postgres
Postgres 19 adds SQL/PGQ graph queries over the tables you already have, an atomic ON CONFLICT DO SELECT upsert, and REPACK CONCURRENTLY that returns disk space without locking the table. Here is what each one replaces, with working SQL and the limits the release notes do not shout about.
docker
Docker Swarm mode is not dead, but it is not growing either. Build a three node cluster from zero, deploy a stack, handle secrets and volumes, dodge the Compose file trap, and see plainly when Kubernetes or Kamal is the better call.
ruby-on-rails
A modular monolith in Rails buys you one thing: a checked constant dependency graph. Packwerk 3.3 and Rails engines run against a real app show where the boundary holds, where a single method argument walks straight through it, and which of your problems stay exactly as they were.
YouTube runs the biggest recommendation feed on the planet, and it does not win with rules. See how its homepage ranks on satisfaction instead of clicks, learns per user, and trusts reactions, then how to build the same ideas into your own app with Ruby and SQL.
Every engineer learns DRY in their first year: don't repeat yourself, extract the common code, keep one source of truth.
Shrink a Rails Docker image from 1 GB to under 150 MB in seven ordered steps: smaller base image, .dockerignore, multi-stage builds, layer ordering, distroless with its honest caveat, and digest pinning. Every step is measurable the same afternoon.
Refactoring a live monolith is a risk problem, not a style problem. A strategy built on real pain points: modular monolith boundaries enforced by tools, boring rollbacks, avoiding the three-week refactoring branch, and saving the expensive decision for last.
AI writes confident Rails code that quietly breaks production standards. Build a quality workflow that catches it: Lefthook gates, Strong Migrations as a hard line, N+1 detection with Prosopite, database checks in layers, and CI that mirrors local hooks.
Removing a Rails database column can crash old processes mid-deploy. This walkthrough shows the safe order: ignored_columns first, migration later, plus how to handle columns that still hold data and what the schema cache does behind your back.
Messaging looks clean on a diagram and gets messy in production. Duplicate messages, slow consumers, fat events, DLQs used as trash cans: what each problem really costs and the patterns (idempotency, event design, async UX) that keep a broker-based system honest.
Fail2ban turns brute-force noise on SSH and your reverse proxy into a boring automated reaction. A practical setup for a VPS or homelab: sane jail config, testing before trusting, notifications without a siren, and where it fits in the wider security story.
DORA metrics explained for engineering leaders: deployment frequency, lead time, change failure rate, time to restore and reliability. What each one really tells you, how to use them to improve delivery, and the anti-patterns that turn them into dashboard theater.
21 engineering lessons that survive contact with production: why simple code beats clever code, technical debt as a communication problem, what actually builds trust in teams, and how senior engineers pick problems. Field notes for real teams, not conference slides.
HTTP caching refresher for engineers: how freshness, revalidation, Vary and Cache-Control really interact, why stale responses can still be useful, what reload actually does, and the boring headers that keep private content out of shared caches.
Rails partials feel free until a page renders hundreds of them. Benchmarks of five rendering strategies (inline ERB, partial loops, collection rendering and more) show which is fastest, why the helper trick backfires, and a practical rule for when to optimize views.
UUIDv7 arrives in PostgreSQL 18 as a built-in uuidv7() function. See what it changes in practice: friendlier indexes than UUIDv4, simpler Rails migrations, a schema example, and the security caveat that time-ordered does not mean secret.
Auditing a legacy Rails codebase starts with one question: what is the team afraid to change? A practical method for finding real hotspots, building a small audit ledger, using tools and AI without drowning in output, and delivering one page that people act on.
Delete first, automate last: Elon Musk's five-step algorithm applied to software. Why questioning requirements and deleting code must come before optimization, and how automating a broken process just gives you a faster version of the wrong machine.
Structured output from Ollama takes more than JSON mode. Learn the reliable pattern for a Rails app: shape the input, constrain output with JSON Schema, parse defensively, and retry with feedback so a local LLM survives production instead of just a demo.
SQLite in production works better than its toy reputation, but only if you know the traps: why WAL mode is non-negotiable, how blue-green deployments break your setup, missing ILIKE, quirky JSON types, and what you honestly gain in return.
Architecture is the set of decisions that feel permanent after six months. An honest tour of monolith, layered, modular monolith, SOA, microservices and event-driven options: what each really costs and why the modular monolith deserves more attention than it gets.
Message ordering in distributed systems does not come from a FIFO queue. What actually keeps events sane: idempotency as the real safety net, partition keys when order truly matters, optimistic locking as a version guard, and the traps of multiple consumers.
Cynefin framework in practice for engineers: classify a problem as clear, complicated, complex or chaotic before picking a process, so you know when Agile experiments beat Waterfall plans and what to do when production is down.