Searching for a Thurston paper with egg / 3-manifold analogy? I remember coming across some article of Bill Thurston’s where he describes a 3-manifold (with boundary?) as being like an egg. In my recollection the interior of the egg, the shell, and even the shell after being cracked all had meaning. As you can imagine, Googling “like an egg three-manifold Thurston” is not very fruitful. If you happen to remember this article’s title, please point me in the right direction!
If I had to guess, I would imagine that this article is more likely to be a research announcement or notes of some kind rather than a paper proving a major result. I guess this because the paper was somewhat “talk-y”. I hope that this paper really exists out there and I haven’t just imagined it!
 A: I believe you're looking for page 211 of Hyperbolic Structures on 3-Manifolds I: Deformation of Acylindrical Manifolds (Annals of Math., 1986), which includes the following paragraph:

A complete description of the three spaces $AH(M)$, $GH(M)$, and $QH(M)$ is
   certainly not rigorously known, but here is a conjectural image, of which certain
   features can be rigorously proved. Let us stick to the case that $M$ is a compact, acylindrical manifold. Then $H(M)$ is a hard-boiled egg. The egg complete with
   shell is $AH(M)$; it appears to be homeomorphic to a closed unit ball. $GH(M)$ is
   obtained by thoroughly cracking the egg shell on a convenient hard surface.
   Apparently no material is physically separated from the egg, but many cracks are
   developed - cracks are dense in the boundary - and at the same time, the
   material of the egg just inside the shell is weakened, so that neighborhood
   systems of points on the boundary become thinner. Finally, $QH(M)$ has uncountably many components, which are obtained by peeling off the shell and
   scattering the pieces all over. Each component is homeomorphic to some
   Teichmuller space - it is parametrized by Euclidean space of some even dimension.  "Most" of the components have dimension zero, for they describe groups
   whose limit set is all of $S_\infty^2$.

Here $H(M)$ is the set of complete hyperbolic manifolds $N$ with a homotopy equivalence $f:M\to N$, $AH(M)$ is $H(M)$ equipped with the 'algebraic topology', $GH(M)$ with the 'geometric topology', and $QH(M)$ with the 'quasi-isometric topology'.
