The stack is based on a (psycholinguistic) paper *.
The main script fixes the first and last char of each input word and randomizes the order of the chars in between.
If the version of LC is greater LC 7 and above, the script uses "truewords" instead of "words". Look while jumbling at "fully-featured" in 'softline' 7 of the right field to see the difference at once.
The demo input is a blog post by Peter about HTML5.
There is a special button for Malte, in case he's using a foreign kezboard ...
Now I hope, zou have fun with the stack

______
* [See "Raeding wrods with jubmled letetrs: There’s a cost"
by Keith Rayner, Sarah J. White, Rebecca L. Johnson, and Simon P. Liversedge
lra.le.ac.uk/bitstream/2381/3897/1/Rayner_White_Johnson_Liversedge_06_PS.pdf
and see also mrc-cbu.cam.ac.uk/personal/matt.davis/Cmabrigde/
for other languages.]
Tested on desktop Mac/Linux/Win with LC 6 and LC 7, especially latest stable versions.
Tested on RaspberryPi B+ and RaspberryPi 2 running latest avaivalable LC 6 (6.5.1) and LC 7 (7.0.4).