Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Community-curated directory of influential CS research papers
.claude/skills/brycewang-stanford-papers-we-love-guide/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 71% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 34% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 131% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 33% | 0% |
Papers We Love (PWL) is a community-driven repository of influential computer science research papers organized by topic, with worldwide reading groups. The repository contains direct links to PDFs and summaries for hundreds of landmark papers across distributed systems, programming languages, machine learning, security, and more. A go-to resource for discovering foundational and impactful research.
papers-we-love/
├── distributed_systems/
│ ├── README.md # Curated list with descriptions
│ ├── lamport-clocks.pdf
│ └── raft.pdf
├── machine_learning/
├── programming_languages/
├── security/
├── databases/
├── networking/
├── information_retrieval/
├── artificial_intelligence/
├── concurrency/
├── operating_systems/
└── ... (40+ categories)| Category | Notable Papers | |----------|---------------| | Distributed Systems | Paxos, Raft, MapReduce, Dynamo | | Machine Learning | Backpropagation, Dropout, Attention, BatchNorm | | Programming Languages | Lambda calculus, Type inference, Hindley-Milner | | Databases | B-Trees, LSM-Trees, MVCC, Column stores | | Security | Public-key crypto, Zero-knowledge proofs, TLS | | Networking | TCP congestion, BGP, Software-defined networking | | Operating Systems | Unix, Microkernel debate, Virtual memory | | Concurrency | CSP, Actor model, Software transactional memory |
bash# Clone the repository git clone https://github.com/papers-we-love/papers-we-love.git # Browse categories ls papers-we-love/ # Each directory has a README with curated descriptions cat papers-we-love/distributed_systems/README.md
pythonimport os import glob PWL_PATH = "./papers-we-love" # List all categories categories = [d for d in os.listdir(PWL_PATH) if os.path.isdir(os.path.join(PWL_PATH, d)) and not d.startswith('.')] print(f"Categories: {len(categories)}") # Find papers in a category ml_papers = glob.glob(f"{PWL_PATH}/machine_learning/*.pdf") for p in ml_papers: print(f" {os.path.basename(p)}") # Search across all READMEs for a topic import re for readme in glob.glob(f"{PWL_PATH}/*/README.md"): with open(readme) as f: content = f.read() if re.search(r"consensus|paxos|raft", content, re.I): category = os.path.basename(os.path.dirname(readme)) print(f"Found in: {category}")
python# PWL chapters host monthly meetups worldwide # Find local chapters at paperswelove.org chapters = { "New York": "meetup.com/papers-we-love", "San Francisco": "meetup.com/papers-we-love-too", "London": "meetup.com/papers-we-love-london", "Berlin": "meetup.com/papers-we-love-berlin", # 40+ chapters globally } # Video talks on YouTube # youtube.com/@PapersWeLove — recorded presentations # Each talk: 30-60 min paper walkthrough by practitioner
python# Curate a personal reading list from PWL essential_distributed = [ "Time, Clocks, and the Ordering of Events (Lamport, 1978)", "The Byzantine Generals Problem (Lamport et al., 1982)", "Impossibility of Distributed Consensus (FLP, 1985)", "Paxos Made Simple (Lamport, 2001)", "In Search of an Understandable Consensus Algorithm (Raft, 2014)", "Dynamo: Amazon's Key-Value Store (DeCandia et al., 2007)", "MapReduce: Simplified Data Processing (Dean & Ghemawat, 2004)", ] essential_ml = [ "A Few Useful Things to Know About ML (Domingos, 2012)", "Dropout: A Simple Way to Prevent Overfitting (Srivastava, 2014)", "Batch Normalization (Ioffe & Szegedy, 2015)", "Attention Is All You Need (Vaswani et al., 2017)", "BERT: Pre-training of Deep Bidirectional Transformers (2018)", ]
markdown## How to Contribute 1. Fork the repository 2. Add paper PDF to appropriate category directory 3. Update the category README.md with: - Paper title and authors - Year of publication - Brief description (2-3 sentences) - Why it matters 4. Submit a pull request ### README Entry Format - :scroll: [Paper Title](link) — Brief description. Authors (Year). *Venue*.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 16,002 | 20,847 | +30% | 1 | 1 | 0% | 2,474 | 4,229 | +71% | 0 | 0 | — |
case-02 | fail→pass | 13,162 | 8,426 | -36% | 1 | 1 | 0% | 2,052 | 2,744 | +34% | 0 | 0 | — |
case-03 | pass→pass | 6,790 | 2,909 | -57% | 1 | 1 | 0% | 1,093 | 1,843 | +69% | 0 | 0 | — |
case-04 | pass→pass | 3,734 | 2,261 | -39% | 1 | 1 | 0% | 613 | 1,742 | +184% | 0 | 0 | — |
case-05 | fail→pass | 4,839 | 2,130 | -56% | 1 | 1 | 0% | 733 | 1,690 | +131% | 0 | 0 | — |
case-06 | pass→pass | 14,774 | 5,143 | -65% | 1 | 1 | 0% | 2,511 | 2,315 | -8% | 0 | 0 | — |
case-07 | pass→pass | 10,218 | 7,490 | -27% | 1 | 1 | 0% | 1,864 | 2,857 | +53% | 0 | 0 | — |
case-08 | fail→fail | 4,088 | 5,664 | +39% | 1 | 1 | 0% | 752 | 2,342 | +211% | 0 | 0 | — |
case-09 | pass→pass | 3,573 | 3,527 | -1% | 1 | 1 | 0% | 558 | 1,968 | +253% | 0 | 0 | — |
case-10 | pass→pass | 3,131 | 2,572 | -18% | 1 | 1 | 0% | 464 | 1,761 | +280% | 0 | 0 | — |
case-11 | fail→pass | 9,224 | 4,282 | -54% | 1 | 1 | 0% | 1,677 | 1,936 | +15% | 0 | 0 | — |
case-12 | pass→pass | 10,316 | 2,716 | -74% | 1 | 1 | 0% | 1,486 | 1,748 | +18% | 0 | 0 | — |
case-13 | pass→pass | 22,776 | 3,095 | -86% | 1 | 1 | 0% | 1,996 | 1,898 | -5% | 0 | 0 | — |
case-14 | pass→pass | 8,059 | 2,335 | -71% | 1 | 1 | 0% | 1,273 | 1,768 | +39% | 0 | 0 | — |
case-15 | fail→pass | 8,462 | 4,720 | -44% | 1 | 1 | 0% | 1,233 | 1,646 | +33% | 0 | 0 | — |
case-16 | pass→pass | 17,764 | 1,780 | -90% | 1 | 1 | 0% | 2,877 | 1,637 | -43% | 0 | 0 | — |
case-17 | pass→pass | 11,377 | 1,552 | -86% | 1 | 1 | 0% | 1,626 | 1,596 | -2% | 0 | 0 | — |
case-18 | pass→pass | 10,287 | 2,867 | -72% | 1 | 1 | 0% | 1,499 | 1,828 | +22% | 0 | 0 | — |
case-19 | fail→pass | 4,315 | 1,797 | -58% | 1 | 1 | 0% | 512 | 1,607 | +214% | 0 | 0 | — |
case-20 | pass→pass | 12,799 | 14,587 | +14% | 1 | 1 | 0% | 2,118 | 3,941 | +86% | 0 | 0 | — |
case-21 | pass→pass | 9,507 | 9,714 | +2% | 1 | 1 | 0% | 1,701 | 2,970 | +75% | 0 | 0 | — |
case-22 | pass→pass | 19,484 | 18,962 | -3% | 1 | 1 | 0% | 3,155 | 4,904 | +55% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +27 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.