A researcher assumes a model $X \sim \text{Exp}(\lambda)$. To estimate $\lambda$, they use an M-estimator derived from the score function $\psi(x, \lambda) = 1 - \lambda x$. What is the asymptotic variance of $\sqrt{n}(\hat{\lambda}_n - \lambda)$?