Given a binary matrix, find the largest rectangle made entirely of 1s and return its area. Checking every candidate rectangle directly works but re-examines huge overlapping regions over and over. The faster path reframes each row as a bar chart: for every column, track how many 1s have piled up consecutively ending at the current row. That single row's bar chart is exactly the "largest rectangle in a histogram" problem, solvable in one linear pass with a stack that only ever holds increasing bars — pop a bar the moment a shorter one appears, and its width is exactly how far it could have stretched as the tallest bar in its own rectangle. Running that once per row, keeping a running maximum, finds the biggest all-1s rectangle in the whole grid without ever double-checking the same region.