion plus the two most important limitations” instead of only “summarise in three bullets”.
Compression should preserve decision-relevant uncertainty.
Summary Compression Can Merge Sources
If SI has background knowledge or multiple pages in context, it may introduce information not present in the target article.
For source-specific reading, say “use only this article” and verify key sentences against the text.
Keep article analysis separate from broader synthesis.
Summary Compression Can Lose Argument Order
Some arguments depend on sequence: definition, evidence, exception, conclusion. A flat bullet list can erase that logic.
Use an argument map when order matters.
The best compression matches the structure of the source.
Article Reading Failure Mode 1 — Headline Reading
The reader accepts the headline without checking the article. Repair by identifying thesis and evidence.
Failure Mode 2 — Summary Dependence
The generated summary becomes the only text read. Repair by close-reading high-value sections and checking quotes or data.
Failure Mode 3 — Equal Attention
Every paragraph receives the same effort. Repair with purpose-driven triage.
Failure Mode 4 — Qualification Loss
May becomes is, association becomes cause. Repair by explicit qualifier extraction.
Failure Mode 5 — Source Forgetting
Notes survive without author, date or URL. Repair with metadata-first note-taking.
Failure Mode 6 — Argument Flattening
Counterarguments and evidence become one undifferentiated summary. Repair with structural or argument maps.
Failure Mode 7 — No Transfer
The reader understands while looking at SI’s explanation but cannot explain the article independently. Repair with retrieval and self-explanation.
A Practice Lab: Map an Article in Five Minutes
Choose a long article. Record metadata, headings, central question, provisional thesis and three evidence types.
Then compare your map with SI’s structural map. Resolve differences by checking the article.
The exercise trains navigation rather than passive summary consumption.
A Practice Lab: Build a Claim–Evidence Table
Select five important claims. For each, record evidence and qualification.
Mark any claim that lacks visible support.
This creates a reading artifact useful for later research.
A Practice Lab: Write Three Levels of Compression
Create a one-paragraph summary, a five-bullet brief and a one-line research note.
Compare what disappears at each level. Restore any qualification whose loss changes meaning.
This teaches controlled compression.
A Practice Lab: Explain Without the Summary
Close the SI output and explain the article in your own words: question, thesis, evidence and limitation.
Return to the source where memory is uncertain.
This tests actual comprehension rather than recognition.
A Practice Lab: Read for One Question
Take the same article and define one narrow research question. Identify only the sections that answer it.
Create a targeted evidence note rather than a whole-article summary.
This demonstrates how reading purpose changes reading depth.
Article Reading Maintenance
Research notes should be refreshed when the article changes, is corrected or becomes superseded.
If the article is evergreen, keep the note stable. If it supports a current claim, record the date and revisit when necessary.
Good notes reduce future rereading without pretending sources never change.
Frequently Asked Questions
Can SI summarise a long article for me?
Often yes when the article content is available, but the summary should be treated as a reading aid. Verify important claims and qualifications against the original.
Should I read the whole article first?
Not always. Map and triage first, then read high-value sections closely. Read the whole article when full context is important to your task.
How do I know which sections matter?
Use your research question. Look for sections containing the article’s central evidence, methods, definitions or conclusions relevant to that question.
Can I cite an SI summary?
Cite the original source rather than the generated summary when the source is what supports the claim.
How do I avoid losing nuance?
Preserve qualifications, definitions, counterarguments and limitations in your reading card.
What if the article is behind a paywall?
Do not pretend to have read inaccessible text. Use legally available abstracts, author versions or alternative sources and state the access limitation.
What comes next?
Continue through Stage 3 with books, PDFs, multiple-source comparison and knowledge-base building in the complete SI learning hub.
Argument Reconstruction
For difficult essays and analytical articles, reconstruct the argument rather than merely summarising paragraphs. Identify the conclusion, then ask which premises are needed to support it.
A useful structure is Main Claim → Supporting Claim → Evidence → Assumption → Qualification. This reveals whether the article’s logic is cumulative, conditional or dependent on one contested premise.
Argument reconstruction is especially useful when an article contains many examples. It helps separate illustrative material from the claims doing the real argumentative work.
Read for Logical Connectors
Words such as therefore, however, because, although, nevertheless, for example and by contrast signal the relationship between ideas.
SI can help identify these transitions and label their function, but the reader should inspect the surrounding sentences because one connector can change the direction of the argument.
Fast reading becomes more accurate when you track logical movement rather than only keywords.
Read for Evidence Density
Some sections contain many citations, statistics or documented examples. Others are interpretive or narrative. Mark evidence-dense sections for closer reading when your purpose is research.
A high citation count does not automatically mean high quality, but evidence density helps you decide where verification effort should be concentrated.
If the central thesis rests mostly on assertion despite many citations elsewhere, that matters to your evaluation.
Read for Claim Strength
Classify important claims by strength: descriptive, correlational, causal, predictive, normative or prescriptive.
Different claim types require different evidence. A descriptive statistic can support what happened. A causal claim needs stronger method. A recommendation includes values and trade-offs beyond factual description.
This classification prevents a summary from converting a descriptive article into a prescriptive conclusion.
Read for Scope Markers
Look for words that define who, where and when the claim applies: in this sample, among surveyed teachers, during 2025, in Singapore, under laboratory conditions.
