Building a uz/ru/en site: routing, separating interface from content, what to do when a translation is missing, and SEO hreflang.
Building file uploads properly: direct upload via signed URLs, validation, the orphaned-file problem, and image processing.
Maintaining three npm packages over years: versioning, breaking changes, dealing with issues, and avoiding burnout.
Running a server on Proxmox: choosing between VMs and LXC, networking, the backup rule, and what you should not keep on your own hardware.
Speeding up a real site: measuring what users feel rather than a Lighthouse score, the three most effective optimisations, and the wasted work.
Building a webhook receiver properly: signature verification, idempotency, fast responses, moving processing to a queue, and out-of-order events.
E2E testing with Playwright: how many tests you need, how to choose selectors, the real cause of flaky tests, and how to run them in CI.
Writing a master-stack tiling script for KDE: why the existing plugin was not enough, how the script is built, and whether such a project is worth the time.
The expand/contract pattern: adding, renaming and dropping columns in a running system — without locks and without downtime.
A local image generation setup for shop product cards: how a 4 GB VRAM limit shaped every technical decision, and what did not work.
Real work with React Server Components: where to draw the client boundary, when Server Actions fit, and the trap that quietly makes a page dynamic.
Building a background job queue: idempotency, retry strategy, a dead letter queue, and choosing between Postgres and Redis.
A reliable single-server production stack with Docker Compose: healthchecks, log rotation, zero-downtime deploys, backups, and what not to do.
Running one project with an agent for months: what belongs in persistent memory, what MCP connections give you, and how to live with the context limit.
Choosing what to cover with tests: which code earns them, when mocks do damage, and why one hundred percent coverage is not the goal.
A real Tailwind 3 to 4 migration: @theme instead of tailwind.config.ts, the font trap, what happens to plugins, and what to drop along the way.
Moving from console.log to structured logs: correlating a request with a trace id, using levels properly, naming fields, and what you must never log.
What a skill is, how it differs from a plain prompt, when it is worth writing and when it is overkill — using the skill behind my own blog as the example.
A debugging story: why a game ran at half the FPS it should on Wayland, why the quick fix was a bad one, and where the real cause turned out to be.
Moving tenant isolation from a WHERE clause into PostgreSQL RLS: writing policies, passing context, testing them, and the traps that bite most often.
Migrating madrimov.uz to Next 16, React 19 and Bun: proxy instead of middleware, async params, pinning TypeScript 6, and three separate PRs.
The downsides of Redux Toolkit, MobX, and Context API, and why switching to the simple, minimal Zustand was the best decision.
A team lead's playbook: review priorities, small-PR culture, comment tone, reviewing juniors, and practical rules against the LGTM stamp.
Build a token bucket with a Map and an LRU + TTL cache on one Node.js instance; Cache-Control, ETag, stale-while-revalidate, and when you truly need Redis.
Discriminated unions, branded types, satisfies, unknown, zod and template literal types — six practical patterns for real type safety in TypeScript.
The real pipeline behind madrimov.uz: Notion as a single source of content, Next.js ISR, and Make.com scenarios for Telegram and LinkedIn distribution.
Lessons from M Smart School: domain modeling, the multi-tenant question, simple RBAC, designing for unstable internet, and pilot rollout takeaways.
Comparing sessions, JWT and refresh token rotation based on real project experience: when each approach is right and where to store tokens safely.
A monorepo is not a monolith: when it pays off, when it is overkill, and how Turborepo pipelines and remote cache work — practical experience and a decision tree.
From SSH and firewall to systemd, nginx, certbot, and backups — a complete practical scheme for production deploys on a VPS, plus an honest take on Docker.
B-tree intuition, 4 cases where an index silently fails, composite index column order, reading EXPLAIN ANALYZE, and partial and INCLUDE indexes — with real examples.
A canonical/hreflang mistake, dynamic html lang, a JSON-LD graph, the next/og prod trap, and soft-404s — practical SEO lessons proven on madrimov.uz.
A serverless blog on Notion + plain fetch + ISR: a 60-line client, block rendering, and a year of practical lessons.
Moving to multi-core with cluster silently breaks EventEmitter and @Cron. The mechanism behind the intermittent bug and the right fix with Redis/BullMQ.
Why I moved from NestJS + Prisma to Bun + Hono + Drizzle: abstraction, performance, event-driven pitfalls, and an honest comparison in numbers.
What a monorepo is, when you need it, and how to build one with Bun + Turborepo — practical experience from real projects.
Moving from Node.js to the Bun + Hono stack — why, how, and when to be careful.