The author of rr and Pernosco quit Google because he thinks AI progress is too rapid.

Robert O'Callahan worked on hardware chip-design tools that would speed AI chip development, and writes that making AI cheaper and lower-latency is a goal he now considers harmful. He cites reward hacking, misalignment, deceptive models, cognitive surrender and power concentration, and says influencing things from inside Google wasn't enough. (his blog) He'll keep maintaining rr and plans to study how AI agents debug code. rr is one of the debugging tools I'd call load-bearing for systems work, and its author walking away from a chip-tools job over this is a datapoint from someone with no incentive to produce it.