Take the 5×4 coordinate grid and add a 'mine' column of 0s (safe) and 1s (mine). Seed with np.random.seed(0) immediately before drawing (inside solve, right before np.random.binomial: it's global, shared state, so seeding anywhere else won't reproduce the same numbers), giving each square a 40% chance of being a mine. Return the DataFrame.