Note-Taking Rules

Notes exist to be retrieved and used later, not to prove that something was read or attended — a beautifully organized note that never gets reopened, or a transcript nobody can skim, has both failed the same test: they don't serve the future reader, who is usually the same person, months later, with no memory of the original context.

Own words over transcription

• Write notes in your own words as you process the material, not as a verbatim or near-verbatim transcription of the source — transcribing while listening or reading engages a different, shallower cognitive process than paraphrasing, and produces notes that weren't actually understood at capture time.
• If verbatim capture is unavoidable in the moment (fast lecture, dense reading), mark it clearly as unprocessed raw material and schedule a follow-up pass to rewrite it in your own words before it's considered a finished note — a permanent stock of unprocessed transcription is not a note-taking system, it's an unfinished queue.
• Test understanding at capture time by attempting to state the idea more simply or give an example of it — if that's not possible, the material wasn't understood well enough yet to make a good note of it, and that's worth knowing immediately rather than discovering it later.

Atomic notes with links

• Write one idea per note — a note that bundles several unrelated ideas because they came from the same source or session can't be independently linked, reused, or resurfaced when only one of those ideas becomes relevant later.
• Link each note explicitly to related notes at creation time, not as a batch cleanup task later — the links are what make a note collection more valuable than the sum of its parts, and a note created without at least one link into the existing collection is effectively orphaned.
• Prefer many small linked notes over few large notes organized as folders/outlines — a large note can only live in one place in a hierarchy, while an atomic linked note can be surfaced from every context where it's relevant.

Separate capture from organizing

• Capture first, organize later, as two distinct passes — trying to decide a note's permanent structure, tags, and links in the same moment as capturing it slows capture down and causes good material to be missed while mid-thought.
• Use a dedicated fast-capture inbox (a single running list, not a filed/tagged system) during the actual lecture, reading, or meeting, and process that inbox into properly linked, organized notes in a separate session afterward — ideally within a day or two, before context is lost.
• Never let the inbox become permanent storage — an inbox that's never processed is functionally the same as no note-taking system, just with an extra unprocessed step in front of it.

Progressive summarization

• On the first pass, capture the raw idea in your own words. On a later pass, when revisiting the note (because it came up again, or during a scheduled review), bold or highlight the single most important sentence or phrase within it — don't try to identify the "most important part" during the same session the note was first written, before you have any evidence of what it's actually useful for.
• Add each layer of summarization only when the note is revisited for a real reason, not on a fixed schedule regardless of use — a note that's never revisited doesn't need progressive summarization; the layering earns its cost by tracking which notes turn out to matter over time.
• Keep every earlier layer intact underneath the highlights rather than deleting the original text — the point is a layered note where a skim of just the highlights is possible, not a note that's been destructively trimmed down to only its summary.

Retrieval-friendly titles

• Title each note with a specific, declarative statement of its core idea, not a generic label copied from the source ("Chapter 4" or "Lecture notes 3/14") — a title should let you recognize and choose the right note from a list of many without opening any of them.
• Phrase titles as claims or specific questions where possible ("Spacing beats massed practice for retention") rather than topic labels ("Spacing effect") — a claim-style title carries the actual content of the note in the title itself, making the note collection searchable by idea, not just by subject.
• Avoid duplicate or near-duplicate titles across the collection — if two notes want the same title, that's usually a signal they should be merged, or that one needs to be narrowed to the specific distinct claim it actually makes.

A review loop that actually happens

• Schedule note review as a fixed recurring habit (e.g. a short weekly pass over recently added or recently linked notes) rather than relying on stumbling back onto old notes by chance — a note-taking system with no review loop slowly becomes a write-only archive.
• During review, actively look for new links between older notes and anything recently added — the value of atomic linked notes compounds specifically through connections made after the fact, often between notes written months apart on topics that didn't seem related at the time.
• Keep the review session short and low-friction (10–20 minutes, a small batch of notes) — a review process that feels like a large undertaking gets postponed indefinitely and defeats its own purpose.