Progressive summarization is a note-taking system developed by Tiago Forte that compresses information across repeated passes. Each pass highlights or rewrites the material so that only the most useful parts survive. See Forte's foundational work at https://fortelabs.com/blog/progressive-summarization/.
The system rests on three core attributes. First, it accepts that a single capture pass leaves too much material. Second, it requires the user to return to the same notes later. Third, each return applies a stricter filter than the one before.
These attributes separate progressive summarization from ordinary highlighting or one-time summarization. The value appears only after several cycles of review.
Layer 1: Capture
Write or save the raw material without immediate filtering. This layer focuses on completeness rather than polish.
Layer 2: Bold
Return later and mark the sentences that still seem important. Bold text stays visible while the rest recedes.
Layer 3: Highlight
On the next pass, highlight only the bolded sections that remain relevant. This step reduces the material again.
Layer 4: Summarize
Rewrite the highlighted points into your own words. The summary sits at the top of the note or in a separate document.
Layer 5: Synthesize
Connect the summary to other notes or projects. This layer creates new context and turns isolated insights into usable knowledge.
AI can perform the mechanical steps of bolding and highlighting. It can also draft the layer-four summary from the highlighted text. However, these outputs often lose nuance and original context that only the note-taker recalls, making full mechanical replacement unreliable.
Q: Does progressive summarization require special software?
A: No. The method works in any plain text or outlining tool. The only requirement is the habit of returning to the same notes on different days.
Q: What happens when AI already summarized the source?
A: Use the AI output as your layer-one capture. Then apply layers two through five yourself. The human passes still remove noise that general models miss.