The map: where everything lives, and how to stay current

A course that promised you wouldn't need other material owes you an honest closing: this field moves, and the durable skill is knowing where the ground truth lives and what to check when things change. This final lesson is that map — the canonical sources, bookmarked by purpose — followed by the recap of what you now carry.

The canonical sources, by question

• "How does the API feature work?" — the Claude Developer Platform docs are the single source of truth for the Messages API, tools, caching, batches, files, and structured outputs. When this course and the docs ever disagree, the docs have moved and they win.
• "What models exist right now, at what price?" — the models overview and pricing pages, plus the Models API from lesson one for runtime truth. Deprecation notices live there too; treat them as calendar items, not news.
• "How do I do X with Claude Code?" — the Claude Code docs cover the CLI, IDE integrations, CLAUDE.md, skills, hooks, subagents, headless mode, and the Agent SDK. The release notes are worth skimming monthly; the tool grows features faster than word of mouth spreads them.
• "Is there an MCP server / how do I build one?" — the MCP site holds the spec, SDKs, and quickstarts; the public registry is where you check before building an integration that may already exist.
• "Why is the API behaving strangely right now?" — the status page, before you debug your own code for an hour.
• "What's the recommended pattern for this use case?" — Anthropic's cookbook repositories and engineering blog carry worked examples and the pattern essays this course distilled; new model releases ship migration guides that name every breaking change — read those the day you plan an upgrade, with your eval suite as the referee.

Model knowledge goes stale at the edges first: exact prices, exact context windows, beta feature names. The structure — tiers, the capability–cost trade, the prefix rule, the loop, the guardrails — is what this course taught, precisely because it survives releases.

What you now carry

Walk the arc once, because it is an architecture, not a topic list. You choose models by evidence and re-choose them per route. You know every surface — apps, Projects, Artifacts, Claude Code, Chrome, API, SDK — and route work to the right one. You speak the API's native shapes: typed blocks, system prompts that steer, schemas that guarantee, tools that act, caches that pay, streams that deliver, stop reasons that tell the truth. You put anything in front of the model — images, PDFs, whole corpora with citations, live screens. You know when a task deserves a workflow, an agent, or a team of agents, and how context engineering keeps any of them sharp over long runs. You extend everything through MCP, drive Claude Code as a specialised, automated, parallel instrument, and re-drew the lifecycle with agents at every stage and humans at every judgment gate. And you hold the two disciplines that make the rest trustworthy: evaluation, which turns opinion into measurement, and security, which assumes the attack and bounds the damage.

If one thing sticks

The teams that win with Claude are not the ones with secret prompts. They are the ones who specify precisely, verify relentlessly, and build harnesses that make capability safe to use at speed. Models will keep improving underneath you — that is the one prediction this course makes without hedging — and every practice here is designed to get more valuable as they do. Go build.