>Papers

• [11/19/25] PyTorch: An Imperative Style, High-Performance Deep Learning Library
• [08/16/25] DistServe: Disaggregating Prefill and Decoding for Goodput-optimized Large Language Model Serving
• [07/23/25] Inference without Interference: Disaggregate LLM Inference for Mixed Downstream Workloads
• [07/10/23] Efficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphs
• [07/01/23] Bitcoin Whitepaper
• [06/23/23] MapReduce: Simplified Data Processing on Large Clusters

>Engineering

• [02/14/26] A guide to PyTorch's CUDA Caching Allocator
• [11/20/25] FineWeb: decanting the web for the finest text data at scale
• [11/18/25] Post Training Ensemble vs. Singular Model Approaches with Tinker
• [09/12/25] Improving Cursor Tab with online RL
• [09/11/25] Claude Memory: A Different Philosophy
• [09/06/25] How io_uring and eBPF Will Revolutionize Programming in Linux
• [06/13/25] Actor Model in Rust with Tokio
• [06/05/25] LLMs & Elixir: Windfall or Deathblow?
• [08/16/24] Dealing with rejection (in distributed systems)
• [03/05/24] Doing Microservices Completely Wrong: Distributed Monoliths
• [09/22/23] How Cloudflare built rate limiting capable of scaling to millions of domains
• [09/22/23] Scaling up the Prime Video audio/video monitoring service and reducing costs by 90%
• [09/21/23] Scaling the GitHub API with a sharded, replicated rate limiter in Redis
• [09/13/23] Why Nvidia's AI Supremacy is Only Temporary
• [09/08/23] How Discord Stores Trillions of Messages
• [09/01/23] Scalable Machine Learning at Cloudflare
• [08/29/23] Reducing Cold Start time for LLM Inference at Scale.ai

>Lectures

• [03/01/26] Building Machine Learning Systems for a Trillion Trillion Floating Point Operations
• [02/27/26] Stop Writing Dead Programs
• [02/14/26] Monarch: A Distributed Execution Engine for PyTorch
• [11/14/25] Making GPUs Actually Fast: A Deep Dive into Training Performance
• [07/23/25] Accelerating LLM Inference with vLLM (and SGLang)
• [07/23/25] Fast LLM Serving with vLLM and PagedAttention
• [08/03/24] How We've Scaled Dropbox
• [08/01/24] The tragedy of systemd
• [10/11/23] Noam Chomsky on Modern-Day American Imperalism: Middle East and Beyond
• [10/11/23] Noam Chomsky on Power and Ideology
• [10/09/23] Yale ENGL 291: Thomas Pynchon, The Crying of Lot 49

>Domestic

• [06/10/24] New York's Governor Is Inept
• [03/05/24] Global Liquidity Is Ready To Push Assets Higher
• [01/26/24] Cooler Inflation Keeps Door Open for Rate Cuts This Year
• [01/22/24] Companies rush to bond market in record $150bn debt splurge
• [01/21/24] This Isn't Your Father's S&P 500. Don't Worry About Valuations.
• [10/09/23] Americans Are Still Spending Like There's No Tomorrow
• [10/09/23] The Free-Money Experiment Is Over
• [10/09/23] Corporate America Is Ignoring Jay Powell and Bingeing on Debt
• [10/09/23] The 5% Bond Market Means Pain Is Heading Everyone's Way
• [09/26/23] The Hedge Fund That Made a Killing Betting Against Lina Khan
• [08/27/23] Quiet Cutting

>Global

• [06/10/24] How the Talmud Became a Best-Seller in South Korea
• [10/19/23] A Financial Crisis in China Is No Longer Unthinkable
• [10/19/23] China's loans pushing world's poorest countries to brink of collapse
• [10/12/23] Lessons from the 1970s Energy Crisis Can Help Prevent the Next One
• [10/12/23] What the West Loses by Trading With Dictatorships
• [10/12/23] The Bank of Japan should stop defending its cap on bond yields
• [10/12/23] Rising bond yields are exposing fiscal fantasy in Europe
• [08/29/23] Xi's Age of Stagnation
• [08/29/23] Germany is Losing its Mojo

Back to Home