Flywheels are great. A pound or two of weight only hurts when accelerating, but the stored energy easily negates that penalty.
Shoving air into a random frame tube would be suboptimal because that's really low volume and you're dealing with loss when converting between kinetic and pneumatic energy.
The ability to put energy into the flywheel by pedaling while coasting the crank->rear wheel drive would be pretty useful.
Most casual bicycles aren't very engineered for weight optimization, so a 2 pound flywheel could be offset by shaving weight elsewhere.
I'm hoping battery technology follows the same curve CPUs did with Moore's law and that we'll see extended ranges. Perhaps pneumatics could serve as an intermediary for a pedal assist drivetrain. Inverters are probably best for high torque applications which I believe should be steered away from due to how it can ravage the lifespan of batteries.