Many chemical reactions-etching, growth, and catalytic-produce highly faceted surfaces. Examples range from the atomically flat silicon surfaces produced by anisotropic etchants to the wide variety of faceted nanoparticles, including cubes, wires, plates, tetrapods, and more. This faceting is a macroscopic manifestation of highly site-specific surface reactions. In this Account, we show that these site-specific reactions literally write a record of their chemical reactivity in the morphology of the surface-a record that can be quantified with scanning tunneling microscopy. Paradoxically, the sites targeted by these highly site-specific reactions are extremely rare. This paradox can be understood from a simple kinetic argument. An etchant that produces atomically flat surfaces must rapidly etch every surface site except the terrace atoms on the perfectly flat surface. As a result, the etch morphology is dominated by the least reactive species (here, the terrace sites), not the most reactive species. In contrast, the most interesting chemical species-the site where the reaction occurs most rapidly and most selectively-is the hardest one to find. This highly reactive site, the key to the reaction, is the needle in the haystack, often occurring in densities far below 1% of a monolayer and thus invisible to surface spectroscopies. This kinetic argument is quite general and applies to a wide variety of reactions, not just etching reactions. Understanding these highly site-specific reactions requires a combination of experimental and computational techniques with both exquisite defect sensitivity and high chemical sensitivity. In this Account, we present examples of highly site-specific chemistry on the technologically important face of silicon, Si(100). In one example, we show that the high reactivity of one particular surface site, a silicon dihydride bound to a silicon monohydride, or an "α-dihydride", provides a fundamental explanation for anisotropic silicon etching, a technology widely used in micromachining to selectively produce flat Si{111} surfaces. Fast-etching surfaces, such as Si(100) and Si(110), have geometries that support autocatalytic etching of α-dihydrides. In contrast, α-dihydrides exist only at kink sites on Si(111) surfaces. As a result, the etch rate of surfaces vicinal to Si(111) scales with the step density, approaching zero on the atomically flat surface. In a second example, we explain the chemistry that underlies pyramidal texturing of silicon wafers, a technique that is sometimes used to decrease the reflectivity of silicon solar cells. We show that a subtle change in chemical reactivity transforms a near-perfect Si(100) etchant into one that spontaneously produces nanoscale pyramids. The pyramids are not static features; they are self-propagating structures that evolve in size and location as the etching proceeds. The key to this texturing is the production of a very rare defect at the apex of each pyramid, a site that also etches autocatalytically. These experiments show that simple chemi...