Heterotic versus Type I

I compare the calculations of special $F^4$ and $R^4$ terms in the (toroidally-compactified) heterotic and type-I effective actions. Besides checking duality, this elucidates the quantitative rules of D-brane calculus. I explain in particular (a) why D-branes do not run in loops, and (b) how their instanton contributions arise from orbifold fixed points of their moduli space. The instanton sum has a simple representation as a sum of the elliptic genera of matrix models.