Frank posted some nice code on his blog:
http://www.lshift.net/blog/2012/09/27/decomposing-the-ulam-spiral

Couldn't help myself coming up with a simpler (admittedly less efficient) version:

	Display restoreAfter: [
		size := 400 squared.
		primes := (Integer primesUpTo: size) asSet.
		step := 1.
		length := 1.
		pen := Pen new turn: 90.
		[step < size] whileTrue: [
			2 timesRepeat: [
				length timesRepeat: [
					pen fillColor: ((primes includes: step)
						ifTrue: [Color black]
						ifFalse: [Color white]).
					pen go: 1.
					step := step + 1].
				pen turn: -90].
			length := length + 1]].

- Bert -