This typing method is amazing, especially for the function of showing recommended word when typing.
However there is one minor drawback of typing in Korean. When you intend to type "국가", it is typed as "구까" because when you type ywice a consonent("ㄱㄱ"), it becomes double consonent("ㄲ"). The problem is that there is no such word "구까", so I expect the word "국가" should be appeared in the recommended words, but it does not. Therefore, when you want to type 국가, you type 국 and wait for a while for the program to recognize it as seperate character and type 가, which hinders greatly the speed of typing.
Also, if you typed "했습니다"(which is one of the most frequently used word in Korean), it became "해습니다".
I just supprised that nobody reported this problem because this flaw is discussed a lot in the Market. I also think this is a critical flaw considering the program's efficiency using recommended words. I hope this problem fixed sooner.
Korean Improvement
Re: Korean Improvement
I think nobody has brought this up because people are mostly bypassing this problem by adjusting the double tap interval and making use of the word autochange feature. Or maybe it's just because us Koreans are generally shy to post anything in English... 
Well, I have a suggestion for the problem above. The problem arises from the fact that multiple uses are sharing one input: double or triple taps on a single input. So we need a way to differentiate these taps in a way that Smart Keyboard can easily understand... Cyril, does Smart Keyboard measure the time interval between each keystroke? Since there is an option that adjusts double tap interval, I assumed so... If it does measure the intervals, that could be used to recognize which use the user intends with double or triple taps.
For example, "국가" is ㄱ ㅜ ㄱ ㄱ ㅏ: if the user intended "국가" the interval between ㅜ and ㄱ has a good chance of being shorter than the interval between ㄱ and the second ㄱ. On the other hand, if he intended "구까", the interval between ㅜ and ㄱ would probably be longer than that between the two ㄱ's. The same to recognize 했습니다 and 햇씁니다.
Now this won't solve the problem in every case, and the triple tap still remains. But the idea might hopefully give you some ideas...
Um, I just thought of a second approach. Could you give us a new option in the gesture menu, like "start this character in a new (compound) letter"? For example, when I type 학교 = ㅎ ㅏ ㄱ ㄱ ㅛ, if I have the upward or downward gesture defined as "start in new letter", I will type the second ㄱ with an upward or downward slide to write 학교. If I wanted 하꾜, I would not use any slides. This could solve all related problems.
Well, I have a suggestion for the problem above. The problem arises from the fact that multiple uses are sharing one input: double or triple taps on a single input. So we need a way to differentiate these taps in a way that Smart Keyboard can easily understand... Cyril, does Smart Keyboard measure the time interval between each keystroke? Since there is an option that adjusts double tap interval, I assumed so... If it does measure the intervals, that could be used to recognize which use the user intends with double or triple taps.
For example, "국가" is ㄱ ㅜ ㄱ ㄱ ㅏ: if the user intended "국가" the interval between ㅜ and ㄱ has a good chance of being shorter than the interval between ㄱ and the second ㄱ. On the other hand, if he intended "구까", the interval between ㅜ and ㄱ would probably be longer than that between the two ㄱ's. The same to recognize 했습니다 and 햇씁니다.
Now this won't solve the problem in every case, and the triple tap still remains. But the idea might hopefully give you some ideas...
Um, I just thought of a second approach. Could you give us a new option in the gesture menu, like "start this character in a new (compound) letter"? For example, when I type 학교 = ㅎ ㅏ ㄱ ㄱ ㅛ, if I have the upward or downward gesture defined as "start in new letter", I will type the second ㄱ with an upward or downward slide to write 학교. If I wanted 하꾜, I would not use any slides. This could solve all related problems.