Another approach is the customized Web application. Developers can write specialized applications in native code that take advantage of all of the features of a specific smartphone or write a Web application that works in the Web browsers of all of the top devices. "We're seeing more and more of this," says Marshall.
Making a better mobile keyboard
Predictive technologies generally try to discern what choices a user wants to make, but the conceptual breakthrough that makes ThickButtons work is actually the opposite. ThickButtons attempts to guess what the user doesn't want to do.
As a user types, ThickButtons widens keys he or she is likely to strike next, and narrows keys that probably won't be used. When the user backspaces to delete an error, ThickButtons widens the keys next to the deleted one, surmising that the user meant to strike an adjacent key.
As you might guess, the ThickButtons app is built with an embedded dictionary -- but it's a small one of only about 16,000 common words. As the user types, the app compares what's typed with the word list. But rather than trying to guess what word the user is composing, the algorithm eliminates unlikely choices. Letters that probably won't be used narrow on the virtual keyboard; unlikely choices are narrowed.
Using a smaller dictionary solves several problems. It saves memory, of course, and makes prediction much simpler. In fact the company experimented with a much larger dictionary, but it discovered when the algorithm had more choices, it tended to widen too many keys, which lessened usability.
Pretty smart -- but it's too bad iPhone users can't buy it.
This article, "Demo's cool software: There's no iPhone app for that," was originally published at InfoWorld.com. Follow the latest developments from the Demo conference and in mobile computing at InfoWorld.com.