Gboard: The master plan to take over iOS

What Google launched today is what I’ve dreamt about for the last couple of years. App switching isn’t too hard. But transferring context between apps isn’t nearly as simple as it should be. The idea of having this context available constantly through a keyboard, or be able to quickly search for something inline is something I’ve been wanting. I’ve prototyped it several times and always thought of it being implemented as an iOS keyboard that uses Google’s incredible Knowledge Graph to show cards of structured information. This is pretty much what Gboard showcases today. What I was unable to make use of in my prototypes were Google’s unparalleled keyword associations and typeahead predictions – something the industry always seems to be playing catch up to.


What I’m surprised about is the decision to roll this out without Google Account login – I imagine this is to prevent the almost predicable initial press about the privacy repercussions of a Google keyboard that not only follows you across iOS, but also remembers all keystrokes to improve its index. There is no question that this was discussed during the design process, since search algorithms are hungry for this sort of stuff.

This is important for Google as it has been maintaining individual indexes for logged-in users over the last couple of years, because it provides them with two things:

  1. Superior Typeahead: A universal index has a universal typeahead. While the prediction is great, Google’s algorithms rely on a whole bunch of signals. When users specify intent more accurately through keywords, they’re bound to get more accurate results. With time, the per-user index gets a more focused list of signals providing the user with a bigger advantage: the same accurate results with fewer keystrokes to indicate intent, the prediction and predictability keep becoming better and better.
  2. The perfect Now Card: While search is all about intent specified through keywords, Google Now has always been, in my opinion, the opposite of traditional Google Search. Instead of keywords, it relies on history, learned behavior and patterns from the user index with one bold idea: never search again. In such a world there would be no search UI: just one exact result for questions you may have or one list to choose from for queries that expect recommendations/choices. The reason we see multiple cards today is to increase the probability of finding something useful in the absence of enough signals to be completely accurate in prediction.

On Android, this sort of index-creation is already happening. Before installing Gboard, I thought this was a similar attempt at play on iOS, where, with authenticated sessions, the user might be able to see Google Now cards generated from all activity across iOS. If marketed the right way, I think users would look past the issue of privacy. While technically very different from Android’s Now on Tap feature, you could get a lot of the same functionality, without initiating much user action. 

The one thing that might be preventing it right now may be a restrictive App Store review process. Cupertino may be scared to open those doors right this moment. While smartphones are becoming way more powerful, they are not likely to be able to compute and deliver information like web services already do today. Allowing this access may only be a matter of time.