Sat, 18th Jun '11, 11:50 am::
People ask me all the time about the status of KType. I had a demo running in February and since then I've spent hundreds of hours coding with nothing to show so people wonder what is it that I'm really doing. I'm doing SCIENCE! And that's the problem. While I can code up a whole new menu system in a day, doing actual research takes a lot of time. But I do have something interesting to share.
KType suggestions are coming along great. When a user types "wy so srs", KType can now suggest "why so serious". KType will run on the iPad and cannot rely on an internet connection to make suggestions. So I have to come up with these suggestions in real-time as the user is typing, based on the data already stored within the KType iPad app. My goal is to make 5 good suggestions within 0.5 seconds of the user pressing a letter. With a dictionary of over 600,000 words and phrases saved on the relatively slow iPad storage, it is no easy task to pick the best 5 choices in under a second. Just reading the 600,000 phrases takes over 8 seconds! So I have to be as clever as I can be and only search within words that I can predict might match.
To put in more bluntly, I have been spending months just to save milliseconds. Because in the end, the user's experience matters more than anything.