Although Alzheimer's disease (AD) was first discovered a century ago, we are still facing a lack of definitive diagnosis during the patient's lifetime and are unable to prescribe a curative treatment. However, the past 10 years have seen a "revamping" of the main hypothesis about AD pathogenesis and the hope to foresee possible treatment. AD is no longer considered an irreversible disease. A major refinement of the classic beta-amyloid cascade describing amyloid fibrils as neurotoxins has been made to integrate the key scientific evidences demonstrating that the first pathological event occurring in AD early stages affects synaptic function and maintenance. A concept fully compatible with synapse loss being the best pathological correlate of AD rather than other described neuropathological hallmarks (amyloid plaques, neurofibrillary tangles or neuronal death). The notion that synaptic alterations might be reverted, thus offering a potential curability, was confirmed by immunotherapy experiments targeting beta-amyloid protein in transgenic AD mice in which cognitive functions were improved despite no reduction in the amyloid plaques burden. The updated amyloid cascade now integrates the synapse failure triggered by soluble Abeta-oligomers. Still no consensus has been reached on the most toxic Abeta conformations, neither on their site of production nor on their extra- versus intra-cellular actions. Evidence shows that soluble Abeta oligomers or ADDLs bind selectively to neurons at their synaptic loci, and trigger major changes in synapse composition and morphology, which ultimately leads to dendritic spine loss. However, the exact mechanism is not yet fully understood but is suspected to involve some membrane receptor(s).