mrtbld

The spikes on the graph for gingerbread at each Christmas is noise which you could have avoided by adding "android" to each search term. Result: http://www.google.com/trends/explore#q=jelly%20bean%20androi...

It detects a false positive in this picture for "full-body person": http://farm9.staticflickr.com/8232/8366217251_972624d84b_b.j...

The man is detected, but also a shape above the umbrella.

Edit: direct link the the result picture: https://s3.amazonaws.com/dextro_detection_results/debug13579...

I like it.

But I was a bit surprised to see a commenting system in it. This isn't what I expected after having read "Why I Made Throwww" [1]. A blog-like commenting system is not what I would associate with simplicity, nor "the purity of just a message".

Also, a little idea for easy improvement (IMHO): display the length of the text, i.e. the number of words and/or the time to read (e.g. based on a naive average of 250 words per minute [2]).

  [1] http://throwww.com/a/1
  [2] http://en.wikipedia.org/wiki/Reading_%28process%29#Reading_rate

The symmetric, "curious dance" between C and C# in 2009-2011 could be due to the fact that "#" is not a alphanumerical character. Some searches for "C#" may have been miscounted as searches for "C" in that period. (But that's just an hypothesis.)