Why this architecture is bulletproof
If you look at the terminal output when running this script, you will see
exactly why this works so well for AI integration:
Validating DAG containing 6 nodes...
-> Graph validation passed. No missing dependencies.

--- AI COMPILATION ORDER (Bottom-Up) ---
1. Build: AST_COMPILER (Leaf Node)
2. Build: MARKDOWN_PARSER (Leaf Node)
3. Build: GRAPH_BUILDER (Needs: MARKDOWN_PARSER)
4. Build: CONTEXT_HYDRATOR (Needs: GRAPH_BUILDER)
5. Build: TOPOLOGICAL_SORTER (Needs: GRAPH_BUILDER)
6. Build: LLM_ORCHESTRATOR (Needs: CONTEXT_HYDRATOR, TOPOLOGICAL_SORTER)

Notice how LLM_ORCHESTRATOR is listed last. Even though it was written at the
bottom of the original markdown, the algorithm mapped the dependencies and
mathematically proved that the AI cannot generate the code for
LLM_ORCHESTRATOR until CONTEXT_HYDRATOR and TOPOLOGICAL_SORTER have been fully
generated and written to disk.