Basic Squirm has no concept of control. There are rules that say where a cell can and cannot move to but no rules that say which choice to make. In the applet on the how Squirm works page the cells choose at random from the moves available to them.
The applet on the Squirm front page, however, demonstrates two types of behaviour. The large creature (400 cells) moves at random while the small creature (9 cells) heads steadily downwards. This is one of the simplest non-random behaviours that can be implemented - where a cell has a choice of squares to move to, it chooses the one that is nearest the bottom.
An even simpler behaviour is for each cell not to move at all - this results in static blocks of cells that act as obstacles for other creatures.
Explore the Squirm worlds from the front page, send in suggestions for Squirm worlds you want to see. Hopefully soon the Squirm applets will be controllable, so you can play with different configurations yourself.
Consider that at every time step each and every cell has a choice of where to move to (or whether to stay still). Different tactics give vastly different overall behaviours. Imagine the strategy being different for different parts of a creature, or for the strategy to be decided on the fly by a central brain, or in response to local stimuli such as the presence of enemies. Imagine the control being evolved over time, with the most successful strategies being retained.
Squirm as it stands at the moment is more of a framework for a potential artificial life simulator than anything else. It demonstrates how flexibility of form can be achieved in a computationally efficient manner.
The java source for Squirm is available here: squirm.zip (7k). Compile using Sun's JDK (http://java.sun.com/products/) with the command javac *.java and then load the html file into a java-enabled browser. Contact me with any questions: firstname.lastname@example.org.