Working on an Emacs-like editor that uses Clojure instead of Emacs Lisp. It has a C kernel and then uses libsci (Small Clojure Interpreter, built with GraalVM, so it has no Java dependency at runtime).
This is a contrarian take: instead of open source models running on private infrastructure being the dominant mode of AI, they instead argue that the returns to scale from cacheing intermediate reasoning steps could provide enough of a performance advantage to centralized models.
He came up with a framework where time is real (assumed), and then the laws of physics evolve with time, and then goes on to develop an evolutionary theory of universes, where universes reproduce by producing black holes, which spawn baby universes, with slightly different laws of physics. He then predicts that it should tend to produce universes that are optimized to produce black holes.
It's still worthwhile to look at 10x and 100x concentrations since these things bioaccumulate. Whatever negative effects are happening at 1x should be studied as we crank that concentration up. Might be fine now, but in 100 years? We should probably have an idea how the harm/effects scale
I would love to know more about Jank, from what I read, it transpiles to C++ right?