Engineering
RSSDeep dives and write-ups from real work — cloud and application architecture, and the reasoning behind the decisions.
-
How to let teams own their repositories and pipelines without the platform team becoming a ticket queue — declarative registries, validation that runs first, reconciliation with drift detection, and a lifecycle that includes retirement.
-
What it takes to host a self-hosted CI platform properly — two agent tiers, queue-driven scaling, breaking the bootstrap dependency, stopping base-image drift, and making the server disposable.
-
A production-safe rollout checklist for infrastructure that cannot be recreated quickly, including retention defaults, staged validation, and manual approval points.
-
How to choose between transit routing and direct peering for managed database access, with practical decision criteria and routing guardrails.
-
When multi-cloud is worth it (and when it isn't), and how to pick the tool that automates it — a practitioner's take on Terraform, OpenTofu, and Pulumi in 2026.
-
A practical CI/CD design for monorepos with multiple infrastructure runtimes, focused on selective pipelines, safer promotion, and clean ownership boundaries.
-
How to keep infrastructure stacks generic while all environment-specific values live in validated config files, reducing drift and risky one-off changes.
-
The working discipline I've settled on for getting dependable, secure, low-regret results from AI coding assistants — prompting frameworks, model selection, context hygiene, and safety rules, learned the hard way.
-
How uv replaces pyenv, venv, pip, pip-tools, pipx, and most of Poetry with a single fast tool — with a quick-start, a migration path, and a replacement cheat sheet.
-
A plain-language tour of the two most consequential Python releases in years — free-threading, JIT compilation, subinterpreters, t-strings, deferred annotations, and what to do about them.