A reading room for system design

How the hardest systems are actually built.

A curated archive of engineering writing from Netflix, Meta, AWS, Cloudflare, Stripe, and others — distilled into concepts, patterns, and systems you can actually cite.

1,571Articles read

542Distilled

2,903Concepts

1,797Patterns

1,693Systems

40Companies

Start here

• Concept

  Control Plane / Data Plane Separation

  The single most-cited pattern in the corpus: keep the fast path independent from the thing that configures it.

• Concept

  Observability

  Logs, metrics, traces — and the unresolved debate about what belongs in each.

• Concept

  Write-Ahead Logging

  How databases, message queues, and filesystems all survive crashes the same way.

• Concept

  Eventual Consistency

  What you actually get when you scale a read-heavy system past one region.

• Concept

  Backpressure

  Unbounded queues are the most common production bug. This is how systems push back.

• Concept

  Tail Latency at Scale

  Why the p99.9 matters more than the mean, and the hedged-request playbook for taming it.

• Concept

  Blast Radius

  The discipline of designing so a single failure doesn't take everything with it.

Latest additions

1. 2026-06-19Adaptive write request scheduling in Redpanda's Cloud Topics
2. 2026-06-19Build your own vulnerability harnessCLOUDFLARE
3. 2026-06-18Building Agents that Don't Break ThemselvesFLYIO
4. 2026-06-18Cloudflare — Bringing more agent harnesses and frameworks to CloudflareCLOUDFLARE
5. 2026-06-18Long Horizon: How Atlassian Built a Reasoning Engine for Complex AI TasksATLASSIAN
6. 2026-06-12Enabling Evolutionary Database Development: Database branching with Lakebase, the conclusion (Part 3)DATABRICKS
7. 2026-06-12Ingesting the Milky Way: Petabyte-Scale with Zerobus IngestDATABRICKS
8. 2026-06-12How Dropbox uses MCP and Dash to close the design-to-code security gapDROPBOX
9. 2026-06-12Scaling Security Insights: how we achieved a 10x increase in global scanning capacityCLOUDFLARE
10. 2026-06-09Scaling beyond one: How Airbnb evolved its data architecture for a multi-product worldAIRBNB
11. 2026-06-11Metric Semantic Layer: How Lyft Governs and Scales Key Data DefinitionsLYFT
12. 2026-06-10Architecting Scalable ML Platforms: The Integrated Infrastructure and Acceleration Behind Rovo

Browse

• Reference

  Concepts — 2903

  The primitives: CAP, WAL, consistent hashing, backpressure, leader election.

• Reference

  Patterns — 1797

  Repeatable solutions: circuit breaker, sidecar, saga, event sourcing, outbox.

• Reference

  Systems — 1693

  Named systems in production: Dynamo, Kafka, Spanner, Borg, Colossus.

• Primary source

  Sources — 542

  Every ingested article with full metadata, tags, and backlinks.

How this works

An ingestion pipeline pulls engineering blog posts from ~30 company feeds, de-duplicates them, and runs each through a curator agent that distills the piece into structured notes and cross-links it against the existing concept graph. See the overview for the full methodology and corpus stats.

The LLM-readable catalog mirrors every page for agent consumption.