Articles
Worth reading
A short take on pieces worth your time — architecture, AI, system design, and engineering careers — each one linked back to its original source.
42 articles
Cloudflare outage on November 18, 2025
Matthew Prince's own postmortem of the outage that took down a meaningful chunk of the internet for several hours, X, ChatGPT, and Cloudflare's own status page among the casualties, and the root cause is almost mundane: a database permissions change caused a bot-management "feature file" to double in size and blow past a hardcoded limit. It's a good reminder that the scariest outages rarely come from novel attacks; they come from ordinary config changes interacting badly with an assumption nobody revisited. Prince's willingness to publish the gory detail, including where their own tooling made diagnosis slower, is what makes this worth reading over a sanitized summary.
Integrating Netflix's Foundation Model into Personalization Applications
Divya Gadde, Ko-Jen Hsiao, Dhaval Patel, and Moumita Bhattacharya walk through the three different ways Netflix integrated a single foundation model into multiple personalization surfaces, embeddings, subgraph, and fine-tuning, rather than training a bespoke model per use case. It's a practical answer to a question a lot of teams are facing right now: once you have one strong foundation model, how do you actually plug it into existing product surfaces without rebuilding everything downstream? The tradeoffs they lay out between the three integration patterns are transferable well beyond recommendation systems.
Effective Context Engineering for AI Agents
Anthropic's Applied AI team makes the case that context, not the prompt, is the scarcest resource in a production agent, and lays out concrete tactics for curating it: tighter tool definitions, deliberate compaction between turns, and structured note-taking instead of just stuffing more into the window. I've started treating "what's actually in the context window right now" as its own design question after reading this, rather than assuming bigger windows solve everything. A natural, more tactical follow-up to Anthropic's earlier "Building Effective AI Agents" piece.
Building Effective AI Agents
Anthropic's own engineers stepping back from the hype to draw a precise line between a "workflow" (LLMs and tools wired through code you control) and an "agent" (the model deciding its own next steps) — a distinction that matters the moment you're deciding how much autonomy to actually hand an LLM in production. The five patterns it lays out — prompt chaining, routing, parallelization, orchestrator-worker, evaluator-optimizer — are the ones I keep coming back to when scoping an AI feature, because they're composable primitives rather than a framework you have to buy into. Worth reading before you reach for a heavyweight agent framework: most production use cases turn out to need a workflow, not an agent.
Netflix's Distributed Counter Abstraction
Netflix's engineers describe the counting service they built on top of their TimeSeries abstraction to handle 75,000 counter requests per second at single-digit-millisecond latency, the kind of "boring" infrastructure problem that turns out to be genuinely hard once you need it fast and consistent-enough at global scale. I like this as a case study in building a narrow, well-defined abstraction on top of an existing platform primitive instead of reinventing storage from scratch for every new use case. Relevant to anyone building shared counting or aggregation infrastructure.