Have a look at Triptych by Chronic Logic. It has an effect similar to your jelly, excepting of course that the blocks never deform like your jelly does.

Just tried the "jelly"; quite interesting
I also looked at the source code. Indents are our friends

Heh, I converted your code to my system before I even tried to read it.

Saturday, I wrote my own jelly in c++, from scratch. It works well but is pretty weak.

I have skipped the diagonal joints, I thinks they are the cause of it's instability.

But without them the jelly crumbels to easy.

I wonder how to make that "transer force" algo should look, I tried to solve it but couldn't hack it.

I also had the idea that a joint should consist of three points and it should struggle to keep the angle between the two resulting lines.
Haven't got that far though.

X-G:
That's not how I would do it.

Trezker:
I would do something more like this:

-First points need to be grouped into 3 (each point can be in more than one group). Note that for this method diagonals are a good idea because groups are more easily formed. Now the point of these groups is to make sure that the triangle that the three points form does not become inverted. If it does, one of the points has moved to where it doesn't belong. Now to check if a triangle has become inverted you need to use something like this:

float a,b,c,d;
a=x1-x2;
b=y1-y2;
c=x3-x2;
d=y3-y2;
if((a*d)-(b*c))<0)triangle_inverted();

where x1,y1,x2,y2 .... are the coordinates of the three points in the group of three.

To see which point is actually causing the triangle to become inverted you will have to increment only its coordinates (ie in my source do: pnt[n].x+=pnt[n].xx,pnt[n].y+=pnt[n].yy for only the point n if you are checking the point n) and then do the above check.

What you actually do when you find a point trying to move into a segment it shouldn't be in I am not exactly sure. You could try something like this:
xx=pnt[naughty_point].xx;
yy=pnt[naughty_point].yy;
pnt[naughty_point].xx=0;
pnt[naughty_point].yy=0;
for(n=0;n<N;n++)pnt[n].xx+=xx/N,
pnt[n].yy+=yy/N;

Where N is the total number of points.

This would distribute its "momentum" or impulse or what ever xx/yy is (I really wish I knew..) to all the jelly. Or maybe try just giving it to points near by or something.

Anyway I could not get this to work but I am reasonably confident that it can work.

Seems like there's a group of jelly-programmers here. Maybe we should start a club...

And then call it JellyHack?

Pretty Nice