I’m searching for advice on a simple way to organize files on my Mac. I’d like to use tags and smart folders, but I also use OneDrive to sync many of my documents, which can strip off the tags as part of the syncing process. I’d like not to use a rigid folder structure if I can avoid it, because then I usually can’t find what I need when I need it. I’d also like something that doesn’t take a lot of thought. Ideally, I’d save everything in a single directory and just have the system sort it out. I don’t think that exists yet.
I’d also like to avoid using a tool like Devonthink because I may need to access the files from other computers or devices (hence the OneDrive syncing).
For the smart folders. saved searches in Finder might be good enough. Use the search bar in the upper right of a finder window, then the + button in the upper right to add predicates, then the Save button next to the +, which saves and offers to put them on your Finder sidebar.
You can just dump all your files into your Documents folder with no manual organization on your part. That part exists. We need to understand the “sort it out” to make further suggestions. Specifically, sort it out by what criteria? And next, sort it out auto-magically (e.g. when you open your Document folder, files have some organization on them already) or sort it out when you do something manually? And finally, do something with the files themselves, or just show you something about your choice to sort it out?
Well, this is counter to my experiences and intuition. At the extreme, are you suggesting that you are OK to rename folders in your Documents folder or shuffle them around to other organization schemes every time you open your Documents folder? Because otherwise, the folder structure is already “too rigid” (as I might read between the lines). Or perhaps better said, what constitutes “rigid” versus (at the opposing extreme) “flexible” to you.
All of your files? Or just subsets of your files? And if the latter, do you realize that you are already putting a subset of your total files into their own directory, potentially thereby making them too hard to find later (by your previous arguments against “rigid” folder structures)?
The aim should be to implement a set up that may take some careful thought up front so that its usage takes little to no thought in practice. You seem to dream about a set up that does not take a lot of thought to implement and that can be used with little or not thought in practice. Here is where I would say in response: “I don’t think that this exists (yet if ever)”.
In summary, can you explain what you mean by “sort it out”, by rigid, and by (NOT) “a lot of thought”. Perhaps also give us specific examples of things that do not work for you.
Anything being actively worked on lives in the Active folder. When it’s complete the named folder is moved to the “Completed Year” relating to the date in the Project Folder name.
I then create a folder in Active for any new project which starts with either the Date I create the folder and start the work, or if it’s prep for an upcoming meeting, the date of that meeting. All dates are most significant part first so any folder for a project started today would start with 20260309 (1 because it meets the ISO standard and 2, when sorted Alphabetically the latest projects are always at the bottom.
So the name starts with the date, then a Signifier of the area e.g. CUST(omer), SUPP(lier), DOC(ument), REP(ort) then an explanation of what the project is
e.g.
20260112 - SUPP - Supplier name - SOC 2 review
20260118 - CUST - Customer name - Security Questionnaire
20260208 - REP - New Monitoring System
The Date sorts nicely so if I know I worked on something in August, I can quickly find it by scrolling. I can search by the customer name, and the middle part in bold helps me parse dependent on what I’m looking for. I can also sort by any part of it e.g. Name:“CUST - Customer name”
On top of this, I keep a Spreadsheet of customer engagements so that I can quickly look for when I completed the questionnaire or had a meeting.
The above means I can quickly see what is currently a Work in Progress, and filing is really easy once completed. I’ve never failed to find something if I’ve gone looking for it.
What I envision is a single directory where I store all of my files and they are auto-tagged with metadata by the OS. Some of that already exists (e.g., created date, modified date, file type, size), but it would be great if it could also set categories by examining file contents. I could then tune its ability to categorize, so over time it gets most things right. I keep waiting for the next versions of the OS to be able to do these types of things. With local AI models, this is probably not too far out there, but also may be totally over-engineering things.
Since the computer is so much better at searching than me, I’d like it to do more of that.
Your approach to file storage, taken entirely as a replacement for self-organization, would throw me completely off track. Also, rather than over-engineering towards a specific solution, I wonder that you are under-appreciating its downside trade offs.
Your central question is less about file organization per se and more about whether an AI app or tool exists in any reasonable form to allow hands-free browsing through a file system. You’ve eliminated a potential app, Devonthink, indicating that the AI app or tool must work coherently side-by-side with the existing macOS Finder so that other apps (OneDrive) can do their things effectively too.
With the above in mind, here is a test question. Have you installed an AI tools on your computer, pointed it at your file system, and asked it to classify your files by whatever search criteria it chooses? Perhaps someone can make a specific recommendation in this regard.
Spitballing, because I haven’t (yet!) tried this myself: Claude Cowork + a Claude Skill that lays out the schema for automatic filing and searching.
Microsoft is actively working with Anthropic to embed Claude / Claude Cowork into Office 265. At the moment, the integration is limited to the enterprise and team plans, but I assume it will roll down to the consumer level at some point.
If OneDrive strips tags and you want to use OneDrive then you can’t use traditional tags.
My system uses Hazel to move and file documents based on keywords in the filename or meta information.
I typically save files into a folder called ‘24’ so I can find and attach them to emails quickly. After, wait for it, 24 hours, Hazel moves them to my Autofile folder and a bunch of rules then move and sort. Anything that doesn’t match the rules sits in the folder to be dealt with manually.
This isn’t ’one big bucket’ but provides some structure.
For finding the files I use Alfred to quickly filter files where I know some of the filename (for at is typically date - tag - description. Where date is yyyy, yyyy-mm or yyyy-mm-dad, and description will include a client name as part of a fuller description.)
If Alfred doesn’t allow me to spot the file I want then I will use HoudaSpot to find it.
HoudaSpot is like smart folders on steroids. I barely scratch the surface with my queries.
DEVONthink has been doing this for 20+ years, with its own internal AI. We’ve been ahead of the curve a long time
I’d also like to avoid using a tool like Devonthink because I may need to access the files from other computers or devices (hence the OneDrive syncing).
Though it’s only my opinion but… “may need” is not compelling, “do need” is. You’ve already acknowledged OneDrive can strip metadata while syncing (which is accurate), so if it becomes a real need, what’s your replacement?
