Explainers
Claude concepts, explained
In-depth, plain-English explainers for the ideas the Claude certifications actually test. Each leads with a direct answer, goes deep, and ends with the practice that turns understanding into judgment.
Agents vs Workflows in Claude, Explained
The difference is who decides the next step — your code or the model — and the rule is to pick the least autonomy that solves the problem.
ReadMCP (Model Context Protocol), Explained
How an open client/server protocol lets you plug tools, data, and prompts into Claude once and reuse them everywhere.
ReadMCP Authentication & Authorization, Explained
How to secure a Model Context Protocol integration with OAuth and per-user delegated tokens, and why the trap answer hands the server a shared secret or a broad token.
ReadRAG vs Long Context: When to Use Which
A practical decision rule for architects: retrieve when the corpus is large, changing, or access-controlled; stuff the window when it is small, stable, and fits.
ReadTool Use and tool_choice in Claude, Explained
How the tool-use loop works, why descriptions drive selection, and the tool_choice trap the exam loves to test.
ReadClaude Context Windows & Context Engineering, Explained
Why the context window is a budget you curate, not a bucket you fill
ReadStructured Output & JSON with Claude, Explained
Force the shape with tools and schemas, then validate the meaning
ReadThe Message Batches API, Explained
What Batches is, when to use it, and the myth that batch jobs cannot use tools.
ReadObservability & Tracing for Claude Agents, Explained
What to instrument in an LLM agent, why per-span traces matter, and how to tell a model failure from an orchestration failure.
ReadCost Governance & Token Budgeting for Claude, Explained
The levers that control Claude spend at scale — model right-sizing, prompt caching, batching, token budgets, and cost SLAs.
ReadEvaluating Claude: Evals & LLM-as-Judge, Explained
How to actually know a Claude-powered system works — and why a model's confidence score is not its accuracy.
ReadPrompt Injection & Tool Poisoning, Explained
Why untrusted content in the context window is an instruction-injection attack, and how to actually defend against it.
ReadFail-Closed Guardrails & Safety Controls, Explained
Why a safety check that blocks when it errors protects you, and why one that fails open is worse than having no check at all.
ReadAgent Security & Least Privilege, Explained
Why the safest agent is the one you gave the fewest tools, and why the exam's favourite wrong answer adds a guardrail instead of removing a capability.
ReadIncident Response for Claude Systems, Explained
How to handle a production incident in an LLM system the way you'd handle any outage: detect, contain, diagnose, roll back, and learn, from a runbook you wrote before the pager went off.
ReadMigrating Claude Model Versions Safely, Explained
Why the release-day upgrade is a production change like any other, and how to move to a new Claude model without shipping a silent regression.
Read