Andy Matuschak's Memory Systems

Table of content

Andy Matuschak helped build iOS at Apple. Then he led R&D at Khan Academy. And then he quit everything to become an independent researcher, funded entirely by Patreon supporters.

What’s he researching? Tools that expand what people can think.

The Problem With Learning

Here’s something you’ve probably noticed: learning often just… doesn’t work. You take a class, read a book, watch a tutorial. A month later? You’ve retained almost nothing.

Matuschak has spent years obsessing over why this happens. His conclusion: existing media—books, videos, courses—completely ignore how human memory actually functions.

Books assume you’ll remember what you read. You won’t. Not without deliberate practice. And that practice needs to happen at very specific intervals, or your brain discards the information as irrelevant.

The Mnemonic Medium

In 2019, Matuschak and physicist Michael Nielsen built Quantum Country—an essay that helps you remember what you read.

It looks like a normal article about quantum computing. But embedded throughout are flashcard-style questions. After reading, you get review sessions spaced over weeks and months. The system handles scheduling automatically.

The result? Readers actually retain the material. Months later, years later. Not because they have better memories—because the medium itself is designed around how memory works.

This became Orbit, an open platform for “programmable attention”—the idea that software can systematically orchestrate what you think about and when.

Evergreen Notes and Working in Public

Matuschak maintains a public wiki of his working notes. Not polished essays—raw thinking, updated constantly.

He calls this “working with the garage door up.” You can watch his ideas evolve in real-time: incomplete thoughts, dead ends, surprising connections.

His note-taking philosophy: evergreen notes. Each note captures one atomic idea, written for your future self. Notes link to each other densely. Over time, this web of ideas becomes a thinking environment, not just storage.

The system isn’t available for others to use—it’s still a research prototype. But the principles have influenced how thousands of people think about personal knowledge management.

Morning Writing Practice

Every morning, Matuschak writes. Not for publication. Just to process, to think, to develop ideas.

This isn’t journaling. It’s more like having a conversation with himself about whatever problems he’s wrestling with. The writing clarifies his thinking. Sometimes it surfaces connections he missed. Sometimes it reveals that an idea is weaker than he thought.

The discipline is the point. Consistent practice compounds.

AI-Augmented Learning

Matuschak’s recent work tackles a question: how should AI change how we learn?

His How Might We Learn? demo (2024) sketches a vision. Imagine you’re diving into a new field. An AI with deep context about your background and goals could:

Suggest tractable entry points that match your skills
Provide guidance embedded in your actual work, not separate from it
Generate dynamic media that lets you play with concepts
Map personalized paths through canonical texts
Create practice tailored to what you’re building

The key insight: people learn best when immersed in authentic projects. But complex material requires scaffolding. AI could finally let us have both—the motivation of real work with the support of good teaching.

Philosophy

Matuschak believes we’re living in a kind of dark age for tools for thought. Computers could enable “transformative tools that expand the set of thoughts it’s possible to think.” But mostly they’ve given us email and spreadsheets.

The dream he’s chasing: interfaces so powerful they feel like magic to us today, the way visual effects would seem to a cave painter.

That sounds grandiose. But writing and mathematical notation already did this. They didn’t just help people think faster—they made entirely new kinds of thought possible.

Why can’t software do the same?

Practical Takeaways

Memory is a design problem. If you want to remember what you learn, build spaced repetition into your practice. Don’t leave it to chance.

Notes should be evergreen. Write for your future self. Make each note atomic. Link densely. Your note system should be a thinking environment.

Work in public. Sharing works-in-progress attracts collaborators, surfaces mistakes early, and forces clarity.

Practice compounds. Daily writing, daily review—small consistent efforts beat sporadic intensity.

Immersion + guidance. The best learning combines authentic projects with cognitive scaffolding. Don’t sacrifice motivation for structure, or vice versa.