When it is time to be at a computer, DT is the centerpiece of my workday and personal interests. Collecting, organizing, researching, extending: all of these are excellent features. DT 2 is more enabling than DT 1, and I anticipate that the remaining features that will bring DT to the end of the beta phase will be equally helpful. However, as more data collects inside my databases, the biggest shortcoming becomes more obvious. The shortcoming is the ability to invoke the powerful potential of the DT database to RE-organize.
On one level, reorganization is possible in small steps. The “AI” lets me select this one thing and see what piles it might belong to. But I can’t select a pile of things and see what other piles it relates to. To speedily create collections without physically moving things around (the key is “speedily”). For example, we can do a “see also” and make a replicant group of those things, but it’s a two step process that requires a pre-existing group to which the replicants are sent, and it’s still limited to the model: show me how this one thing relates to a bunch of other stuff. In that one respect, in my personal opinion, the AI interface stops at around 2002. I’m confident the database has the subtle connections that would enable a more robust interface. I hope that the DT team is thinking of the future Version 3 and beyond, where the more robust interface is possible.
What is “more robust”? These things: tools to investigate many-to-many and many-to-one relationships in my data, not only one-to-many. Tools that rely on the data itself and not metadata like tags (metadata is stateful by definition - data exploration should be stateless). Transparency in the AI so that I can easily see and tweak the factors that go into the “score”. AI that depends not on the piles of things I’ve already made (groups) but on the intrinsic features and relationships within the documents and between the documents. Tools that suggest new categories and collections. Tools that work across databases, even if a database is not open (ok, that might be going a bit too far, but it’s a nice thought). Much easier abilities to create and destroy relationships on the fly. (A simple but powerful partial step in this direction would be to change replicate/duplicate/move to prompt for, and create, a new group rather require pre-existing groups). Beyond these, there’s the possibilities of visualization, but that might also be beyond the definition of DT.
I firmly believe that Eric, Christian, Annard, Bill and the rest of the DT team have very successfully met their publicly stated objectives with DT 2, even as it stands with PB 7. It’s not only best in class, it is the class. Because of that, I’m confident that there can be a DT 3 that vastly expands the capabilities and shoves the “AI” well beyond what any Mac app has delivered thus far.