As far as I'm aware, so don't take this as gospel or anything:
Its total potential to do work within the reference frame of any given system, which is what we're interested in here, doesn't remain a constant. It's like... if you think of it like water: If you have all your water at the top of a cliff in a big reservoir and it runs down into a turbine linked up to a generator, then you can get a lot of work out of that water. But once it's gone down the cliff, despite it still being the same mass of water, your ability to extract useful work from that water is dramatically less. You'd have to pump it back to the top of the cliff to repeat the exercise (at a loss since the system you're playing with is never going to be perfectly efficient, i.e. the second law of thermodynamics - so there'd basically be no point in doing so.)
What we really want are high concentrations of energy; low entropy sources. We're looking for those situations where the water is still at the top of the cliff. A star is fussing down through the various elements, and in the process it emits a lot of radiation during its lifetime. And it's not getting that energy back. Once the elements have undergone the process you'd have to put energy back into them to return them to their earlier states. But that energy's been leaving the star for years, and some of it travels at the speed of light so it's very far away - so... it's just not gonna happen. Stars do go through life cycles and produce various phenomena but always at a loss of their potential to do useful work, always at an increase in entropy.
^^;
