The article introduces a novel method called Adaptive Seeds to improve the efficiency and sensitivity of comparing large genomic sequences. Traditional methods like BLAST use fixed-length seeds, which can be inefficient for sequences with non-uniform nucleotide compositions. Adaptive Seeds, on the other hand, choose matches based on their rarity, ensuring that the number of matches increases linearly with sequence length. This approach is implemented in the LAST software, which can handle large, non-uniformly composed sequences quickly and accurately. The authors demonstrate the effectiveness of Adaptive Seeds through various tests, including comparisons of human and chimpanzee Y chromosomes, malaria genomes, and protein sequences. The results show that Adaptive Seeds can achieve better performance in terms of both sensitivity and running time compared to traditional fixed-length seeds.The article introduces a novel method called Adaptive Seeds to improve the efficiency and sensitivity of comparing large genomic sequences. Traditional methods like BLAST use fixed-length seeds, which can be inefficient for sequences with non-uniform nucleotide compositions. Adaptive Seeds, on the other hand, choose matches based on their rarity, ensuring that the number of matches increases linearly with sequence length. This approach is implemented in the LAST software, which can handle large, non-uniformly composed sequences quickly and accurately. The authors demonstrate the effectiveness of Adaptive Seeds through various tests, including comparisons of human and chimpanzee Y chromosomes, malaria genomes, and protein sequences. The results show that Adaptive Seeds can achieve better performance in terms of both sensitivity and running time compared to traditional fixed-length seeds.