In the context of the author's comments, you're scaling to meet demand, and delaying experiments with profit. Unless you're serving the bot community, it's hard to imagine how you would scale without demand.
Well, scale can mean "spend time and money building infrastructure that supports lots of traffic" -- doing this before getting the traffic is premature optimization.
But, scale can also mean "get lots of traffic" -- this is growth, not optimization; in this case, trying to optimize revenue before the growth happens could be considered premature optimization.
Why bother making something scale before you know whether there is demand or profit?