The amyloid hypothesis is still used to explain the pathogenesis of Alzheimer's disease. Despite all progress made, however, the molecular causes of the amyloid pathology, and of the tau pathology, tend to be ignored in most patients with this disorder (sporadic, late-onset). Mutant genes for amyloid precursor protein (APP) or presenilin cause early-onset familial Alzheimer's disease (<1% of all cases) and have helped to elucidate APP processing and amyloid-peptide formation by alpha, beta, and gamma secretases. Inhibition of production of amyloid peptides by inhibitors of beta and gamma secretases has been suggested as the rational and most specific therapeutic approach. Alternatively, or additionally, the activation of alpha secretase would increase non-amyloidogenic processing of APP. Here we review fundamental, genetic, and clinical arguments on which the therapeutic strategies for design of secretase agonists and antagonists are based, with special attention to physiological model systems to assess the potential of current efforts.