Toward a Wikipedia model of drug discovery
medium.com
medium.com
The FDA has continuously increased the standards required to get a drug approved. Take a look at Librium (chlordiazepoxide) for example. It was the first benzodiazepine discovered. The molecule was first synthesized in 1956. The anxiolytic effect was discovered in 1957. FDA approval came only 3 years later in 1960 [1].
There is no way that would happen today. The truth is, as data requirements increase, so do costs. Increasing costs make development of certain drugs unviable from a commercial perspective. If you're trying to come up with a better cholesterol drug now, you're expected to do two multi-year (2-4 yr), 1000+ patient clinical studies. That alone could run you $500M. Even though is plenty of room for improvement in treating hypercholesterolemia, companies are shying away from pursuing it due to the unattractive economics.
And as for the limitations of small molecules, I agree that current drugs cover a lot of space right now ("we've picked all the low hanging fruit"). However, what drive new drug is new science that determine disease mechanisms. Many of the latest immunotherapy cancer drugs that are coming out (and resulting in very impressive outcomes) are small molecules.