7 Chapter Quizzes · 350 Questions
One quiz per chapter, 50 questions each, all written directly from the chapter content. Take the quiz right after finishing a chapter — it's far more effective than re-reading. Once you've completed all seven chapters, there's a comprehensive cross-chapter final exam waiting for you.
LLM Fundamentals50 questions
Three training stages, tokens & tokenization, parameter freezing, temperature sampling, four hallucination types, RAG & four mitigation strategies
AI Harness50 questions
Context window & overflow, prompt engineering, injection attack & defense, tool calling & MCP, KV cache & cost optimization
Practice · From Demo to Product50 questions
Image generation productization, Agent deadlock & failsafes, context compression, long-term memory, multi-Agent collaboration, MCP ecosystem
AI Engineering Design Patterns50 questions
Workflow vs Agent, the three context levers, ACI tool design, evaluation methodology & three Grader types, sandbox isolation
Harness & Self-Improvement50 questions
Three Harness design patterns, automatic evolution of context engineering, workflow search, recursive self-improvement & the seven checkpoints
Anatomy of Grok Build50 questions
Runtime loop & three Actors, Compaction, tool approval, dual-path memory retrieval, five sandbox types, MCP integration
Vibe Coding Methodology50 questions
Flow control & manual breakpoints, quality floor, rejecting incremental delivery, documentation practices, three gates for destructive operations
After each submission, your results are recorded by topic. Topics where you miss the most questions will appear as weak spots in your Miyang profile, with direct links back to the relevant lessons. So don't fear a bad score — a bad score tells you exactly what to review.
All questions are written from the course content, and every question includes an explanation. If you spot an issue with a question, click "Issue with this question" on the quiz page to report it directly.