One of the hardest decisions an agent gets to make is to choose between

and

Otherwise put, are we to allow evolution to proceed?

The post continues after this message:

‘Don’t mend it if it’s not broken’ makes perfect sense for something which already exists. And works ‘good enough for us’.

But when you get to make something a second time? Or even the one million-th time?

What keeps you from ‘experimenting’?
If the expected marginal benefits are greater than the potential marginal costs, of course!