Refine GRPO evaluation details and clarify model performance comparisons in Blog.md

Revise reasoning for choosing SFT + agentic loop over SFT + GRPO in deployment documentation

Add initial blog post detailing ArchitectureEnv and its design approach