Cloudflare Blog1 min readMar 24, 2026

Sandboxing AI agents, 100x faster

We’re introducing Dynamic Workers, which allow you to execute AI-generated code in secure, lightweight isolates. This approach is 100 times faster than traditional containers, enabling millisecond sta

Cloud AI

Dynamic WorkersAI-generated codeAPI DesignSecurity

AI Summary

Cloudflare has developed a new sandboxing technology, Dynamic Worker Loader, which is 100x faster than traditional containers. It uses isolates, instances of the V8 JavaScript execution engine, which take only a few milliseconds to start and use a few megabytes of memory, allowing for unlimited scalability and zero latency. This technology enables agents to write code on-demand, run it in a secure sandbox, and access specific APIs, all in a scalable and efficient manner.

Read full article →

A one-line Kubernetes fix that saved 600 hours a year

A large Terraform project management tool called Atlantis in Kubernetes was experiencing slow restarts due to a volume ownership update bottleneck. The Kubernetes default setting for updating group ownership on persistent volumes was set to "Always," causing delays as the volume grew to millions of files.Changing the `fsGroupChangePolicy` field in Atlantis's pod specification to `"OnRootMismatch"` greatly improved restart times, allowing for a 30-second restart instead of the previous 30-minute delay.

CloudPerformanceBackend

KubernetesAtlantisKubernetes OptimizationCI/CDPerformance OptimizationSecurityCloud Native

Cloudflare Blog1 min3d ago

Launching Cloudflare’s Gen 13 servers: trading cache for cores for 2x edge compute performance

Cloudflare upgraded its server fleet to Gen 13, based on AMD EPYC 5th Gen Turin processors, which offer 2x core count improvements over Gen 12 but reduce per-core cache. FL2, a Rust-based rewrite of Cloudflare's request handling layer, was crucial in unlocking the performance potential of Gen 13 without latency penalties. By transitioning to a cleaner architecture with improved memory access, FL2 enabled 2x edge compute performance while maintaining service-level agreements (SLAs).

CloudPerformanceBackend

AMD EPYCTurin CPUsEdge ComputeCompute PerformanceServer ArchitectureCloud Infrastructure