Scope markers are often small phrases with large implications. Preserve them in notes.
When comparing articles, scope markers help you decide whether findings are genuinely comparable.
Read for Definitions That Shift
Some long articles redefine a term gradually. “Intelligence”, “literacy”, “productivity” or “success” may change meaning between sections.
Record the author’s initial definition and any later refinement. A moving definition can be legitimate, but it must remain visible.
SI can help collect every explicit definition or use of the key term for comparison.
Read for Examples Versus Evidence
Examples make arguments concrete, but they are not always evidence. A hypothetical scenario can illustrate a principle without showing that it happens frequently in the real world.
Label examples as hypothetical, anecdotal, case-based or empirical. Do not let a memorable story carry more evidentiary weight than the source warrants.
This distinction is especially important in opinion and business writing.
Read for Hidden Assumptions
Ask what the article takes for granted. A productivity article may assume output is measurable. An education article may assume examination performance is the primary goal.
Hidden assumptions can determine the conclusion before the evidence is considered.
Write one or two assumptions in your reading card and ask what would change if they were false.
Read for Incentives and Perspective
Consider who wrote the article and why. A company blog, advocacy essay, academic paper and investigative report can all be useful while serving different institutional purposes.
Perspective does not make a source invalid. It helps you understand why some evidence is emphasised and other evidence may be absent.
This belongs in source evaluation as well as reading comprehension.
Read for Missing Voices
Long articles can present a coherent narrative while excluding relevant stakeholders or evidence. Ask whose perspective is represented and whose is absent.
For policy, this may include implementers and affected populations. For technology, users and operators may reveal issues absent from vendor discussion.
Missing voices are a prompt for further research, not automatic proof that the article is wrong.
Read for Unanswered Questions
As you read, maintain a short list of questions the article raises but does not answer. These may become future search queries or comparison points.
This keeps reading active. The article becomes part of a research conversation rather than a closed authority.
An unanswered question can be more valuable than another summary sentence.
Read for Source Lineage
When the article cites other sources, note which ones appear central to the argument. A recurring citation may be the real evidence source worth opening.
Follow the most important lineage rather than every link. This converts reading into targeted research.
The article becomes a map to stronger evidence instead of the final destination.
The Compression Ladder
- Full source.
- Section map.
- Claim–evidence table.
- One-paragraph summary.
- Five-bullet brief.
- One-line research note.
- Keyword or tag set.
Each level loses detail. The reader should know which information must survive each compression step.
Do not jump from full source directly to one-line note before understanding the article’s structure.
Compression Invariants
Define what must survive compression: central question, thesis, decisive evidence, qualification and limitation.
Optional examples and rhetorical transitions can often disappear. Scope-changing caveats should not.
A good summary is not the shortest possible version. It is the shortest version that preserves the information relevant to the reader’s task.
Copyright-Safe Summarisation
When using SI to work with published articles, summarise in your own words and keep quotations brief and purpose-specific. Do not ask the system to reproduce the article or reconstruct large portions of protected text.
For research notes, paraphrase the argument and preserve source citations. Quote only when exact wording matters, such as a definition or precise policy phrase.
The goal is comprehension and analysis, not creating a substitute copy of the source.
Reading an Article You Disagree With
Do not begin by searching for flaws. First reconstruct the strongest version of the argument: thesis, evidence and qualifications.
Then evaluate sources, assumptions and counterevidence. This prevents disagreement from becoming misrepresentation.
SI can help generate questions, but your summary should be recognisable to a fair-minded reader of the original article.
Reading an Article You Agree With
Agreement creates a different risk: accepting weak evidence because the conclusion feels right.
Ask the same questions you would ask of an opposing source. What evidence supports the claim? What is the population? What would falsify the conclusion?
Symmetrical scrutiny improves research quality.
Reading Multiple Articles on the Same Topic
Use a common note schema so the articles can be compared: Question, Thesis, Method or Evidence, Population, Date, Limitation and Relevance.
Do not write each note in a completely different format. Consistency reduces cognitive load during synthesis.
Article 27 later in the series develops multi-source comparison in detail.
Article Notes as a Knowledge Base
A strong reading note should be reusable months later. Include enough source identity and evidence context to understand why you saved it.
Tags can help group notes by topic, method, population or source role. Avoid creating so many tags that retrieval becomes harder.
The note is an index into the source, not a replacement for it.
Reading for a Presentation
If the final deliverable is a presentation, identify the one or two claims from the article that advance the narrative. Record the evidence needed to support those claims and the source citation.
Do not turn every section of the article into a slide. The presentation has its own receiver and objective.
Keep backup notes for questions that may arise during discussion.
Reading for a Decision Memo
Decision reading focuses on evidence that changes the choice: expected effects, risks, costs, uncertainty and implementation conditions.
Extract the parts relevant to the decision criteria and preserve limitations that could reverse the recommendation.
Background narrative can be compressed aggressively if it does not affect the decision.
Reading for a Literature Review
Literature-review reading requires more method detail. Record research question, design, sample, outcome, effect and limitation.
The article’s role in the literature also matters: foundational, replication, critique, synthesis or new evidence.
A generic summary is insufficient because later comparison depends on these structured details.
