This looks fairly promising, but it's not there yet. For example, try drawing a capital sigma ( Σ ); if you don't get your top and bottom bars perfectly parallel there's a strong chance you'll get gibberish back. Or try pi ( π or Π ); a 1° change in the angle of your bars might make the difference between a handful of relevant results and a full page of nonsensical Devanagari and Gurmukhi characters. Or do a double dagger ( ‡ ); if your middle bar isn't perfectly straight up and down it might think you're doing cent ( ¢ ) or integral ( ∫ ). It seems like it's extremely difficult to get the same results for the same 'search' — every time you get wildly different answers because the slightest little difference in how you draw it changes how it reads it.
So i think the recognition algorithm needs a little work. I know you mentioned font limitations, so i suspect adding more of those would be a good start.
Also, i don't see Japanese support, which is unfortunate. If you could get that working i would be an instant fan. (I deal with transcription of Japanese a lot, and although i'm proficient in kana, my kanji skills are pretty poor and probably always will be. What i've always wanted is a tool that can return kanji characters based on a drawing WITHOUT requiring strict adherence to stroke order. That last requirement rules out every single tool i've found...)
It does seem similar. I actually like Detexify's interface a little bit more because it doesn't require pressing a button after you've drawn your character, although I imagine that is a matter of preference.
Detexify's popularity and utility shows that this sort of thing could be very useful--I suspect I might end up using it myself in the future.
Love the idea, thought the same thing after having seen detexify and doing web development looking for suitable Unicode glyphs (glyph lists are a pain in the, well...). Keep it up and by all means fix the current error. ;)