The ability of peptides to construct specific secondary structures provides a useful function for biomaterial design that cannot be achieved with traditional organic molecules and polymers. Inhibition of amyloid formation is a promising therapeutic approach for the treatment of neurodegenerative diseases. Existing peptide-based inhibitors are mainly derived from original amyloid sequences, which have very limited sequence diversity and activity. It is highly desirable to explore other peptide-based inhibitors that are not directly derived from amyloid sequences. Here, we develop a hybrid high-throughput computational method to efficiently screen and design hexapeptide inhibitors against amyloid-β (Aβ) aggregation and toxicity from the first principle. Computationally screened/designed inhibitors are then validated for their inhibition activity using biophysical experiments. We propose and demonstrate a proof-of-concept of the "like-interacts-like" design principle that the self-assembling peptides are able to interact strongly with conformationally similar motifs of Aβ peptides and to competitively reduce Aβ-Aβ interactions, thus preventing Aβ aggregation and Aβ-induced toxicity. Such a de novo design can also be generally applicable to design new peptide inhibitors against other amyloid diseases, beyond traditional peptide inhibitors with homologous sequences to parent amyloid peptides.