Roman Leventov comments on Incubating AI x-risk projects: some personal reflections