Jay Bailey comments on Levelling Up in AI Safety Research Engineering