How to build a visual second brain
without building another system to maintain.
A second brain is supposed to reduce cognitive load. If maintaining it becomes another job, something has gone wrong.
A visual second brain should make it easier to keep useful things, understand what they contain, find them later, connect related ideas and rediscover old material. It should not require perfect filing every time you save something.
Quick answer
The short version
Build around what you already save, keep capture effortless, treat visual material as real knowledge, organize meaning rather than everything, and design for the day you remember only part of it.
Updated September 2026
01 · Section 01
Start with what you actually save.
Don't design a knowledge system around an imaginary future version of yourself. Look at what you already keep: screenshots, websites, PDFs, photography, interfaces, notes, articles, research.
Those are the raw materials of your real memory system. A system that ignores them in favour of tidy prose notes will be abandoned in a month.
02 · Section 02
Make capture nearly effortless.
If every saved item requires a filing decision, the system creates friction at the worst moment, while you're in the middle of the work that produced the thing worth keeping.
Capture first. Understand later. This ordering is the single most important design choice in a personal memory system.
03 · Section 03
Treat visual information as knowledge.
A screenshot can contain an idea, a design pattern, research, text, a conversation, a place, a product. Visual material shouldn't become second-class information simply because it isn't plain text.
Most note-first systems quietly demote images to attachments. Whole categories of thinking, anything you learned by seeing, then live outside your knowledge base.
04 · Section 04
Organize meaning, not everything.
Use collections when they help answer a real question. Don't create categories because the system expects every item to belong somewhere.
Universal filing is the most common way a second brain becomes a chore: the taxonomy grows faster than the thinking it was supposed to serve.
05 · Section 05
Build for imperfect recall.
The real test isn't “Can I store this?” It's “Can I find this six months from now when I remember only part of it?”
Design around that future query. It's vague, partial and visual, and it's the only query your system will ever actually be asked.
01CaptureAnything, instantly.→
02UnderstandText, subject, description, date.→
03ConnectRelated things find each other.→
04RecallA partial memory is a valid query.→
05ThinkOld material informs current work.→
06RediscoverThings return when they become relevant.
The loop a second brain needs to close.
06 · Section 06
Keep context.
Useful memory usually includes what you were working on, what else you saved, when you found it, what idea it connected to and why it mattered.
Context is also what makes a saved thing legible years later, once the reason has faded. Without it, an archive becomes a pile of orphaned artefacts.
ScreenshotA dense analytics view
Kept for the hierarchy, not the numbers.
Related to: Dashboard research
ArticleAn essay on local-first software
Read the week we chose an approach.
Related to: Sync decision
PDFA typography specimen
Three pages that mattered.
Related to: Brand refresh
NoteTwo sentences about why this matters
Written once, worth more than five tags.
The raw materials, kept with their context.
07 · Section 07
Let old material become useful again.
The value of memory isn't only retrieval on demand. Sometimes an old reference becomes useful because today's work gives it new meaning, you didn't have the question when you saved the answer.
A strong system makes that possible without becoming noisy. Resurfacing that interrupts is a feed; resurfacing that waits until it's relevant is memory.
08 · Section 08
Keep the system yours.
A personal memory system reveals what you're interested in, what you're working on, what you're researching and what you're thinking about next. That's an unusually complete picture of a person.
So the architecture is part of the design, not a footer statement. Know the answers before you commit years of material to something.
Where your data lives
On your machine, in someone's cloud, or both.
What is processed locally
Reading, describing and indexing can happen on-device.
What leaves the device
And when, and whether you can turn it off.
What external AI receives
Content, or only what you explicitly send.
How you export
A memory system you can't leave isn't yours.
09 · Where Shelf fits
How Shelf approaches the visual second brain
Shelf starts from a simple idea: you shouldn't have to remember where something was saved in order to remember it. It treats the things you choose to keep as memories that can carry meaning, visual information, context, relationships and time.
Capture is a shortcut. Reading, describing and indexing happen on your Mac. Search works on words and sense together. Older things come back when they line up with current work, and your library exports.
The goal isn't to make you maintain a better database. It's to make the things you kept easier to return to.
A second brain should remember more work than it creates.
Shelf V3 is the current build. Shelf does not use automatic updates, so download the latest build here and install it over your current version.
This is intentional. Shelf is built around a local-first, privacy-focused philosophy. An automatic updater requires additional background processes, network connections, update infrastructure, and another system with permission to modify the app. We prefer to keep Shelf’s network footprint and background activity as small as possible.
Manual updates keep the process explicit: Shelf only checks the internet when a feature actually needs it, and you decide when a new version is downloaded and installed.