engblogs

Engineer a real-time distributed graph query service over billions of nodes and edges, using gRPC, breadth-first traversal, asynchronous I/O, and selective caching to deliver sub-100ms, multi-hop queries.

Netflix's Real-Time Distributed Graph powers fast, multi-hop queries through a BFS-based, async-first serving layer with selective caching and gRPC, delivering sub-100ms responses at scale.

Explores Netflix's Real-Time Distributed Graph (RDG) query layer, detailing a breadth-first, async-first, gRPC-driven design that streams adjacency lists, applies per-edge-type filters, and uses selective caching (EVCache) to deliver sub-100ms multi-hop queries at scale.

A technical overview of merging AI Gateway and Workers AI into a single AI control plane, delivering automatic observability, unified billing, and model-first routing across providers.

Cloudflare unveils a revamped community program featuring Cloudflare Ambassadors and Cloudflare Community Engineers, plus an additional $1M in open-source funding to empower developers, sustain open-source projects, and strengthen engagement on Discord and local communities.

Radar Researcher combines AI and API-backed access to Cloudflare Radar data, allowing plain-language queries to generate interactive charts and explain results in context.

Explores how to manage 'agentic' traffic by shifting from static checks to continuous, trust-based risk assessment using tools like Precursor and BotBase, and outlines adaptive defenses for a safer Internet.

Unifies AI Gateway and Workers AI into a single AI control plane that centralizes access, observability, billing, and model-first routing across providers.

Cloudflare debuts Ambassadors and Community Engineers programs to empower global developer communities, backed by a fresh $1M in open-source funding to sustain and grow open-source projects.

Radar Researcher is a plain-language AI tool that lets users explore Cloudflare Radar's open data through the API, delivering interactive charts and step-by-step explanations for fast, verifiable insights.

Explores continuous, trust-based detection of agentic traffic on the Internet, detailing BotBase and Precursor and the new Adaptive Intelligence to distinguish good and bad behaviors.

Performance-focused comparison of diffusion and autoregressive language models showing how DLMs achieve higher arithmetic intensity through parallel decoding while facing long-context scaling, with block-wise decoding and batched inference shaping throughput and latency.