Why We Built SummaryMaps (and Where We're Going)

SummaryMaps started as a private notebook. We were reading too much and remembering too little, and we wanted a tool that respected both the book and our brain. After a year of building, we still think that's the right framing — and we've learned enough about what that means in practice to write it down properly.

The original problem

We started with a simple observation. We could remember the layout of a friend's apartment from a decade ago, but not the third major argument from a book we'd read the previous month. Spatial memory was effortless. Sequential memory was a chore. And yet most of what we tried to learn arrived as text — sequential by nature.

The handful of times we'd actually retained book content for years, it was because we'd drawn something. A diagram, a flowchart, a sketch in the margin. Drawing made the book stick. But drawing took time, and we'd given it up by the second semester of university because nobody had time to be a slow reader.

SummaryMaps is the answer to that. Drop in a book, get a beautiful, interactive map. Study mode quizzes you on what you almost forgot. Audio reads it back to you on your walk. Book clubs let you share what you've mapped with the people you actually want to talk to. The thing that used to take an hour now takes ninety seconds, and the thing that used to fade in a month now lasts.

What we got wrong in year one

Three things, mostly.

We over-indexed on automation. Early on, the maps were entirely AI-generated. Users liked them but didn't keep coming back. It turned out the cognitive payoff of mapping comes from making the map, not from owning one. We've since added an explicit 'co-write the map' mode where the AI proposes and the reader confirms, branch by branch. Retention is dramatically higher.

We treated every book the same. The first version produced the same shape of map for Atomic Habits and for Sapiens. Argument books and Framework books and Playbooks all came out as identical hub-and-spoke diagrams. That was wrong. Now the AI detects the book's pattern (or asks if it's unsure) and produces a map that fits the shape.

We tried to be a study tool first. What users actually want is something between a journal and a library. The maps aren't there to pass a test, they're there to remember what you read. We've reorganised the app around that.

What's next on the roadmap

Three big bets for the next year.

Better maps for fiction. Most of our energy so far has gone into non-fiction, because non-fiction maps more obviously. But the cognitive payoff for fiction is real — character relationships, thematic spine, plot structure. We're prototyping a 'scene-theme' map for narrative books, and the early tests are promising.

Expert-made maps for classics. Our library of user-generated maps is growing fast, but for the canonical titles — Thinking Fast and Slow, Sapiens, Atomic Habits — we're commissioning maps from researchers in the relevant fields and offering them as a baseline you can fork.

Auto-scheduled review. Spaced repetition for books. The system surfaces the right map at the right interval based on how well you remembered the last review. Most people forget books in months; with two five-minute reviews at the right intervals, you can keep a book for years.

The journal you're reading

This journal — SummaryMaps Journal — exists for a more selfish reason. We want a place to write quietly about books, learning, and mapping, without the algorithmic pressure of a social platform. The articles end with mind maps because we believe in our own product. They're free and they'll stay free.

If you've subscribed to the weekly newsletter, thank you. If you haven't, the signup is at the bottom of every page. One essay every Sunday, no clickbait, no five-thousand-word listicles. The kind of email you actually want to open.

The bigger bet

Reading is a long game. The tools should be too. We've designed SummaryMaps to still be useful to you in ten years, with the maps you build today still readable, still searchable, still yours to export. We've avoided dark patterns deliberately — no streaks, no badges, no public follower counts. The point is to read better, not to be seen reading better.

That's the bet. Quiet writing, slow reading, durable memory. We think there are a lot of people who want this. If you're one of them, you're in the right place.