I'm no expert and this article is clickbaity, but I think the basic claim here does not sound crazy. If regular freestanding graphene does have a natural "ripple" of current caused by thermal energy[0], it seems reasonable to place an electrode very close to it and harvest the work done when charges move relative to each other. The circuit described in the youtube video[1] does seem like it would work, if the "ripple" is predictable and slow enough to switch the circuit in phase with.
It's not "limitless" obviously; it's just converting heat energy to electricity. But if this principle works, and if it can be implemented in the real world (which seems like a big if - I can imagine it being fouled up by temperature fluctuations, physical effects on the graphene like sound vibrations, etc) it sounds like it might be very useful for things like sensors with very low power needs.
0: The same team claimed to describe this in a 2014 paper - see https://phys.org/news/2014-04-track-ripples-freestanding-gra...
> The system reaches thermal equilibrium and the rates of heat, work, and entropy production tend quickly to zero. However, there is power generated by graphene which is equal to the power dissipated by the load resistor.
Looks like the article is also on ArXiV: https://arxiv.org/abs/2002.09947
https://scholar.google.com/scholar?oi=bibs&hl=en&cluster=103...
Is it really a closed system at equilibrium?