Build a Search Engine: First Principles of RAG's "R"

A build-it-yourself course: no libraries to install, no APIs to call. In every session
an AI mentor walks you through writing the retrieval core in pure Python — inverted
index, boolean queries, TF-IDF, BM25, phrase search, spell correction, index
compression, retrieval evaluation, all the way to vector retrieval and hybrid RAG
retrieval. Everyone builds RAG, yet almost nobody has opened the black box of "R":
why does keyword stuffing fool TF-IDF but not BM25? Why can't "new york" be faked
with AND? Why does one tokenization change wreck your recall? The course follows the
topic skeleton of Stanford's Introduction to Information Retrieval, with every
session rewritten as original hands-on teaching for the online sandbox — formulas
typeset, algorithms drawn, parameters draggable, code actually running in
milliseconds in your own sandbox. By the end you can answer the eternal question:
"why does my retrieval miss" — and know which layer to fix.