Reading for Examination Revision
For study, convert the article into questions after reading. Ask for definitions, mechanisms, evidence and counterarguments, then answer without the source open.
Use SI to generate retrieval prompts, not only summaries. The learner’s unaided answers show whether comprehension survived compression.
Return to the original article for correction rather than trusting the generated answer key automatically.
Reading for Vocabulary and Language
If the goal is language development, identify unfamiliar words from context before asking SI for definitions. Record the sentence role and your own paraphrase.
Then test the word in a new sentence. Reading becomes vocabulary transfer rather than dictionary lookup.
Keep copyrighted source quotations brief and necessary.
Reading for Writing Craft
A writer may study structure, transitions, pacing and argument without copying phrasing. Ask SI to describe techniques abstractly: how the introduction sets stakes, how evidence is sequenced, how the conclusion returns to the question.
Create your own outline using the technique on a different subject.
This preserves learning while avoiding imitation of protected expression.
Reading With Time Budgets
Set a time budget according to source importance. Five minutes may be enough for triage; thirty minutes for an important research article; longer for central evidence.
Time budgets prevent low-value sources from consuming the same attention as decisive ones.
If the article proves more important than expected, upgrade the budget deliberately.
Reading Under Deadline Pressure
Under pressure, protect three things: source identity, decisive evidence and qualifications. Do not sacrifice them merely to produce a faster summary.
Use section maps and targeted find operations to locate the most relevant passages. Keep unanswered questions explicit.
A bounded partial understanding is better than a confident fabricated full understanding.
The Comprehension Test
After reading, close the article and answer four questions: What was the question? What was the thesis? What was the strongest evidence? What was the main limitation?
If you cannot answer one, reopen the relevant section rather than rereading everything.
This test distinguishes comprehension from familiarity.
The Transfer Test
Apply the article’s concept to a new example. If the article explains a mechanism, identify whether that mechanism appears in another case.
Transfer demonstrates that you understood the concept rather than memorised wording.
For research, transfer can reveal whether the author’s framework generalises beyond the source context.
The Verification Test
Choose one central factual claim and open its underlying source if available. Confirm that the article represents it accurately.
This test is especially valuable for articles that synthesise research or statistics.
You do not need to verify every sentence equally; prioritise claims central to your use of the article.
The Counterargument Test
State the strongest objection to the article’s conclusion. Then ask whether the article addresses it and with what evidence.
This deepens comprehension and prevents passive agreement.
If the counterargument depends on external facts, research them rather than inventing them.
The Receiver Test
Can the final reading note serve its intended receiver? A future researcher should see evidence and limitation. A student should see concepts and questions. A manager should see decision relevance.
If the note contains everything but no useful priority, compress again around receiver needs.
Reading is complete when the information is usable, not when every sentence has been paraphrased.
A Full Worked Reading Example
Imagine a 4,000-word article claiming that a new study method improves retention. First record source metadata. Scan headings. Read introduction and conclusion. Identify that the article cites one experiment and several expert comments.
Map the central claim, then read the experiment section closely. Extract sample, duration, outcome and limitation. Note that expert comments explain possible mechanisms but do not add independent outcome data.
Create a one-paragraph summary: question, finding, evidence and limitation. Then write a one-line note stating why the source matters to your research.
Finally, open the original study if the claim is important. The article has now served as both source and map to stronger evidence.
A Full Worked Reading Example: Policy Commentary
An opinion article argues that schools should restrict AI use. Map thesis, values, empirical claims and policy recommendations separately.
Check which factual claims have sources and which are normative judgements. Identify whether the article represents current rules accurately.
Your note should preserve the author’s argument without presenting the recommendation as official policy.
A Full Worked Reading Example: Technical Tutorial
A long tutorial explains a software workflow. Record product version, prerequisites and target outcome. Scan code headings and identify the section matching your task.
Read that section closely and test the instructions in a safe environment. Note any version mismatch or deprecated API.
The article is useful only if its technical state matches the environment you are using.
A Full Worked Reading Example: Investigative Feature
A feature contains documents, interviews and narrative. Build an evidence inventory: official records, named interviews, anonymous-source claims and reporter analysis.
The summary should distinguish these evidence types. Do not convert an anonymous allegation into established fact.
If the feature matters to a consequential decision, seek corroborating evidence where appropriate.
Article Reading Handoff
When passing an article to another person, include why it matters, the two or three most relevant sections, the main evidence and the limitation.
This is more useful than sending a bare URL or a generic summary.
A good handoff saves the next reader from repeating your navigation work while allowing them to verify the source.
Article Note Versioning
If an online article is updated, your note may need revision. Record access or update date for volatile sources.
When the article changes materially, preserve the older note if it supported a historical decision and create an updated note for current use.
Notes should reflect source versions rather than silently mixing them.
A Final Long-Article Reading Gate
- Source identity is recorded.
- Central question is clear.
- Thesis is stated in your own words.
- Structure is mapped.
- High-value sections were read closely.
- Important evidence is extracted.
- Definitions and qualifications are preserved.
- Counterarguments are distinguished from the author’s view.
- One key claim has been verified when necessary.
- Summary does not exceed the source.
- Reading note serves the receiver.
- Original article remains available for rechecking.
