Online edition for BLOGE 0.9.8-RC1 · facts verified 2026-09-15 · 中文

Arc 7 Recap — Graduation

Chapters 33–34 end the book with one standard: graduation is not “using every BLOGE feature.” It is choosing the smallest model that makes the important behavior visible—and knowing when to stop.

The final decision loop

Step	Question	Output
STOP	Is this only sequential code, one remote call, or CRUD?	A documented reason not to use BLOGE
PROBE	Is orchestration risk real but still uncertain?	One thin vertical slice and a comparison baseline
ADOPT	Do dependencies, waiting, recovery, or evidence justify a graph?	An ADR with constraints, costs, and exit conditions
GRADUATE	Can another team operate and challenge the result?	Runnable system, tests, diagrams, evidence, and limits

Chapter 33 starts from constraints rather than features. Chapter 34 then offers three capstone levels: foundation, advanced, and graduation. The final level requires six artifact classes—story, executable flow, behavior tests, architecture view, operational evidence, and a limits statement. Optional features earn no credit unless they resolve a stated constraint.

Try It: Take One Feature Away

Choose your capstone design and remove its most impressive BLOGE feature. If the business story, recovery boundary, and evidence remain clear, leave the feature out. If clarity collapses, write the exact constraint that justifies putting it back. That sentence is the heart of your final ADR.

The book's last claim

You now have a method, not a catalog: model dependencies, name lifecycle owners, control time and inputs, bind claims to evidence, and stop at the strongest conclusion the evidence supports.