Beyond synthetic testing: Capturing and replaying real database workloads at Airbnb
Airbnb details its approach to capturing real production traffic across hundreds of MySQL-compatible database clusters running at millions of queries per second. The captured traffic is replayed offline to run realistic load tests, plan capacity, and safely de-risk major infrastructure upgrades.
Airbnb · 10 October 2026 · Read the original →
NTS: Authenticated Time at Meta
Meta implemented Network Time Security (NTS) on its public time infrastructure to enable authenticated packet verification without in-flight tampering. To scale the service without storage bottlenecks, the servers operate statelessly by deriving cookie keys rather than storing or replicating per-client state.
Meta · 10 October 2026 · Read the original →
Building Git infrastructure for agent-scale development
GitHub outlines the architectural changes required to adapt its core Git infrastructure for agentic software development workloads. As automated agents join human developers in generating code and repository activity, the platform is evolving its architecture to handle the resulting surge in access patterns.
GitHub · 10 October 2026 · Read the original →
Metrics Board: Building an Agent-ready Metrics Layer
Pinterest describes the design of its Metrics Board, a unified metrics layer created to provide reliable data across business performance and feature experimentation. The system is engineered to expose standardized, trustworthy metrics that can be consumed directly by automated AI agents and human decision-makers.
Pinterest · 10 October 2026 · Read the original →
The keys to the Internet change on October 11. Are you ready?
Cloudflare explains the technical mechanics and operational risk of the DNS root Key-Signing Key (KSK) rollover. Because the key anchors the cryptographic chain of trust for DNSSEC, validating resolvers must trust the new key prior to the switch to prevent resolution failures for valid sites.
Cloudflare · 10 October 2026 · Read the original →
From Activity to Intent: Generating User Journeys with LLMs
Pinterest details how machine learning engineers use large language models to transform low-level user activity data into high-level user intent journeys. The approach aggregates granular interaction logs into structured behavioral sequences to better model user intent across the platform.
Pinterest · 10 October 2026 · Read the original →
Building an evidence-grounded agentic security operations harness on Cloudflare
Cloudflare details the design of its multi-agent security operations harness built to mitigate alert fatigue during security incidents. The architecture uses specialized AI agents to correlate simultaneous environmental signals, group related security events, and assist human analysts during high-volume alert spikes.
Cloudflare · 10 October 2026 · Read the original →
We are building infrastructure for agents faster than we can agree on how time works.
7 stories, every Saturday
Published here every week. Follow by RSS to get it as it lands.