Notes on software, architecture, and the details that make systems work. Written from experience, shared for the next build.
Five principles that separate code you can still touch in six months from code that becomes a minefield. Each principle explained in plain English with before/after code in TypeScript, Java, Python, and C#.
Unit, integration, contract, and e2e — when each actually pays off, the modern test pyramid, and a practical policy for flaky tests. Concrete advice for a service that ships.
Instrumenting a production Node.js service for traces, metrics, and logs with OpenTelemetry — plus correlation by trace ID, exporter choices, and what breaks in production.
How to change a production schema without taking traffic down — expand/contract, backfills, dual-writes, online DDL, and the migrations that always bite you.
From kill switches to percentage rollouts — how to decouple deploy from release, evaluate self-hosted vs SaaS flag platforms, and keep flag debt from eating your codebase.
The 2021 OWASP Top 10, but with actual vulnerable Node.js and Next.js snippets and the fixes I ship in production. Concrete examples beat awareness posters.
Cache-aside, write-through, write-behind, read-replicas, edge caches, HTTP ETags, and stale-while-revalidate — how to pick the right caching pattern and how to invalidate without tears.
Chunking, embeddings, hybrid retrieval, reranking, citation, and evaluation — a practical guide to building a RAG pipeline that survives real traffic and messy documents.
How to test prompts, catch regressions, score outputs, and defend against prompt injection — with a small eval harness you can ship this week.
When a monorepo pays off, which toolchain to pick, and what a sensible layout looks like for a small-to-mid team — pnpm workspaces, Turborepo, Nx, and the newer options.
