Collections looked like a simple feature. They’re turning into the core product decision.

I’ve been spending a lot of time lately refactoring the collection experience in Voyager.

Voyager started from a pretty simple belief: file management is still stuck in a folder-first model, while actual work is much more contextual. So instead of treating files as things that only live in folders, we’ve been building around attributes, related files, and collections. 

What surprised me is that collections looked like a straightforward feature at first, but they’re turning into one of the hardest product decisions.

If collections are too lightweight, they feel disposable. People create them, but never come back.
If they’re too structured, they feel heavy. People hesitate before creating them at all.

That tension has been driving a lot of my recent work.

I’m refactoring the flow around:
• when a collection should be created
• how much should be automatic vs explicit
• what makes a collection feel like a working set instead of passive storage

This has changed how I think about Voyager.

I used to think the hard part was “help users find the right files.”
Now I think the harder part is “give them an output they can actually keep working from.”

We already built system-attribute-based collection definition and natural-language collection creation, and now I’m trying to make the collection behavior itself feel much more coherent. 

Would love to hear from anyone who has worked on search, PKM, productivity, or file workflows:
how did you think about the balance between lightweight and structured?

If you’re curious what Voyager looks like today, it’s here:
voyager.fm