Concrete example first
Let (R=\mathbf{Sh}(T,J)). Start with (X_0=\emptyset). Suppose your closure rules say:
- include the terminal object,
- include (H) and (H^*),
- include objects needed to interpret RTL connectives on included objects.
Define (X_{n+1}=UG(X_n)). When (X_{n+1}=X_n), you reached a fixed point.
Formal lesson content
The generative closure operator is [ UG:P(\mathrm{Ob}(R))\to P(\mathrm{Ob}(R)). ] It is monotone and inflationary, so iterative chains from (\emptyset) increase until a least fixed point is reached (under the paper’s set-theoretic setup).
GFRTU is that least nontrivial fixed point: the smallest object collection closed under the universe’s semantic and dynamic requirements.
Why this lesson matters
Without (UG), you can describe pieces but not the minimal whole universe. (UG) is what converts local rules into a complete, initial semantic environment.
Typical pitfalls
- Adding objects “because they seem useful” instead of because closure forces them.
- Defining closure rules that do not preserve monotonicity.
- Forgetting that RTL interpretation also imposes closure obligations.