There is one problem with a y-combinator approach to capital intensive work: someone has to be in charge. There is 'a' party responsible for delivering the product to the customer, and that drives everything else.
Put ycombinator in charge of a shipyard and a lot of cool stuff will happen, but delivering a finished boat on time and on-cost is not one of them.
The past 100 years are a great example of how a lot of change can happen in a short amount of time.
There are some things that you can automate -- but a lot of things in the oil & gas industry cannot be automated. Specifically, the leg work required to get things started & keep them maintained. Automating those sorts of things requires a pretty advanced AI and I don't see that coming around the corner anytime soon.
Automation can get us close. Already has, but this just moves the labor out of the field and into a cube.
But (I think others have pointed out downthread by now) capital intensive projects are always going to need someone to go out and kick the tires, work a nut loose, apply some ingenuity.
If you don't have those guys hanging around to kick the tires and whack a lug nut things will stop until you send a guy out to do that.
The probes we have on Mars .. you can wait a few months for the guys at mission control to work out a problem. You can't do that with an oil well.
Optimal size will still depend on the industry.
Seventy guys can operate a dropbox.
You'll need a few thousand guys, minimum, to run a shipyard, assemble tractors, extract oil.
Doesn't mean they are inefficient - just that some tasks require a lot of capitol and manpower.