How I Built MagicCode in 2 Weeks — With a Little AI-Cadabra

The behind-the-scenes story of launching a tech consulting business using AI tools, 30+ years of experience, and zero developer team.

People keep asking me: “Wait — you built all of this yourself? In two weeks?”

The short answer is yes. The longer answer is more interesting — because I didn’t do it alone. I had help from AI agents that worked alongside me around the clock, and the experience completely changed how I think about what one person can accomplish with the right tools.

Here’s the full story.

A Quick Backstory

I’ve spent 30+ years in technology. I studied computer science at UC San Diego, worked my way up to director-level roles leading global engineering teams, founded and sold a company called WebDispatch.com, and spent years at companies like Ayla Networks (IoT) and Contentstack (content management). I’ve seen a lot of tech cycles come and go.

But when I retired and moved to Danville, I noticed something that all those years in enterprise tech didn’t prepare me for: my neighbors, friends, and local small business owners were genuinely struggling with everyday technology. Not because they weren’t smart — but because nobody was taking the time to help them one-on-one, without the jargon, without the condescension.

I wanted to fix that. And I wanted to build the business fast, because the need was already here.

The Two-Week Sprint

Most people would hire a web developer, a designer, maybe a marketing consultant. Wait a few months. Spend a few thousand dollars. I took a different path — I used AI tools to do almost everything, directing the work like a tech lead running a small team.

Here’s roughly how those two weeks played out:

• Days 1–2: Strategy and Brand I used ChatGPT to brainstorm the business model, define my target market, nail down pricing tiers, and draft all the core messaging. Together we landed on the wizard mascot and the “AI-Cadabra” theme — because making technology feel like magic is exactly what I do. I also used ChatGPT’s image generator to create every branded visual, from the wizard hero image to social graphics.
• Days 3–5: Website Build This is where Claude came in. I set up a WordPress site on Hostinger using the Twenty Twenty-Five theme, then handed the heavy lifting to Claude Code — Anthropic’s AI coding agent. Claude Code wrote all the page HTML and CSS, set up the REST API and WP-CLI workflows, built custom page templates, and handled all the technical plumbing. Every page — Home, Services, About, Contact, Book, Pricing, Blog — was built through code, not drag-and-drop.
• Days 6–8: Infrastructure Claude Code set up Google Analytics 4 with custom conversion events (booking clicks, contact form submissions, city page views, Calendly clicks). It configured Tawk.to live chat, integrated Calendly booking, built Pushover push notifications so I get alerts on my phone when someone visits or submits an order, and created a session-tracking system. All of this through terminal commands — no clicking around in admin panels.
• Days 9–11: SEO and Content We built 8 local SEO city pages — Danville, San Ramon, Pleasanton, Blackhawk, Alamo, Walnut Creek, Dublin, Livermore — each with unique content and internal cross-links. I drafted blog posts with ChatGPT’s help and Claude Code published them. We set up Google Search Console and linked it to GA4.
• Days 12–14: Polish and QA This is where my second AI agent came in. I call her Coco (short for Claude Cowork). Coco works in a browser — she handled visual QA, checked every page on desktop and mobile, configured wp-admin settings, set up Calendly event types, and caught layout issues that a terminal-only agent couldn’t see. Meanwhile, Claude Code built a regression test suite that checks all 44 critical site elements automatically.

The AI Team

Let me introduce the crew — because this is the part that fascinates me most.

The agents even communicate with each other through a GitHub Issues message board I set up. Claude Code leaves notes for Coco about what changed, Coco reports visual issues back, and I review everything. It’s a real workflow — just with AI team members instead of human ones.

What Actually Got Built

I want to be specific here, because “I built a website” undersells what happened. In two weeks, the following went live:

No developer team. No design agency. No marketing consultant. Just me and three AI tools, working together the way I used to lead engineering teams — except faster, cheaper, and available 24/7.

What This Taught Me

Building MagicCode this way wasn’t just efficient — it was genuinely fun. And it crystallized something I’d been thinking about since I first started using AI tools seriously:

AI doesn’t replace expertise. It amplifies it.

I couldn’t have built this in two weeks if I didn’t have 30 years of knowing how websites work, how to structure a business, how to lead a team, and how to spot problems before they become expensive. The AI tools were fast and capable, but they needed direction. They needed someone who could say “that CSS is wrong” or “that business model won’t work in this market” or “we need regression tests before we touch the homepage again.”

That’s the part most people miss about AI. It’s not magic on its own — it’s magic when someone who knows what they’re doing is at the controls.

Why I’m Sharing This

Partly because I’m proud of it. Partly because I think it’s a glimpse of where things are heading. But mostly because I want people to know that this kind of thing is possible — and it’s not just for software engineers.

You don’t need to write code to benefit from AI. You need curiosity, patience, and someone who can show you the ropes. That’s what I do every day at MagicCode, whether it’s helping someone set up ChatGPT on their iPhone, fixing a small business website, or walking through Apple Intelligence features one by one.

The tools are here. They’re ready. You just need a good guide.

Written by Ron Breger, founder of MagicCode — technology consulting for real people in the East Bay and beyond.