AI-native development is not only about asking a model to write code. The useful version is a workflow where AI agents, local tools, and production infrastructure work together in a tight loop.
My current stack centers on OpenAI Codex for repository work, Claude Code for implementation sessions, Gemini CLI for fast command-line assistance, MCP workflows for connected tools, Bun for the JavaScript runtime, React for interactive interfaces, and Cloudflare Workers for deployment.
What Matters
The most important part is not the number of tools. It is the feedback loop:
- define the product behavior clearly
- build the smallest useful version
- test the rendered UI and command output
- deploy to a real URL
- iterate based on what actually works
This keeps the work close to production from day one.
Why Bun and Workers
Bun keeps local development fast and simple. Cloudflare Workers gives the site a global static delivery layer with a deployment flow that fits the way I ship: build locally, deploy through Wrangler, verify the live URL, and continue iterating.
For static portfolio and content sites, this is a strong baseline. It avoids unnecessary backend complexity while leaving room for future Worker logic, AI endpoints, forms, analytics, or edge features.
The Working Style
AI tools are best when they are treated as collaborators inside an engineering process. I use them for planning, code generation, refactoring, testing, review, and deployment checks, but the product direction stays grounded in user experience and real outcomes.
The goal is simple: turn ideas into functional web experiences faster, without losing the discipline needed for production.
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.