Resources

A curated library of books, essays, papers, and tools that shaped how I think about growth systems, discoverability infrastructure, and the AI-search era, annotated with operator commentary, not just bibliography.

Everything here earned its place. If a book is missing, it’s either because I haven’t read it yet, or because reading it didn’t change anything I do.

Some book links are Amazon affiliate links, which means I earn a small commission if you buy through them. It costs you nothing extra, and nothing is on this list because it pays. Every book here is one I have actually read.

PaperCore2020

Dense Passage Retrieval for Open-Domain Question Answering

Karpukhin et al

The paper that showed retrieval works on learned embeddings alone, beating BM25 by 9 to 19 percent. This is the mechanism that ended keyword matching as the whole story.

PaperCore2020

ColBERT: Late Interaction over BERT

Khattab and Zaharia

Scoring at token granularity rather than document granularity. This is the technical reason assistants retrieve passages instead of pages, and therefore the reason a section has to make sense on its own.

PaperRecommended2021

BEIR: zero-shot evaluation of retrieval models

Thakur et al

Eighteen datasets, ten retrieval systems, one uncomfortable finding: BM25 is a hard baseline to beat out of domain. Useful antidote to the assumption that embeddings win everywhere.

PaperRecommended2022

ColBERTv2: retrieval via lightweight late interaction

Santhanam et al

The follow-up that made late interaction affordable, cutting the storage cost six to tenfold. Read after ColBERT if the first one convinced you.

PaperCore2023

Lost in the Middle: how language models use long contexts

Liu et al

Models attend to the beginning and end of a context window and lose the middle. If your answer is buried in paragraph nine, it may as well not be there.

PaperRecommended2021

Pretrained Transformers for Text Ranking: BERT and Beyond

Lin, Nogueira and Yates

Book-length and worth it. The clearest survey of how neural ranking actually works, written for people who want the mechanism rather than the metaphor.

EssayRecommended

How LLMs retrieve and cite (Perplexity Engineering blog)

Perplexity Engineering

Operator-level writing from the team building the retrieval layer. The closest you’ll get to ground truth.

NewsletterCore

One Useful Thing (newsletter)

Ethan Mollick

The most consistently useful operator-level writing on practical AI. Subscribe and read every issue.

Know a resource that should be here?

I’m always looking for sharp writing on growth systems, AI-search, and operator practice. If something genuinely changed how you work, I want to read it.

Send a recommendation