I'm thinking about adding something on achieving performance at data center scale.
It's not software, but the Jupiter networking papers from Google really catch my attention.
And I'd love to have something on the distributed in-network computation of inference a la the new Nvidia Vera Rubin.