Hello All,
I remember reading a blog post about solving a system of charged particles with differentiable programming (Flux with Zygote perhaps) using a gradient based optimization method instead of explicitly solving the dynamics of the system - and I cannot find it anymore. There was an animation with dot sizes corresponding to charges, it was really nice.
If this rings a bell, please tell me!
Thank you, but that’s not it unfortunately.
Hmm, this post does miss a really essential point about the physics; simulating the dynamics of an N-body system of charges is very much not equivalent to minimizing the energy function. At least, not without some kind of dissipation in addition to the normal Coulomb force between charges. If you imagine the charges are immersed in a very viscous medium the analogy between gradient descent and dynamics would be more reasonable.
Yes, this is it, thanks 
I now feel stupid for not finding it at fluxml.ai