Where is that document…?
One thing that’s bothered me about Patchwork for a while is the friction around where to save documents and how to find them again. My dissatisfaction with the current model of locating documents is compounded by the Add to sidebar tool. This tool should not need to exist.
Add to sidebar was created as a practical solution to a real problem - users who opened documents in Patchwork and then closed the tab had no way to find the document again within the system unless they had saved the link somewhere. It solves the problem of locating previously opened documents by adding them to the sidebar, where it creates a new problem of unbounded sidebar growth. I feel distracted by even the thought of the hundreds of documents in the sidebar of @pvh and others who have spent years using Patchwork. Add to sidebar means now documents are there and they can be found, but it’s still not necessarily easy, especially since search is currently limited to the document title. Ultimately, it is unsatisfying to live in a system where documents have to be in the sidebar to be found again.
The solution to an overwhelmingly long list of documents in the sidebar is to take the time to organize all those documents within the sidebar or to consciously think about where to save a document and to organize each one as they come in.
The first approach requires dedicating time regularly, which can be hard to find. The number of systems that have been proposed to solve this problem of digital organization [ref: PARA, GTD, etc] suggests it’s not natural and most people won’t do it.
The second approach - finding each document an appropriate home as it enters the system - often has high friction. If I’m creating a wiki, then thinking about where a document should go in relation to other documents is both natural and a useful practice. If I have an idea and want to sketch it on a canvas before I forget, then stopping to think about where this nascent thing should live at some point in the future when it’s in some form that I can’t yet predict will be a huge interruption in my creative process (spoiler: I’m not going to do it). (To be clear: my new canvas doesn’t need to be added to the sidebar, but it does live at the top level adding to my unbounded list of documents unless/until I do something about that.)
Ultimately the problems I want to solve aren’t unique to Patchwork, but crop up in most digital systems:
- it’s hard to find things again later unless you implement some form of organization
- there’s friction and high effort requirements in deciding how/where/when to organize things
- personally, I find it noisy and highly distracting to see lots of things (e.g. document & folder titles) that are irrelevant to my current work
What is the best way to find something?
How could we do better? It seems to me that we can address this by providing two things:
- flexibility in how you group & browse
- efficient search over the contents of local documents
For this exploration, I focused on the first point, motivated by the belief that the mechanism for storing a document should match the mode you’re in. In other words - there is no best way to organize something and a corollary is that there is no best way to find something. It will always be contextual, and the best way will be the one that works. If you’re building a Wiki, you’re probably happy to think about folders and organization. If you’re in early stages of research or creation, you likely don’t know what things will be or how they will fit together, let alone where they should go. I focused on how to find those documents that end up at the bottom of the sidebar list or get dropped into some unsorted misc folder.
I’ve been thinking about 4 mental models for organization:
- the filesystem
- the desk
- the timeline
- the notebook
The filesystem provides fully structured organization with everything in its proper place (one place), and you have to save it in the right place or lose it (or at least need to search for it). This is basically what we have right now, except that things aren’t added to any kind of default bucket unless you click the “Add to sidebar” button. (There is a discussion to be had about filesystem vs. tags which I won’t go into here.)
The desk is a spatial view where you can place things and remember where they are in “physical” relation to each other. chee’s new “Space” spatial folder viewer tool is an exploration of this.
The timeline is simply a linear historical view of documents - it’s browser history. I find it to be an incredibly useful way of finding lost things. I can search back to around the time that I think I last saw it. I can look at all of the other things and find the ones that I remember interacting with around the same time. It enables me to use my own stored context and memories for retrieval of a document in the system. If I can’t remember any text from a document I want to find, then search won’t help me. But if I remember that I looked at it last Wednesday between checking my email and ordering dinner then browser history is the tool that will let me find it. An additional benefit of browser history is that it keeps all the noise out of site until I need it. The unbounded list of all the websites I’ve visited is not something I want to see during daily work and life.
The notebook brings a physicality to the timeline and gives it a sense of space and size. When I look for something in my physical notebook, I think about similar things to when I’m searching in my timeline. When did I start this page? What did I do before it? What did I do after it? But there are additional benefits to a physical notebook. I might have a sense of how big a “chunk” of pages there is between where I am now and where the page I want is. As I flip back through it, I see the content of pages, which makes it easier to hone in on what I’m looking for. Sometimes, when I’m not looking for anything in particular, I’ll flip through on a retrospective tour of specific time periods just to be reminded of things I’ve forgotten, enjoy moments of life and work, or be reinspired.
Browsing timelines and reliving history
A digital version of the notebook model is really just a timeline with enhancements - it’s not that different from a timeline in a photos app or a social media feed, which made it natural to explore the timeline and notebook modes together.
First, I implemented a simple timeline akin to browser history. One patchwork tool listens to all of the “open document” events in the system and logs them into an “Account History” document. For each entry, it keeps track of the timestamp, the document url and the heads at the time it was opened, and the tool it was opened with. This minimal set lets you see exactly what was viewed, when, and how.
The default history viewer tool is a simple version of browser history in Patchwork. It shows opened documents in a list which can be grouped by date or by document. This already gives me as a user a lot more flexibility and control over how to find and interact with my documents. When I click on a link in the account history, it opens the document in the tool it was viewed in. To avoid confusion, it opens the most recent version of the document (even though we saved the heads), because that is the expected user experience for anyone familiar with regular browser history.
However, this approach doesn’t yet make use of additional properties of Automerge and Patchwork or the saved heads in the account history. For that, I created the notebook viewer. The notebook viewer is an initial exploration of what it would be like to combine the powers of browser history with the Wayback Machine. Instead of finding a document in browser history, clicking, and seeing a 404 or a completely new frontpage, you can find it and see exactly what you were looking at in that moment way back when. It explores “reliving” history instead of just retrieving it.
What’s interesting about Automerge is that the document history enables us to browse through and see our Patchwork experience meandering through time, much like a timeline view of a photo album. But unlike a photos or social media timeline, we can do this with living documents. We can see the history of our interactions with a document interleaved with the other documents we referenced or used in between. This triggers memories and gives us insight into our thought process at that point in time. We can also scroll back through our history to find some half-recalled paragraph I read on 13 March 2026 and then jump to the most recent version.
What did we learn?
Offering flexibility in how we browse our documents feels empowering because it lets us map our memories and context to how we interact with the system. Having more ways of finding a document lets me use the model that is easiest and most natural at the moment I’m looking for it. When I know I’m looking for the document called “Patchwork Lab Note: Account History”, I can look in my “Lab Notes” folder. Sometimes I just want “that really cool thing I was looking at last Monday while I was having my afternoon coffee break, I think it was around 2pm, it was so inspiring, why can’t can’t remember what it was called or any of the details?!” That’s what account history is for. When I know I can find things later, I don’t need to worry about organizing documents when I’m in creation mode or getting-things-done mode, which is freeing.
We’ve talked about extensions to this. Documents are local and account history is local, so you could automatically add all sorts of local context as well - weather, location, soundscape (e.g. songs or background chatter), or you could manually add notes and tags that are about the moment in time and your context rather than about the specific document. Then you could look for the document you read while basking in a sunbeam at Prufrock. What we have now reduces friction and gives users more power and control, but I think there are many possibilities from the useful to the whimsical that we haven’t imagined yet.