That is a valid criticism of the MPI system and maybe an indication that MPIs should build smaller departments headed by more junior researchers at the group leader level (permanently).

I blame the reviewers and culture on the algorithm theory side. I wanted to name my first paper “Trix with time” as a pun. My advisor and later, reviewers talked me down to a generic bland “informative” title. The PL theory community seems to have better paper titles.

The lean Zulip has an active jobs channel. I believe the Lean + distributed systems is still a small community compared to other provers. The loom+veil combo is currently the best tool.

As an aside, i am on campus (SB), work on lean, and work on distributed algorithms (not systems). Happy to chat.

Glad to meet a fellow pundit

How should I read this?

There is some illuminatingchatter on the lean Zulip about this. Including the authors.

I am on an infinite ad loop with the link. Is it just me?

It would have to be faster in GPU hardware and they would have to be able to budget for it, given how expensive GPUs are at scale.

I recall that there is some research into sparsifying neural network weights as well. This sort of speed up might be cheaper.

Maxwell’s equations were perfectly sufficient for radios.

I think I recognise Indian English there. “I’d love a pointer” is Indian corporate-speak.