Next.js
App Router, RSC, Vercel best practices, avoid waterfalls, optimize bundle size, and use the React/Next.js skills in the project.
How I use AI to improve code quality and productivity every day.
Great code starts with strong foundations.
When people ask if I use AI in my day-to-day work, this is my answer: a repeatable workflow that keeps quality high and output predictable across stacks.
Productivity & quality2x
The workflow, step by step
Strong foundations
Before generating a single line, define the base: clear requirements, scope, and success criteria.
Choose the right model
Pick a capable model for the task: Gemini, Claude, Kimi, or ChatGPT — depending on context length, code quality, and availability.
Choose the right tool
Cursor, Windsurf, Claude Code, Codex, or similar — the IDE or agent that best fits your stack and workflow.
Clear user stories & acceptance criteria
Well-written HUs with explicit acceptance criteria so the AI (and the team) know exactly what "done" looks like.
Skills & rules
Use project-specific skills (e.g. React best practices, SOLID) and Cursor rules so the AI follows your standards.
MCPs that extend the AI
My go-to MCPs: Playwright, Context7, Supabase, Chrome DevTools, GitHub, Notion — for docs, DB, browser automation, and repos.
Give the tool context
Attach relevant files, paste error messages, and describe the goal so the AI has enough signal to generate correct code.
Clear architecture & patterns
A defined project structure and design patterns for both backend and frontend — so the AI generates code that fits the system.
Data contract
Define schemas and contracts (e.g. API types, DB models) so the AI doesn't invent or remove fields unexpectedly.
Code with best practices
Clean Code, SOLID, and framework-specific guidelines (e.g. Vercel React/Next.js best practices) applied consistently.
Testing
Unit tests, E2E tests, and automated Playwright flows to catch regressions and validate behavior.
CI/CD pipelines
GitHub Actions (or similar) to run tests, lint, and deploy — so every change is validated before merge.
Code review
Human or tool-assisted review (e.g. GitHub PRs, SonarQube, peer review) to keep quality and consistency.
Applied to every stack
The same workflow adapts to the stack: foundations and quality first, then stack-specific practices.
Java Spring Boot
Layered architecture, REST conventions, Spring Security, clean code and testing (JUnit, MockMvc, Testcontainers).
Python
FastAPI or Django, type hints, pytest, and clear project structure (e.g. src layout, dependency injection).
React Native
Expo when possible, performance awareness, native modules when needed, and the same quality bar as web React.