When these conditions are met, you have not merely obtained a summary. You have built a reusable model of the article that is compact enough for fast work and strong enough for later verification.
Read for the Article’s Job, Not for Every Sentence
Long articles usually contain several layers: orientation, argument, evidence, examples, background, caveats and conclusion. Efficient reading begins by identifying which layer your task actually needs.
If you are verifying one statistic, you do not need to spend equal attention on the opening anecdote. If you are studying the author’s argument, you need the thesis, supporting claims and counterarguments. If you are researching methodology, you may need the methods and limitations more than the narrative introduction.
SI helps by mapping the article before deep reading, but the user still decides the reading purpose.
The Argument Skeleton
An argument skeleton reduces a long article to its logical structure: central question, thesis, major supporting claims, evidence for each claim, counterarguments, qualifications and conclusion.
Ask SI to produce the skeleton with section locators. Then compare the skeleton against the article. Missing or invented support is easier to catch when each claim is mapped to evidence.
The skeleton is more useful than a general summary when you need to understand how the article reaches its conclusion.
The Evidence Skeleton
An evidence skeleton lists only the evidence: studies, statistics, quotations, examples, documents, interviews or datasets used by the author.
For each item, note what claim it supports and whether it is primary, secondary, anecdotal or interpretive evidence.
This lets you distinguish an article with many paragraphs from an article with genuinely diverse evidence.
The Qualification Layer
Long articles often contain important qualifiers that summaries remove: “in this sample”, “under these conditions”, “may”, “associated with”, “preliminary”, “not statistically significant” or “the authors caution”.
Ask SI to extract every qualification attached to the main claims. Then read those passages yourself.
A fast reader who preserves qualifications can be more accurate than a slow reader who remembers only the headline conclusion.
The Definition Layer
Authors often define key terms differently. One article may use “productivity” as output per hour; another may use manager ratings. One education paper may define “engagement” as attendance; another as self-reported interest.
Extract definitions before comparing sources. A disagreement may disappear once terms are aligned.
SI can build a glossary, but verify definitions against the article itself.
The Assumption Layer
Some assumptions are explicit; others are embedded in framing. An article may assume that faster is better, that one measure captures success or that one population represents another.
Ask: what must be true for the article’s recommendation to work? Which assumptions are empirical and which are value judgments?
Fast reading becomes deeper when assumptions are surfaced rather than skipped.
Read Figures and Tables Separately From the Prose
Charts and tables can contain evidence the prose simplifies. Examine titles, axes, units, denominators, sample sizes, notes and time periods.
Ask SI to describe what a figure shows, then compare the description with the visual. If precise numbers matter, read them directly or use a data-extraction method you can verify.
Do not rely on a paragraph summary of a chart when the chart itself is central to your claim.
Read Footnotes, Endnotes and References Strategically
Footnotes can reveal qualifications, source origins and disputes hidden from the main narrative. References can show whether an article depends on original evidence or mostly cites secondary commentary.
You do not need to open every reference. Follow the citations attached to the claims that matter most to your task.
This selective source tracing is faster than treating every citation equally.
Read the Methods When the Claim Depends on the Method
For empirical articles, methods determine what conclusions are justified. Sample selection, measurement, comparison groups and statistical design matter.
If you are only orienting yourself, a methods skim may be enough. If you plan to use the article as evidence, read the methods connected to the central claim.
SI can translate technical methods into plain language, but the source remains authoritative for what was actually done.
Read Results Before Discussion in Academic Work
Discussion sections interpret results. Reading the results first can reduce the risk of accepting the authors’ preferred interpretation without seeing the underlying findings.
Ask SI to separate measured result, author interpretation and limitation. Then compare each part with the paper.
This is especially useful when press coverage or abstracts use stronger wording than the measured result supports.
Read for Contradictions Between Abstract and Body
Abstracts compress aggressively. Occasionally, the nuance in the body is weaker or more conditional than the abstract suggests.
Compare the abstract’s strongest claim with the detailed results and limitations. Record any narrowing of scope.
Do the same with executive summaries in policy or business reports.
Read for What Was Excluded
Exclusions shape conclusions. A study may exclude participants with missing data. A report may exclude small markets. A review may limit language or date range.
Ask SI to identify inclusion and exclusion rules. Then assess whether the excluded material matters to your research question.
Fast reading should notice the boundaries of the evidence, not only what remains inside them.
Read for Time Structure
Long articles often mix historical background, current evidence and future speculation. Label each section by time role.
This prevents a historical statistic from being remembered as current or a forecast from being repeated as fact.
For fast-changing topics, the date structure of an article can matter as much as the argument structure.
Read for Geographic and Population Scope
Mark where the article changes population or geography. A narrative may begin globally, cite US evidence and end with universal recommendations.
Ask SI to produce a scope map: claim, population, geography and time period.
This helps determine which parts transfer to your own context and which require local evidence.
The Compression Ladder
- Full article: original text with all nuance.
- Section map: headings and purpose of each section.
- Argument map: thesis, claims, evidence and qualifications.
- Research note: source, finding, limitation and relevance.
- One-paragraph summary: bounded synthesis.
- One-sentence memory: what this source contributes to your project.
Move down the ladder only after the level above is stable. Each compression step should preserve the information needed for your task.
