the point he is making is that users shouldnt have to do this work. The circle feature implies that they will automatically will take care of it when in fact it doesnt. It's a lot of manual labor as he elaborates.
So, it's not magic because it can't read my mind, thus it's a bad feature because it takes work to fully, fully utilize? I guess I'm still missing the point. It'd be nice if Google knew my relationship with these people or somehow knew my best friends from my colleagues or tech interests, but that's asking a bit much.
It's a suboptimal feature. Most developers like to write code that gets used a lot (at least I do). If a feature that you hoped will be popular is only used by a tiny subset, then you probably failed.
In Facebook's case, they may have written the initial lists feature as way to show they're "doing something" about privacy, so simply deploying the feature is a win.
But in Google's case, they've made Circles a central feature of their platform and it's coming up short (so far at least).
Google (or any network) could make it easier for us to use the feature. Facebook has figured this out and actively has some features to help.
The main ones that come to mind are the ability for other people to create relationships, whether personal or work related. This eliminates some of the work when building your profile. Then Facebook implemented the auto-grouping of lists depending on employers, previous interaction, locations, etc. And lastly, they monitor your graph on a daily basis and serve up possible additions to your friend lists which can be added with a simple click.
Google could have done a little of this using Google data such as Gmail lists and all, but they haven't (yet). Personally, I think Facebook's lists and grouping feature is far superior and I'm actually finding out from non-technical friends that they're actually using the friend lists. I suppose this has to do with Facebook automatically pushing a number of default lists to users...