Enviar por SMS: Pseudo-derandomizing learning and approximation