This review will focus on the process of amyloid-type protein aggregation. Amyloid fibrils are an important hallmark of protein misfolding diseases and therefore have been investigated for decades. Only recently, however, atomic or near-atomic resolution structures have been elucidated from various in vitro and ex vivo obtained fibrils. In parallel, the process of fibril formation has been studied in vitro under highly artificial but comparatively reproducible conditions. The review starts with a summary of what is known and speculated from artificial in vitro amyloid-type protein aggregation experiments. A partially hypothetic fibril selection model will be described that may be suitable to explain why amyloid fibrils look the way they do, in particular, why at least all so far reported high resolution cryo-electron microscopy obtained fibril structures are in register, parallel, cross-β-sheet fibrils that mostly consist of two protofilaments twisted around each other. An intrinsic feature of the model is the prion-like nature of all amyloid assemblies. Transferring the model from the in vitro point of view to the in vivo situation is not straightforward, highly hypothetic, and leaves many open questions that need to be addressed in the future.