Scientists have figured out how to harness Brownian motion -- literally the thermal energy of individual molecules -- to make electricity, by cleverly connecting diodes up to pieces of graphene, which are atom-thick sheets of Carbon. The team has successfully demonstrated their theory (which was previously thought to be impossible by prominent physicists like Richard Feynman), and are now trying to make a kind of micro-harvester that can basically produce inexhaustible power for things like smart sensors.
The most impressive thing about the system is that it doesn't require a thermal gradient to do work, like other kinds of heat-harvesting systems (Stirling engines, Peltier junctions, etc.). As long as it's a bit above absolute zero, there's enough thermal energy "in the system" to make the graphene vibrate continuously, which induces a current that the diodes can then pump out.
Original journal link: https://journals.aps.org/pre/abstract/10.1103/PhysRevE.108.024130
Hey, I'm just an aero/structural engineer - this microscopic and quantum level stuff is well outside of my daily practice, too. The theory (of which I am innocent of all detail) says that this shouldn't be possible - using Brownian motion as a source (directly or as a pump). If this is an end-run around classic physics, that's okay, as long as the overall energy balance can be shown to be maintained.
Edit: Usually in threads like this I hope to say something wrong, or apply the wrong principle, and then someone who is an expert comes in and corrects me. Then I go look up whatever it is they say and I get to learn something new for the day. Either that or someone who knows more than I do agrees with me and expands on the description in a really insightful way, and I get to learn something more in depth that day.