I went back through my Kindle highlights last month. Hundreds of them. A lot of noise, very little signal. Mostly yellow lines under sentences I already agreed with when I read them.
Highlighting felt like learning. It wasn't. So I made rules.
Highlight these four
- Surprise - contradicts what you believed. Highest value.
- Mechanism - how something worked, not that it did. "Knight financed growth by riding the float between Nissho's credit and Bank of California" beats "Knight had cash problems."
- Reusable phrasing - a line that compresses an idea so well you'd use it in a meeting. Rare. Maybe 3 per book.
- Disagreement - where you think the author is wrong. Mark it. Your own reaction is the asset, not the text.
Don't highlight
- Things you already agree with. This is 90% of most people's highlights, and why nobody rereads them.
- Facts you could Google.
- Anything that felt "profound" without being specific. That's mood, not information.
- Long passages. If it's more than 3 lines you're marking a topic, not a point.
The rule that matters more than the criteria
A highlight you never revisit is worthless. Highlighting feels like retention and isn't.
So: margin note > highlight. 4-5 words in your own words next to the mark.
"Same as Grove's 1x" / "disagree, survivorship" / "steal for AM SOP."
Forces processing at read-time, and makes review actually usable later. A bare yellow line means nothing to you in six months.
On Kindle
No margins, but you can still write notes. Select the text, tap Note instead of Highlight, type your 4-5 words. All of it shows up at read.amazon.com/notebook.
Practical system
- Highlight freely while reading, don't overthink it.
- On finishing: one pass, cut to 10-15 keepers, and write why each one survived.
- That distilled list is what you reread quarterly. Not the book, not the raw highlights.
For biographies
Highlight decisions under uncertainty, not outcomes.
What did they know at the time? What did they choose?
Outcomes are hindsight, poisoned. Decisions transfer.