Yes a "look" is basically determining what algorithm to apply to the cube based on the current state.

1 look last layer means that after the first two layers are solved, you would know an algorithm to apply to solve the cube based on the state of the last layer.

With the typical speed solving method (CFOP), the last layer requires 2 "looks" (one for orientation of the last layer, one for the permutation)

Standard speed cube method uses 9 “looks” each followed by an optimized algorithm. This development (1LLL) combines the final two looks (each normally requires choosing one of roughly 40 unique algorithms ) into one monster step with 4000. The number is big because there’s many possible embeddings. Each look is only at the next subsection of the cube you plan to solve.