What we're building
No code today, this is the last lesson, and it's a bridge, not a capstone. lesson.py prints a summary of five specific limits this course ran into along the way, each paired with the RAG architecture later in this series built to address it. Nothing here is new information, every limit was already demonstrated hands-on in an earlier lesson; this just names them together, in one place, before this course ends.
Why this matters
The rag-fundamentals course mapped nine RAG architectures in prose, as a conceptual survey. This course took the simplest one, Naive RAG, and built it by hand, which meant actually hitting its limits rather than just reading about them: Lesson 16's multi-hop question genuinely failed with k=1. Lesson 12's scoped search genuinely returned a confidently wrong answer. That hands-on failure is worth more than an abstract description of "Naive RAG has limitations," because you've now watched each one happen on purpose, with your own code, against a document you can read yourself.
Every architecture named below solves a real problem this specific course ran into, not a hypothetical one.
Where to go from here
Naive RAG is the foundation every one of those architectures builds on, not a discarded first draft. Hybrid RAG (this series' next course) adds keyword search alongside the embedding-based search this course already built; Graph, Corrective, Agentic, Multimodal, and Adaptive RAG each add one specific capability on top of the same chunk-embed-retrieve-generate shape this course started from. Understanding this course thoroughly, rather than skimming it, is what makes every one of those additions legible instead of magical.
Congratulations on completing the course.