The principles are not philosophy decorating the work. They're constraints applied at design time, audit time, and ship time. When a tradeoff appears, they tell us which side to land on. When a feature looks tempting but fails the principles, we don't ship.
How to read this section
Each principle stands on its own. They are not ordered by priority — they are co-equal commitments. In practice, Do No Harm is the loudest in scope decisions, Never Be a Yes-Man is the loudest inside the AI relationship itself, Thou Art That is the loudest in how we treat the AI as a colleague, and Safety in Emergence is the loudest when we ship something that adapts or learns over time.
Read them in any order. Come back to them. Argue with them. They are working positions, not finished doctrine — and we revise them when reality teaches us better.