Choosing a development loop

Documentation status: guide — see Maturity and evidence.

Choose the starting loop according to project risk, not ideology.

Project situation	Recommended start	Why
Well-known process with short delivery cycle	Model-first	Build a working conceptual application quickly.
Existing CRUD or workflow application to modernize	Model-first	Replace implementation structure progressively with concepts, actions and projections.
Ontology emerging during product discovery	Model-first, then H-Logic synthesis	Let the running application reveal useful distinctions before consolidating them.
Rule-heavy or regulated domain	H-Logic-first	Validate vocabulary, constraints and deductions before generating runtime artifacts.
Knowledge integration from heterogeneous sources	H-Logic-first	Resolve overlap and contradictions before choosing projections.
LLM-assisted knowledge acquisition	H-Logic-first	Treat generated knowledge as a candidate that must be parsed and validated.
UI-centric prototype with simple semantics	Model-first	Avoid logical complexity until it creates concrete value.
Decision support requiring explanations and inference	H-Logic-first	The symbolic model is part of the product contract.

Switching loops

A project can move from model-first to H-Logic-first when cross-model coherence becomes difficult to review. It can move back toward model-first once the semantic core is stable and the main work becomes views, services, processes and deployment.

A useful mixed cycle is:

business text -> models -> runtime feedback -> H-Logic synthesis -> coherence review -> regenerated model skeletons -> runtime feedback

Do not force every developer to understand parser internals or inference algorithms. The methodology should expose validated conceptual operations, not engine internals.