The Reconstruction Test
After reading the summary or map, close it and explain the article in your own words. State the question, main claim, strongest evidence and main limitation.
Then reopen the article and check what you distorted or omitted.
This test distinguishes recognition from understanding. If you cannot reconstruct the argument, the summary may have made the article feel easier without actually becoming learned.
The Citation Test
Choose one important claim from your summary. Find the exact section of the article that supports it. If you cannot locate the support, the claim may be overcompressed or invented.
Repeat for the article’s central conclusion and its main limitation.
A fast reading method should always be able to return from compressed notes to source evidence.
The Counterargument Test
Ask what the strongest objection inside the article is. If the article presents none, ask what evidence or perspective would challenge its conclusion.
This prevents the summary from becoming one-sided simply because the article is one-sided.
The counterargument test is particularly useful for opinion, policy and advocacy writing.
The Receiver Test
A student, manager and researcher need different notes from the same article. The student may need concepts and examples. The manager may need decision implications and risks. The researcher needs evidence, method and limitations.
Ask SI to restructure the same verified reading note for the receiver rather than rereading the entire article from scratch.
Keep the source facts invariant while adapting presentation.
The Relevance Test
Long articles contain interesting information that may not matter to your current project. Add a field called Relevance: why does this source or section change your understanding, evidence or decision?
If you cannot answer, the material may belong in background rather than active notes.
Relevance is one of the main ways to read faster without reading carelessly.
The Stop Rule
Stop reading when the information need is satisfied, not when every paragraph has been consumed. This is appropriate for targeted research, not for situations where full reading is the learning objective.
A stop rule might be: central claim understood, key evidence located, main limitation recorded and relevant passage captured.
If the article becomes central to your argument later, return for full close reading.
Fast Reading Is Different From Skimming
Skimming can be passive. Controlled fast reading is purposeful. It uses a question, structural map, evidence targets and verification.
The user knows what was not read closely and does not pretend the entire article was mastered.
This honesty about coverage is essential when SI accelerates reading.
A Worked Example: Long News Feature
Task: understand whether a reported policy change is official and what changed. First map the article: opening anecdote, policy description, stakeholder reactions, historical background.
Read the policy description and source links closely. Treat stakeholder anecdotes as experience evidence. Follow the official source for the actual rule.
Your research note may be much shorter than the article because only some sections answer the question.
A Worked Example: Long Opinion Essay
Task: understand the argument, not accept it. Map thesis, reasons, examples, assumptions and counterarguments.
Separate factual claims requiring verification from value judgments. Search independent evidence for central factual premises.
The final note should say what the author argues, what supports it and what remains contested.
A Worked Example: Academic Paper
Task: determine whether the study supports a claim. Read abstract for orientation, methods for design, results for measured findings and discussion for interpretation.
Extract sample, outcome, effect, uncertainty and limitations. Compare conclusion with data.
Do not rely only on the AI summary of the abstract.
A Worked Example: Technical Documentation
Task: implement a feature in the current software version. Search within the article for version, prerequisites, examples, warnings and API signature.
Read the relevant section closely and test the example in the actual environment.
Ignore unrelated historical sections unless compatibility matters.
A Worked Example: Long Policy Report
Task: identify recommendations affecting secondary education. Map executive summary, methodology, findings, recommendations and annexes.
Read the methodology enough to understand evidence basis. Extract only recommendations relevant to secondary education and record whether they are evidence findings or policy proposals.
Check annex tables when recommendations depend on data not visible in the executive summary.
A Worked Example: Vendor White Paper
Task: understand a product claim. Map claim, evidence, benchmark conditions, customer examples and limitations.
Treat customer examples as illustration. Follow cited benchmarks or independent sources for comparative performance.
A vendor white paper can be useful without being independent evidence.
A Worked Example: Investigative Journalism
Task: understand a complex allegation. Map reported facts, documents, interviews, denials and unresolved points.
Distinguish what reporters directly obtained from documents from what unnamed sources said and what remains allegation.
Read corrections or later updates before using the article in current research.
Reading Several Long Articles Together
Do not create one independent summary per article and stop. Build a comparison matrix across Question, Claim, Evidence, Method, Scope, Limitation and Relevance.
This reveals agreement and difference more efficiently than rereading all summaries.
Article 27 in the curriculum will develop multi-source comparison further.
Reading Long Articles Into a Knowledge Base
Store structured notes rather than raw AI summaries. Include source identity, date, central claim, evidence, limitations, tags and relevance.
Preserve locators for important passages. When the knowledge base is used later, the source can be rechecked.
A knowledge base should accumulate verified insight, not untraceable compression.
Reading Long Articles for Vocabulary and Learning
Students can use SI to identify unfamiliar terms, but definitions should remain tied to the article’s context. Ask for sentence-level meaning and then check against the source.
After reading, close the summary and explain the argument or answer retrieval questions.
Fast reading should strengthen independent comprehension rather than replace it.
Reading Long Articles on Mobile
Mobile reading increases the value of section maps and article cards. Use SI to identify the highest-value sections before scrolling through a very long page.
Save key passages and source metadata in a compact note. Avoid maintaining dozens of open tabs with no record of why each matters.
The technique improves navigation without reducing evidence standards.
