Articles
In-depth research, comprehensive guides, and long-form thoughts on modern technology.
How to Test AI Agent Skills Without Hitting Real APIs
Evaluating an AI agent skill against a live API costs money, mutates real data, and produces results nobody can reproduce. Here's the fix.
Why Testing Agent Experience Changes Before Shipping Matters
A Microsoft team found that AI agents behave counterintuitively to documentation tweaks, and built a way to test changes before shipping them.
Ray Adds Native Support for Google Cloud TPUs
Ray 2.55 brings first-class Google Cloud TPU support, letting developers run distributed workloads on TPU slices with familiar Ray APIs.
Tunix Adds High-Throughput Agentic RL Training on TPUs
Google's JAX-native post-training library Tunix now supports high-throughput agentic reinforcement learning, targeting multi-turn tool-use training.
Amazon SQS Turns 20: Two Decades of Message Queuing
Amazon SQS launched in July 2006 as one of AWS's first three services. Twenty years later, its core decoupling pattern still holds.
How the 2026 World Cup Reshaped Global Internet Traffic
Cloudflare Radar data shows how kickoff times, halftime breaks, and even three-minute hydration pauses reshaped internet traffic worldwide.
How Dropbox Used DSPy to Improve Dash Chat's Judges
Dropbox used DSPy to calibrate its LLM judges against human labels, then used those judges to sharpen Dash chat's own system prompt.
Inside Riviera: Cloudflare's Content Platform for AI
Cloudflare's internal content processing platform, Riviera, was built for parsing web data. Here's how it's being reshaped for AI workloads.
How Spotify Scaled Data Insights with an AI Context Layer
Scaling internal data queries requires more than a raw language model. Spotify built a curated context layer to ensure its assistant remains highly accurate.
Engineering Netflix's Real-Time Service Topology at Scale
Creating a real-time dependency map at Netflix scale required abandoning simple batch processing and building a robust reactive pipeline to handle massive load.
How Netflix Scaled In-House LLM Serving with vLLM and Triton
Netflix runs its own generative inference stack using vLLM and Triton. Moving away from third-party APIs required solving complex state and deployment bugs.
Cursor's Latest Data Reveals How Developers Actually Use AI
Two years of Cursor usage data reveals that top developers generate massive amounts of code, relying heavily on context caching to reduce API overhead costs.