It's an average for the network, not an actual expense you pay to run a transaction. Stakers are the ones mainly using that energy. If you're just a regular user and send a transaction, you just spend the tiny bit of energy needed to make a cryptographic signature and transmit it.
I think the issue here is that 35Wh per transaction is still a million times less efficient than traditional databases. The MacBook example is just meant to give a frame of reference for how much 35Wh is.
Sure, no argument there. I'm really glad proof-of-stake is close. It's just that GP made it sound like issuing three transactions would actually drain your battery, so I wanted to clarify that.
