Observed arrival · 2026-09-10
RAG Patterns, Side by Side
A hands-on comparison of seven retrieval-augmented generation designs, tested against one reader's bookshelf and a fixed set of questions.
Field notes
The evaluation corpus combines structured and unstructured material: reading dates, ratings, page counts, reviews, Kindle highlights, recommendation links, and photographed pages. The test questions deliberately include filters, contradictory signals, images, and relationships between books, exposing cases where similarity search alone may be insufficient. Scores use a nine-point scale with half-credit for partial answers, while cost is compared with a naive baseline. The page identifies two measured patterns and labels five others as predictions.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue