I've been working intensively with Claude API and Claude Code for over a year — not as a curiosity, but as my primary work tool. I've gone through all the phases: initial enthusiasm, model limitations, architectures that don't scale, and ones that do.
Before "vibe coding" was a trend, I was already building complete projects with Claude Code as the agent. I've seen what actually works and what's marketing. This blog exists to write about the former.
I have no patience for noise. If an article doesn't contain real code, concrete architecture decisions, or clear opinions on when not to use AI, I don't publish it.
If you have a project where Claude API, Claude Code, or AI agents could make a difference, let's talk. No commitment.