Defend One Architecture Across Six Contexts
Architecture is the set of boundaries that still hold when the scenario changes, a tool fails, and the evidence is incomplete. Defend architecture choices across all five CCAR-F domains. Adapt one decision method to the six public scenario contexts without memorizing one topology. Implement deterministic checks for orchestration, tools, Claude Code, structured output, and context reliability. Build failure packets that test partial results, stale state, unsafe tools, and invalid output. Produce a reviewer-ready architecture packet with explicit tradeoffs and escalation. An architect prepares six diagrams for six expected use cases. The support diagram uses an agent loop. The code diagram uses Claude Code. The research diagram has subagents. The extraction diagram uses JSON. During review, the architect cannot explain why one step is a tool and another is a subagent. The diagrams omit retry semantics, configuration scope, partial results, source versions, and human authority. Each design works only on its happy path. Scenario-based architecture questions test transfer. The names and business details change, but the same decisions recur: What sequence is deterministic, and what choice requires model reasoning? Which context should own each concern? Which tools are visible, authorized, and retryable? Where does shared Claude Code guidance live? How does a typed result become semantically and evidentially valid? What state survives failure, compaction, resume, and human handoff? This capstone builds…
Defend One Architecture Across Six Contexts: Architecture is the set of boundaries that still hold when the scenario changes, a tool fails, and the evidence…
This free lesson is part of the AI Engineering from Scratch curriculum. Read the full explanation, run the lesson code, and verify the result in the interactive reader or from the repository source.
Browse the complete course catalog or open this lesson on GitHub.