Reading Behind Paywalls
Respect access controls. Use material you are authorised to access. If only a snippet or abstract is available, do not ask SI to reconstruct the unseen full article as if it had read it.
Look for lawful author manuscripts, institutional repositories, abstracts, press materials or related accessible sources where appropriate.
State clearly when your understanding is based only on an abstract or partial text.
Reading PDFs
PDFs can contain columns, footnotes, tables and scanned pages that complicate extraction. Verify page references and inspect figures directly.
Ask SI to cite page numbers for important claims where the system can access them. Check those pages yourself.
Article 26 in the curriculum will focus on PDF analysis in more detail.
Reading With Accessibility Needs
SI can help transform long text into simpler language, audio-friendly structure or chunked sections. Preserve definitions, qualifications and source identity during transformation.
Accessibility transformation should not become factual simplification. The core argument and evidence remain the same even when presentation changes.
Receiver-centred reading is compatible with rigorous reading.
The Reading Note Template
- Source and date.
- Why I am reading it.
- Central question.
- Thesis or main finding.
- Three strongest claims.
- Evidence used.
- Key definitions.
- Qualifications and limitations.
- Important counterargument.
- Relevant passage/locator.
- What the article does not establish.
- Relevance to my project.
- Follow-up source or question.
The Deep-Reading Trigger
Fast reading should trigger deeper reading when the source becomes decision-critical, highly contested, methodologically complex or central to your own conclusion.
Do not treat time saved during orientation as permission to avoid close reading where close reading matters.
The goal is to allocate attention intelligently, not to minimise reading at all costs.
A Reading Confidence Label
Label your note by reading depth: Triage, Mapped, Evidence-Checked or Close-Read.
This is more useful than a vague sense that you “read” the article. It tells future you how much trust to place in the note.
Do not use a compressed Triage note as if it came from a full close reading.
A Long-Article Reading Audit
- Did I identify the reading purpose?
- Did I map the article before summarising?
- Did I locate the central claim?
- Did I identify evidence types?
- Did I preserve qualifications?
- Did I read key figures or tables?
- Did I distinguish author interpretation from evidence?
- Can I locate support for my summary claims?
- Did I record what I did not read closely?
- Can I reconstruct the argument without the AI summary?
A Full Reading Practice Project
Choose a 3,000–6,000 word article relevant to your work. Spend five minutes mapping headings, thesis and evidence. Use SI to produce an argument skeleton and evidence inventory with locators.
Read the two most relevant sections closely. Correct the map. Create a research note using the template above.
Close both article and AI summary. Explain the article from memory. Reopen the source and identify three distortions or omissions in your reconstruction.
Finally, write one sentence describing exactly what the source contributes to your project. This converts fast reading into durable research knowledge.
A Final Fast-Reading Examination
Take a new long article. Before using SI, write your reading question and stop rule. Ask for a section map, argument skeleton and qualifications.
Verify the map against the source. Read only the sections required by the question, plus the methods or evidence needed to judge them.
Create a one-paragraph bounded summary and attach one key source locator. Then explain what the article does not establish.
The examination is complete when you can save time without losing the argument, evidence boundaries or ability to return to the source.
Reading Architecture: Map Before You Compress
Long articles become much easier when the reader understands their architecture. Before summarising, create a map of sections and what each section is doing.
A useful map distinguishes definition, background, claim, evidence, example, objection, qualification and conclusion. This functional map is stronger than a list of headings because it explains why each section exists.
SI can generate the first map quickly. The reader should then check whether the map matches the article rather than accepting a plausible reconstruction.
The Argument Spine
The argument spine is the smallest sequence of claims needed to reconstruct why the author reaches the conclusion.
For each major section, ask: what does the author want me to accept here, and what later claim depends on it? The answers form a chain from premise to conclusion.
A fast summary that preserves the spine can be short without becoming shallow. A summary that removes the spine leaves only conclusions and anecdotes.
Evidence Density
Different sections have different evidence density. Introductions and narrative examples may be long but contain little direct evidence. A short methods or results section may contain the material most important to the research question.
Use SI to identify where quantitative data, citations, experiments or primary records appear. Then allocate close-reading time accordingly.
This is one reason reading by page count is inefficient: not every paragraph has equal evidentiary value.
Qualification Density
Some articles carry most of their nuance in a few qualifying phrases. These phrases limit population, certainty, time or causal interpretation.
Ask SI to highlight the clauses that constrain the strongest claims. Examples include “in this sample”, “may”, “under these conditions”, “associated with” and “the authors caution”.
Then compare the compressed summary with those clauses. If the summary becomes stronger, repair it.
The Claim–Evidence Table
- Claim.
- Evidence used.
- Evidence type.
- Locator.
- Scope.
- Qualification.
- Author interpretation.
- Reader note.
The table creates a reading record that can later become part of a research evidence ledger.
It is especially useful when one article contains several claims of unequal strength.
The Author–Reader Separation
Long reading becomes confused when the reader’s own reactions are mixed into the source summary. Keep separate fields for Author Says and Reader Thinks.
This is useful for contested topics, opinion essays and research interpretation. The reader may strongly disagree while still reconstructing the article accurately.
Accurate reconstruction should come before evaluation.
The Fact–Interpretation–Implication Split
A long article may report a fact, interpret what it means and then discuss implications. Those are three different layers.
