Step through how gradient descent minimizes Mean Squared Error for linear regression — watch the line fit the data as weights update each epoch.