Blog
Technical notes on cloud and AI infrastructure, automation and operations. Trade-offs, failure modes and checks to make before committing engineering time or budget.
AWS promises 4x faster stack operations and hypervisor-level microVM isolation for dynamic code. An opinion on what each actually changes, who should care, and the fine print — because deployment speed and code isolation are different problems wearing one announcement.
Azure Linux 4.0-style minimal hosts — no GUI, SELinux enforcing by default, atomic rollbacks — are becoming the compliance-friendly default for container fleets. The operational and regulatory case, the migration friction nobody advertises, and an opinion on timing.
Why 'we have backups' is not an answer to a continuity question, and how engineering leaders should structure — and test — recovery before an incident or an audit forces the issue.
Decision criteria for CTOs weighing managed databases, Kubernetes, and serverless against self-operated infrastructure: the real cost comparison, the migration signals, and the lock-in question answered honestly.
Anthropic's gateway component for Bedrock and Google Cloud centralizes network policy, RBAC, and cost control for agent traffic — a corporate firewall for AI. Why the category needed to exist, what centralizing buys and risks, and an opinion on adoption timing.
CLI, dockerd, containerd, runc, and the kernel primitives underneath — namespaces and cgroups. Why the layered anatomy matters operationally, what each layer owns, and where incidents actually live. A back-to-fundamentals piece.
How engineering organizations facing ISO 27001, SOC 2, NIS2, or DORA can make audit evidence a byproduct of normal operations — and why infrastructure as code is the cheapest compliance tool you already own.
The typical findings of a short, structured cloud cost review — idle spend, mis-sized capacity, wrong service tiers — what each is worth, and why the durable savings come from process, not deletion.
A build-vs-borrow framework for engineering leaders: which infrastructure work justifies external senior capacity, what it should cost relative to hiring, and the three signs you are about to buy the wrong thing.
EKS, GKE, or AKS — or none of them. How to decide whether managed Kubernetes fits your organization, what it really costs to operate, and the simpler alternatives that are often the right answer.