I don’t code. I prompt.
That’s not a flex — it’s a description. I spent 20+ years in tech, mostly in cybersecurity, and I never became a software developer. I understood systems, infrastructure, and business logic. I could read code well enough to know if something was broken. But writing it from scratch? Not my strength, and honestly, not my interest.
Then AI tools got good enough that it stopped mattering.
What “Building with AI” Actually Looks Like
Let me be specific, because the internet is full of vague AI hype. Here’s what my typical build session looks like:
- I open Claude Code — a CLI tool that gives an AI agent access to my filesystem, terminal, and development tools.
- I describe what I want in plain English. Not pseudo-code. Not technical specifications (usually). Just “I need a Python script that connects to the WordPress REST API, pulls credentials from 1Password via SSH, and provides get/post/delete methods.”
- Claude writes the code. I review it. I ask for changes if something looks off. We iterate.
- I test it. If it works, we move on. If it breaks, I paste the error and we fix it together.
That’s it. That’s the workflow.
What I’ve Built This Way
In the last few months, using this approach, I’ve built:
- A complete e-commerce store — WooCommerce, custom theme, automated product publishing pipeline, SEO optimization. It sells real products to real customers.
- A multi-site blog engine — PostgreSQL database, content templates for different post types, automated publishing to WordPress via REST API, scheduled content calendar.
- A self-hosted infrastructure stack — Proxmox hypervisor, Docker containers running PostgreSQL, n8n, vector databases, local LLMs. All configured through conversational AI.
- This website — the theme, the deployment scripts, the content pipeline. All of it.
None of this is trivial. These are production systems handling real traffic and real money.
The Honest Parts
Here’s what the AI hype machine won’t tell you:
It’s not magic. You still need to understand what you’re building. If I didn’t know what a REST API was, or how databases work, or what a Docker container does, Claude couldn’t help me. The AI is incredibly capable at writing code, but it needs someone who understands the architecture to point it in the right direction.
You will debug. A lot. AI-generated code works maybe 80% of the time on the first try. The other 20% requires iteration — sometimes quick fixes, sometimes rethinking the entire approach. The debugging process is faster because the AI can help, but it’s not zero.
Context management is a skill. The biggest challenge isn’t getting AI to write code — it’s giving it enough context to write the right code. I maintain detailed project files (CLAUDE.md), infrastructure documentation, and decision logs specifically so the AI has the context it needs to make good choices.
You need to verify everything. AI can write plausible-looking code that’s subtly wrong. You need enough understanding to catch those errors. I test every feature, check every deployment, and never assume something works just because it was generated without errors.
Why I’m Sharing This
Because the gap between “AI can sort of do stuff” and “AI can build real products” is closing fast, and most people don’t realize it yet.
I’m not a developer. I’m a builder who happens to use AI as his primary tool. And if I can build production systems, e-commerce stores, and automated infrastructure this way — what could you build?
That’s what this blog and YouTube channel are about. Real builds, real costs, real lessons. No hype, no toy demos.
Welcome to the workshop.