DeepMind London: Ivan Leo on Agents, and the People Building With Them
Ivan Leo on Gemini interactions, video tools, and managed agents, alongside builders using agentic coding to solve everyday problems.
6 min2026-09-20Murali Suriar: Diagrams and Shared Understanding | Compile London
Murali Suriar on diagrams as shared understanding: choose the right view, compare mental models, and review architectural change.
4 min2026-09-20Pauline Brunet: AI Deployments at Scale | Compile London
Pauline Brunet session notes: use-case selection, forward deployed engineers, production ownership, evaluations, permissions, and outcomes.
5 min2026-06-29AI Tool Gateways: Sandboxing Agent Access in Kubernetes
AI agents have unlimited ambition and undefined access boundaries. You have sandboxed everything else. Here is how to sandbox your agents too.
9 min2026-06-22Observing LLM Inference: The Metrics That Actually Matter
TTFT is your SLO, not throughput. Here is what to measure for LLM inference, and what Google ADK, LangChain, and LangGraph give you out of the box.
8 min2026-06-15Network Control with Cilium and Kyverno: Policies That Actually Work
Network policies in most clusters are cargo cult: written, accepted, changing nothing. Cilium actually enforces them, and shows you the traffic.
8 min2026-06-08Multi-Tenant Observability: LGTM at Platform Scale
Your tenants want dashboards. Your security team wants isolation. Your SREs want a single pane of glass. Here is how to build all three with the LGTM stack.
8 min2026-06-01Gateway API in Practice: From Ingress Migration to Envoy Debugging
Kubernetes deprecated Ingress in spirit years ago. Gateway API replaces it with a model that actually separates platform from application concerns.
7 min2025-07-017 Books That Shaped How I Build Platforms Developers Actually Use
Seven books that shaped how I build platform products people trust, adopt quickly, and keep using.
10 min2021-11-10Writing a Go Client Library Worth Using
Most Go client libraries are a pain to use, test, or extend. How to write one that isn't: functional options, context propagation, rich errors, retries.
9 min