while i've not spent much time with the "nosql" products, its mostly because i've decided to stay SQL. The bottom line is that most humans think of data the same way SQL thinks about data. When you talk to people outside of engineering (i.e. mgmt or customers), they expect the product to be able to do things that SQL does naturally (normalized data for easy changes to things like 'user name', join and sort based on a variety of cross cutting properties, etc). You can always make both do the same thing eventually, but one will fight you more than the other....
looks great - doc site is cleaner yet still very familiar. Everything else aside, I'd kill for better doc search - or just simply allowing () to search for functions. Trying to refresh my memory on how the .first() or .latest() method works, for example, requires browsing. no real way to search for it.
I use Tower daily, and used the Tower 2 beta for a bit. I have to say I think they are missing an opportunity. There are a few cool new features (like autofetch) but for the most part, the upgrade smacks of change-for-change-sakes not actual progress towards a good workflow. More clicks are required, keyboard-only navigation is tough, diffs are hidden by default.
One thing most of these guys do is simply smack a UI on top of a command line, rather than explore ways to manipulate the available data into a better workflow than you could possibly get via commands. the conflict wizard is a good step, but that is only part of the story. Repo's are full of data ready to be mined in ways that aren't just 'git pull' and 'git merge'. There is a visual file history, changes made per-person, etc.