Example: fact—attendance fell by 10%. Interpretation—the policy may have reduced participation. Implication—the organisation should change the policy. Each step requires different support.
SI can help label these layers so the summary does not collapse them into one statement.
Fast Reading of Methods
When methods matter, read enough to answer: who or what was studied, how data were collected, what comparison was used, how outcomes were defined and what limitations were acknowledged.
Do not skip methods merely because the article is long. Methods determine what the results mean.
For non-research articles, the analogous question is: how did the author obtain the information?
Fast Reading of Tables and Figures
Tables and figures can carry more evidence than pages of prose. Ask SI to describe what each visual measures, then inspect axes, units, legends and source notes yourself.
Do not accept a generated interpretation if the visual is hard to read or contains complex statistical notation. Return to the original figure and its caption.
For exact numerical work, use deterministic calculation or data tools where appropriate.
Fast Reading of Footnotes and Endnotes
Footnotes often contain source provenance, methodological caveats or important exceptions that the main text compresses.
You do not need to read every footnote equally. Prioritise notes attached to central claims, disputed facts and surprising statistics.
SI can help flag which notes appear structurally important, but the original note remains the source.
Fast Reading of Hyperlinked Articles
Modern long-form web articles often outsource evidence to links. A paragraph can look well supported while the linked source says something narrower.
For critical claims, follow the link. Record whether it is primary, secondary, current and directly supportive.
This connects fast reading with the source-evaluation process from Article 23.
Reading for Research Versus Reading for Learning
Research reading asks what this source contributes to a project. Learning reading asks what you need to understand and retain.
For research, extract evidence, scope, methods and source role. For learning, add explanation, recall and transfer tasks.
One article can be read twice under different purposes without contradiction.
Reading for Decision-Making
Decision reading focuses on facts, trade-offs, assumptions, risks and information that could change the decision.
Ask SI to identify which claims in the article are decision-relevant and which are background. Then verify the high-impact claims yourself.
A decision brief should preserve uncertainty more aggressively than a casual reading summary.
Reading for Writing
Writers often read for structure, argument pattern, evidence and style. Keep style inspiration separate from factual sourcing.
You can ask SI to reverse-outline the article and identify how the author introduces examples, handles objections and closes sections.
Do not copy distinctive phrasing or long passages. Learn the structural technique instead.
Reading for Examination Study
Students need to identify definitions, mechanisms, examples and likely misconceptions. Ask SI to convert the article into retrieval questions only after the key concepts are mapped.
Close the article and answer from memory. Return only to sections linked to mistakes.
This prevents summarisation from replacing studying.
Reading an Article With No Headings
When an article has weak or no headings, create synthetic sections based on paragraph function. Group paragraphs that define, argue, evidence or conclude.
Ask SI to propose section boundaries, then check where the topic or argumentative role actually changes.
A synthetic outline can make an essay much easier to navigate without changing the source.
Reading an Article With Many Headings
Some web articles use dozens of headings for scanning. Not every heading represents a major argument step.
Cluster headings into larger themes. This prevents the article from appearing more fragmented than it is.
The final map should reflect conceptual structure, not merely HTML structure.
Reading Articles With Repetition
Long online articles sometimes repeat key points for different audiences or search terms. SI can identify repeated claims and compress them into one accepted statement.
Before removing repetition from notes, check whether later repetitions add a qualification, new example or different evidence.
Compression should remove duplication, not unique content.
Reading Articles With Strong Rhetoric
A persuasive article can use vivid anecdotes, loaded language or emotional framing. Identify which parts are evidence and which are rhetorical support.
Ask SI to restate the factual argument in neutral language, then compare with the original to see what rhetorical elements were removed.
This helps the reader evaluate the reasoning without denying the role of style.
Reading Articles With Weak Evidence
An article can be useful for understanding an argument even when its evidence is weak. Mark it as Perspective or Commentary rather than Reliable Evidence.
The source can still contribute vocabulary, hypotheses or examples.
Source role should follow evidence quality rather than a binary keep/discard judgement.
Reading Articles With Mixed Quality
Some sections may be strong and others weak. Evaluate at claim level when necessary.
A technical explanation may be excellent while the market forecast is speculative. Preserve the strong material without granting the whole article equal authority.
This is another reason claim-level notes are more useful than one reliability score.
The Three-Pass Reading Method
Pass 1 — Orientation
Scan source identity, headings, introduction and conclusion. Decide relevance and reading purpose.
Pass 2 — Structural reading
Map claims, evidence, definitions and qualifications. Read high-relevance sections closely.
Pass 3 — Verification and recall
Check critical claims, produce synthesis and test your own understanding.
The three passes can be short or long depending on article difficulty. Their value is separation of navigation, comprehension and validation.
The One-Question Reading Method
When time is extremely limited, bring one explicit question to the article. Search within the text for sections that could answer it.
Extract the answer, evidence and qualification. Then decide whether the article is strong enough to keep in the research set.
This is much safer than asking for a generic full-article summary you do not need.
The Five-Question Deep Reading Method
- What is the article trying to answer?
- What is the author’s answer?
- What evidence supports it?
- What limits or challenges it?
- What does this change for my research, learning or decision?
These questions form a compact deep-reading framework across many article types.
