One point that stuck with me is that nearly everyone attempts to forecast the future by carrying out current trends into the future with a linear graph. For example, here is a natural gas consumption graph from EIA which shows a steady increase in consumption through 2040.
As Tony pointed out, successful technology is adopted on an S-curve, which makes prediction especially tricky since there is a long period of slow adoption followed by very rapid adoption.
Any rate of increase or decrease is ultimately unsustainable. I wonder how much lower the cost of PV and battery production can go before reaching the limits? Then again, if the future involves robots doing the vast majority of all work, then the cost of most goods should approach $0.