Papers We Love

What it is

Papers We Love is both a repository and a community. The repository organises papers into more than 70 topic directories, including distributed systems, databases, data structures, operating systems, programming languages, cryptography, security, machine learning and theory. Papers marked with a scroll icon are hosted in the repository; most are links to external sources because of licensing. A small script can download the linked papers for offline reading.

The community side consists of local meetup chapters around the world, where someone presents a paper and the group discusses it. Many talks have been recorded, and there is a Discord server for discussion.

As of September 2026 the repository has about 110,000 GitHub stars and 6,400 forks, and commits were still landing that month.

Why read papers at all

The ideas that modern infrastructure depends on (MapReduce, GFS, Dynamo, Bigtable, Paxos, Raft, Spanner and many more) came from a modest number of papers. Most are more readable than their reputation suggests. Reading the original gives you the assumptions and limitations, which are exactly what gets stripped out of blog summaries and exactly what experienced engineers ask about.

Where to start

• Distributed systems: begin with the Dynamo, Raft and Google File System papers. They are readable and hugely influential.
• Databases: look for papers on B-trees, log structured merge trees and the classic "Architecture of a Database System".
• Programming languages: "Why Functional Programming Matters" and papers on type systems are approachable.

Work through one paper properly rather than skimming five.

How to read a paper

1. First pass (15 minutes): read the abstract, introduction, section headings and conclusion. Decide whether it is worth a full read.
2. Second pass (an hour): read the whole paper, skipping proofs. Note the problem, the key idea and the evaluation.
3. Third pass (optional): work through the details, and try to explain the design to someone else.
4. Watch a recorded talk from a Papers We Love chapter if the paper resists a first reading.

Weak spots

• It is a list, not a curriculum. There is no suggested order or difficulty rating.
• Some directories are much better curated than others, and some links go stale.
• Recorded talks vary in quality and some are several years old.

Common questions

Do I need a research background? No. Many systems papers were written by engineers for engineers. Start with the classics above.

How much time should I budget? One paper a week, read properly, is a strong and sustainable pace.

Is there a Pakistani chapter? Chapters come and go; check the repository's chapter list for current locations, or start a reading group with colleagues.

Pair it with

• MIT 6.5840's reading list, which gives a structured order through the distributed systems classics.
• Designing Data-Intensive Applications, whose references point to many of the same papers.

Best for: engineers ready to move beyond secondary sources, and anyone who wants to understand where core computing ideas came from.