AI Note-Taking vs Manual Notes: When to Use Each Method
Compare AI and manual note-taking across learning, meetings, and research contexts. Find the right balance for your knowledge workflow.
Second Brain
Master progressive summarization for note-taking. Learn Tiago Forte's layering technique to distill captured content into action-ready insights.
You capture an article.
It's 3,000 words.
You skim it once and save it.
Six months later you need that article.
But you forgot what it was about.
You read the entire 3,000 words again.
You waste 45 minutes to find one useful sentence.
This is why progressive summarization exists.
Progressive summarization is the technique of layering highlights on notes over time until only the most essential ideas remain visible.
Each layer removes noise.
Each pass makes the note more distilled and actionable.
After three layers, a 3,000-word article becomes a 200-word actionable summary.
A multi-pass highlighting technique where you:
Each pass removes noise. Each pass makes the note smaller and more useful.
Not one-pass highlighting. One pass leaves too much highlighted. The whole note is yellow. No distillation happens.
Not summarization at capture time. You don't understand the material yet. Summarizing too early loses nuance.
Not mandatory complexity. Start with one or two layers. Add more if needed.
You save too much. You remember too little.
Without progressive summarization:
With progressive summarization:
Time saved: 90% of rereading time
When: During first review (days after capture)
How: Mark sentences that stand out to you.
What to highlight:
Volume: Highlight 10–30% of article (not everything)
Example:
Original text: "Machine learning models trained on historical data reproduce systemic biases from that data. Smith et al. found that criminal risk assessment algorithms predicted 20-30% higher risk for minority defendants compared to majority defendants, even when controlling for crime type and severity. This is not a quirk—it's a predictable outcome of training data bias."
Highlighted version: "Machine learning models trained on historical data reproduce systemic biases from that data. Smith et al. found that criminal risk assessment algorithms predicted 20-30% higher risk for minority defendants compared to majority defendants, even when controlling for crime type and severity. This is not a quirk—it's a predictable outcome of training data bias."
When: On second review (1–2 weeks after capture)
How: From your highlights, bold the most important ones.
What to bold:
Volume: Bold 5–10% of highlights (not all highlights)
Example:
From previous highlights, bold the most important:
"Machine learning models trained on historical data reproduce systemic biases from that data. Smith et al. found that criminal risk assessment algorithms predicted 20-30% higher risk for minority defendants compared to majority defendants, even when controlling for crime type and severity."
The bold part is the core claim with evidence.
When: On third review (1–2 weeks after bolding)
How: Write a short summary capturing the article's core idea.
Template:
"This article argues that [main claim]. Evidence: [key fact]. Implication: [why it matters]. Key source: [who to cite]."
Example:
"This article shows that AI systems reproduce racial biases from training data. Smith et al.'s evidence: criminal risk assessment algorithms show 20-30% higher error rates for minority defendants despite controlling for crime factors. Implication: without auditing, AI amplifies historical injustice. Must cite Smith et al. (2023)."
(~50 words)
When: During weekly review (optional, only if connection matters)
How: Link this note to related ideas in your system.
Example:
Volume: 1–3 links per note (not all possible connections)
Recall the CODE framework from Building a Second Brain:
Progressive summarization is the Distill stage.
Layer 1 (Highlight): During first review
Layer 2 (Bold): During second review (1–2 weeks later)
Layer 3 (Summary): During third review (1–2 weeks later)
Layer 4 (Links): During weekly review (optional)
Notion:
Obsidian:
Roam Research:
All tools work. Choose based on your second brain platform.
Week 1 (Capture):
Week 2 (Layer 1 – Highlight):
Week 3 (Layer 2–3 – Bold + Summarize):
Weekly review (Layer 4 – Link):
Mistake 1: Highlighting too much (first pass)
You highlight 50% of article. No distillation.
Fix: Max 20% on first pass. Let importance emerge.
Mistake 2: Delaying layers
You capture and immediately bold and summarize.
You don't understand the material yet.
Summaries are wrong.
Fix: Wait. Let time pass between layers. Understanding deepens.
Mistake 3: Linking everything
You create links to every vaguely related note.
System becomes noise.
Fix: Link only when connection is strong. 1–3 links per note, not 10.
Mistake 4: Never revisiting
You create summary but never use it.
System becomes archive.
Fix: Reference your summaries when writing. The value emerges when you use them.
Single pass (immediately after reading):
Multiple passes (over 2–3 weeks):
Psychology: Sleep and time between passes help consolidation. Each pass is an active retrieval that strengthens memory.
Reading full article: 30–45 minutes
Next time you need it:
After 10 references to same article:
Time saved: 290+ minutes = ~5 hours per article
Multiply that by 100 articles in your system. Progressive summarization saves you hundreds of hours.
Per article:
For 100 articles per year: 20–40 hours total
Return: Hundreds of hours saved rereading and researching.
Week 1: System feels like extra work
Week 4: You reference first summary (instead of rereading full article). Value emerges.
Month 3: You're synthesizing ideas across 20+ summaries. Connections become clear.
Month 6: Your system is compounding. Writing is 2x faster.
A: No. Start with layers 1–3 (highlight, bold, summarize). Layer 4 (linking) is optional and only useful once you have 20+ notes.
A: Reread it. Highlight is only helpful if you understand the material. If first read is confusing, spend more time on comprehension before highlighting.
A: You can, but it's less effective. Your understanding improves over time. Summaries made 2 weeks later are better than summaries made immediately.
A: 30–50 words is ideal. Long enough to capture meaning, short enough to scan in 30 seconds.
A: Best for: articles, books, research papers, educational content.
Okay for: videos (requires taking notes first), podcasts (requires transcripts or notes).
Not ideal for: social media posts, news headlines.
Progressive summarization turns captured content into distilled, actionable knowledge.
The layers:
The benefit:
Start this week:
After one month, you'll have 4 deeply distilled summaries. After one year, 50. Each one saves you 30 minutes of rereading.
For more on PKM, see Building a Second Brain. For notes, check Fleeting to Permanent Notes.
Summarize progressively. Think efficiently.
More WebSnips articles that pair well with this topic.
Compare AI and manual note-taking across learning, meetings, and research contexts. Find the right balance for your knowledge workflow.
Upgrade your note-taking with first principles thinking. Learn to capture the why behind information so your notes actually build understanding.
Compare mind mapping and linear note-taking across use cases. Learn when visual notes outperform text and how to combine both for maximum insight.