DuckDB Internals: Why is DuckDB Fast? (Part 2 Vectorized Execution)

A deep dive into DuckDB's vectorized execution: how processing data in batches of 2,048 rows, instead of one row at a time, makes queries fast.

greybeam.ai

DuckDB Internals: Why is DuckDB Fast?

Walk through DuckDB's internals: how it skips serialization overhead, parses and optimizes SQL, and stores data in columnar row groups with zone maps.

greybeam.ai

PostgreSQL's MVCC is bad. So is everyone else's.

The critics are right: PostgreSQL's MVCC bloats tables and amplifies writes. Oracle, InnoDB, SQL Server, MongoDB, LSM stores, and etcd answer the same four design questions differently, and each pays…

boringsql.com

The Bottlenecks for AI-Driven System Design: why Principal Engineers are not (yet) obsolete.

I’ve been building systems, teaching system design, and writing papers about novel designs for a couple of decades now. In my view, agents are magical self-writing distributed programs: my personal...

maheshba.bitbucket.io

How MVCC and Transactions Work in RocksDB - Artem Krylysov

How MVCC and Transactions Work in RocksDB

artem.krylysov.com

https://www.usenix.org/conference/osdi26/presentation/vickers

usenix.org

Aurora DSQL: Scalable, Multi-Region OLTP

Aurora DSQL is a serverless SQL database designed for cloud-scale transaction processing with multi-region active-active capabilities. Built on a disaggregated architecture, DSQL separates compute, storage,...

arxiv.org

Shard your locks: benchmarking 6 Go cache designs

I built the same in-memory cache six ways with the Go standard library and benchmarked them across read/write mixes and core counts. Lock striping wins by up to 8×, sync.RWMutex turns out to be a trap,...

strebkov.dev

Breaking Database Lock-in: Agentic Regeneration of High Performance Storage Readers for Database Bypass

Analytical workloads operating on data stored in external database systems face a fundamental bottleneck: data access is guarded entirely by the database driver, like JDBC or ODBC, forcing all reads…

arxiv.org

Introducing Meerkat- an experiment in global consensus

Cloudflare Research is building a global consensus service called Meerkat that uses a new consensus algorithm called QuePaxa. We plan to use Meerkat to build a strongly consistent, fault-tolerant key-value...

blog.cloudflare.com

The Log is the Agent: Event-Sourced Reactive Graphs for Auditable, Forkable Agentic Systems

Most agent frameworks are built around the language model: a conversation loop comes first, then tools, then rules, and finally a logging layer bolted on for observability, with state persisted as…

arxiv.org

Zalando Engineering Blog - Client-Side Load Balancing at a Million Requests Per Second

How we built an in-process client-side load balancer for a million requests per second of internal fan-out traffic, what we layered on top (N-ring fade-in, occupancy-based bounded load, and...

engineering.zalando.com

https://cacm.acm.org/federal-funding-of-academic-research/the-llvm-compiler-infrastructure/

cacm.acm.org

https://chay2u.medium.com/the-agent-sandbox-trap-8b9e75b8e66a

chay2u.medium.com

Farewell, GitHub

After 5.5 years at GitHub, 12 different managers, and only two teams: Data Infrastructure and Git Systems, I’ve decided it’s time for a new chapter. Looking back, these years have been a real…

migue.github.io

DuckDB-Quack announcement at AI Council

Slides: https://blobs.duckdb.org/slides/hannes-muehleisen-quack-ai-council-2026.pdf

youtube.com

Event-driven Programming is Usually a Poor Architecture | DBOS

Postgres creator and DBOS co-founder Dr. Mike Stonebraker explains why workflow architectures are preferred over event-driven and why AI frameworks are converging on a workflow-style model and…

dbos.dev

Supercharging Redpanda Streaming with profile-guided optimization

A behind-the-scenes look into how we optimized performance and cut latency in Redpanda 26.1 with profile-guided optimization.

redpanda.com

Decoupled DiLoCo: Resilient, Distributed AI Training at Scale

Google’s new distributed architecture keeps AI training runs on track across distant data centers, with exceptional efficiency – even when hardware fails.

deepmind.google

VillageSQL: Decisions, Principles, & Lessons in Teaching MySQL New Tricks (Steve Schirripa)

CMU Database Group — PostgreSQL vs. The World Seminar Series (Spring 2026) Speaker: Steve Schirripa (https://www.linkedin.com/in/steve-schirripa-99a1651/) April 20,…

youtu.be

The invisible engineering behind Lambda’s network

The subtle inventiveness that reduced cold start setup from seconds to 200μs.

allthingsdistributed.com

Design and Implementation of DuckDB Internals

DuckDB is an in-process SQL database management system focused on analytical query processing. It is designed to be easy to install and easy to use. DuckDB has no external dependencies. DuckDB has…

duckdb.org

Your Container Is Not a Sandbox

The microVM ecosystem was battle-tested long before agentic AI created the demand. A landscape survey: every VMM, the shared Rust crate ecosystem, a dozen AI sandbox platforms, and honest trade-offs.

emirb.github.io

WAL as a Data Distribution Layer

Introduction

richyen.com

S3 Files and the changing face of S3

Andy Warfield writes about the hard-won lessons dealing with data friction that lead to S3 Files

allthingsdistributed.com

Your RAM Has a 60 Year Old Design Flaw. I Bypassed It.

Modern DRAM is based on a brilliant design from IBM. But, we're still paying for a latency penalty that's existed since the 60s! In this video, I'm introducing my personal research project…

youtube.com

Why Are You Not Claudemaxxing?

From mass-dismissing AI coding tools to managing 5 Claude sessions on a walk. A prompt school origin story -- and your first useful AI win.

blog.pumpurlabs.com

Memory Isn't Learning

27 Mar 2026

christophermeiklejohn.com

The Cost of Concurrency Coordination with Jon Gjengset

In this talk, Jon Gjengset explores the true cost of concurrency coordination – from Amdahl's law down to CPU cache lines – and what we can do about it. Modern hardware is moving to more cores…

youtube.com

TurboQuant: Redefining AI efficiency with extreme compression

We strive to create an environment conducive to many different types of research across many different time scales and levels of risk.

research.google