Anthropic's AI-Native SDLC Playbook Praised but Seen Missing Architectural Governance Layer

Anthropic last week published an AI-native software development lifecycle playbook arguing that code generation is no longer the bottleneck in delivery, but that surrounding stages like planning, review, and release still operate at human speed. The playbook proposes a structured artifact chain — including intent, spec, and plan documents — so human attention is focused at stage gates rather than spent reconstructing context repeatedly. It outlines a hierarchy of controls, from advisory skills to deterministic hooks, with the guidance that anything that must always hold requires a deterministic backstop. However, one analysis argues the playbook stops short of addressing architectural governance, such as database choices, dependency rules, or module boundaries, which cannot be reduced to simple path-based hook predicates. The missing layer, the argument goes, involves compiling Architecture Decision Records into active, scope-aware controls that can evaluate whether a proposed change violates a recorded architectural decision.