Davidad's Provably Safe AI Architecture - ARIA's Programme Thesis — AI Alignment Forum