The Summary Pyramid
Build notes from detailed to compressed: evidence notes → section summaries → argument map → paragraph synthesis → one-sentence takeaway.
This direction is safer than starting with one sentence and trying to reconstruct missing detail later.
The pyramid gives you short notes for recall while preserving deeper layers for verification.
Summary Drift
Every compression step can introduce drift. A section summary can lose a qualifier; the article synthesis can merge two distinct claims; the one-sentence takeaway can become stronger than the article.
Check the shorter layers against the deeper ones. Critical nuance should survive compression.
When a takeaway cannot preserve an essential qualification, make it longer rather than misleading.
A Reading Error Log
Keep recurring reading failures: missed qualification, misunderstood term, wrong thesis, ignored method, unsupported inference.
Use the error log to improve future reading prompts and checklists.
Fast reading becomes a skill when the user learns which parts they personally tend to miss.
Reading Handoff to Research
When an article enters a research project, convert reading notes into the project’s source register and evidence ledger.
Do not leave important evidence only in a chat summary. Store source identity, locators and claim fit in the project record.
This preserves traceability across dozens of sources.
Reading Handoff to Knowledge Base
When an article becomes durable knowledge, store the minimal useful representation: source, thesis, key evidence, limitation, tags and relevance.
Do not store giant generated summaries by default. Knowledge bases become harder to use when every source produces many pages of notes.
Save detail in proportion to future value.
Reading Handoff to Writing
Before citing an article in new writing, return to the source and verify the exact claim. Do not cite from the memory of a generated summary.
Keep interpretation and quotation boundaries clear.
The reading note speeds retrieval; the source remains the authority.
A Final Long-Article Reading Gate
Before declaring an article understood, explain its central question, thesis, three important supports, one qualification and one unresolved issue without looking at the summary.
Then reopen the article and check your explanation. Repair the first missing or distorted element.
If the article is part of research, also state exactly what claim it can support and what it cannot.
The article has been read productively when you can both compress it and reconstruct why its conclusion follows.
Reading Governance: Keep Notes Tied to the Source
When long-article notes become part of a larger project, govern them like research assets. Each note should preserve source identity, access or publication date where relevant, and the role the article plays in the project.
Do not allow a generated summary to become detached from its source and circulate as if it were an independent authority. Keep the URL or citation with the note, and preserve the original article for important verification.
This is especially important when notes are later reused in presentations, reports or other SI conversations.
Reading Notes Need Provenance
A note should distinguish what the author said, what SI summarised and what you inferred. These are different layers.
Use labels such as Author Claim, Evidence, SI Compression and My Interpretation when the distinction matters. A future reader should not mistake your inference for the source’s own conclusion.
Provenance prevents interpretive drift as notes are copied between projects.
Article Note Freshness
Online articles can be updated. Product guidance, policy explanations and live journalism may change after your first reading.
For volatile sources, record the date you read or checked the article. When the note supports a current claim later, refresh the source before reuse.
Evergreen essays may need less frequent review, but current operational claims should not rely on stale notes automatically.
Reading Notes and Source Hierarchy
A long article may be useful context without being the strongest source for a claim. Mark its source role: Primary, Secondary, Commentary, Orientation or Experience.
When the article points to original data or policy, follow that source when the claim matters. Keep the article for explanation if it adds useful interpretation.
This allows fast reading to feed strong research rather than replace it.
Reading Note Deduplication
Research libraries often contain several articles that repeat the same study or press release. After reading, identify common source lineage and avoid storing five notes as if they represent five independent findings.
Merge redundant notes or mark them as secondary coverage of the same underlying source.
Deduplication reduces both research clutter and false consensus.
Reading Note Retirement
Retire notes when the article is superseded, corrected or no longer relevant to the active question. Archive rather than delete when the older state is historically important.
A reading system becomes more useful when inactive material stops competing with current evidence.
The goal is not to accumulate summaries forever. It is to maintain a library whose notes still have a clear role.
A Reading Handoff Checklist
- Source title and link are present.
- Date or freshness is visible when relevant.
- Central question and thesis are clear.
- Important evidence and qualification are preserved.
- Source role is labelled.
- Your interpretation is separate from the author’s claim.
- High-value sections are identified for reopening.
- One-line relevance note explains why the article matters.
- Any required verification is listed.
- Superseded or duplicate notes are marked.
This checklist turns a personal reading note into something another researcher can safely use.
A Final Comprehension-and-Use Gate
Before calling the reading complete, close both the source and SI summary. Explain the article’s question, thesis, strongest evidence, key qualification and main limitation in your own words.
Then reopen the source and correct your explanation. Finally, state why the article matters to the current task and what claim, if any, it can support.
If you cannot do those things, the reading may have been fast but not yet productive. Return only to the section that failed the test instead of rereading the entire article.
Fast reading with SI works when compression follows understanding, provenance survives the compression and the resulting note can guide future work without pretending to replace the original source.
Fast Reading Is Controlled Compression
The goal is not to make long articles disappear. It is to identify the structure, evidence and sections that deserve your attention, then compress what you have genuinely understood.
Use SI as a navigator, explainer and note-building assistant. Keep the original article available as the source of truth.
When mapping, close reading and compression work together, you can read more quickly without turning every article into a shallow three-bullet summary.
