Yes, the latter- you can take either side ad absurdum, though the teaspoon side is easier to mischaracterize. Single-variable optimization is easy; multivariate is much harder especially when wetware is involved.