How I built a 47-class AI coding academy with zero lines of hand-written code
I'm Robert McCullock, solo founder of Design Delight Studio in Boston. I run a sustainable streetwear brand. About a year ago I started rebuilding my entire business infrastructure with AI agents. The DDS Vibe Academy is the publishing arm of that work — 47 free classes on AI coding, updated Monday through Friday.
Why I built it
I didn't set out to build an academy. I was learning to use Claude Code, then Cursor, then Antigravity, then Ollama for local inference. Every time I figured something out, I wrote it down so I wouldn't have to figure it out again. Those notes became the first classes.
The academy exists because being the only solo founder in my industry with this material is lonelier than running it as a one-person publishing house. The classes are free because the trade is fair: if a class helps someone ship one thing, they owe nothing. If it helps them build a business, maybe they send a customer to my store someday.
How it was built
The academy hub at ddsboston.com/pages/dds-vibe-academy is itself a working artifact of the methodology. Here's the exact build chain:
- Claude Opus 4.7 (Anthropic) authored every Liquid section. 12 sections, 1 JSON template, approximately 6,400 lines, approximately 500 KB of code.
- Google Antigravity (Google, running Claude Opus 4.6 as its agent model) deployed every file to the Shopify storefront through the Shopify MCP protocol.
- Cowork (Anthropic) ran an autonomous browser audit that found 30 broken URLs and a z-index collision on the sticky navigation.
- Shopify MCP served as the write channel for all theme asset operations.
I did not write a line of code. I did not upload a file. I did not click "save" on the Shopify admin. I designed the constraints — the page structure, the CSS scoping rules, the color palette, the section architecture — and the agents handled the implementation.
There's a Build Manifest on the live page that documents this chain. It's a deployment receipt proving the page's AI-native construction.
What's in the curriculum
47 classes across four stages:
Foundation (17 classes): What Is Vibe Coding, Your First AI Environment, Prompting as Specification, Reading & Directing AI Code, The Vibe Build Loop, Antigravity Onramp, The Claude Onramp, The Cursor Onramp, Ollama for Windows, ChatGPT Foundations, The Sidekick Onramp, Shopify Liquid Foundations. Get a working setup. The shortest distance from zero to "I built something."
Development (12 classes): Claude Code Masterclass (Part 1 & 2), Gemini 3 Deep Think Patterns, Antigravity First Light, Antigravity In Production, Cursor 3 Masterclass, Sidekick Power Patterns, Ollama Cloud Routing, Antigravity 2.0 Masterclass, ChatGPT Images (Part 1 & 2), Hydrogen Storefront Build. The patterns that survive contact with production.
Application (11 classes): Sovereign RAG Masterclass, Claude Landing Pages, VibeTube AI, NicheForge V3, Homepage Upgrade Masterclass, Atelier OS, AGI Nexus, Synthetic Director, Sovereign Orchestrator. Where individual tools become end-to-end workflows.
Mastery (10 classes): MCP Server Forge, SEO Magnet V2, Trilogy Case Studies (Sovereign Orchestrator, Synthetic Director, NicheForge), DVA Desktop Apps Portfolio, Gemini 3.1 Pro: The Pivot, Multi-Model Orchestration. Forensic records of shipped infrastructure.
Plus monthly digest classes (The Vibe Coder's Haven) and daily updates from the sister blog.
The rules
- Free forever. No paywall, no membership, no certificate, no upsell.
- No signup. Not even an email gate. Permanent URLs you can bookmark and share.
- Published as web pages, not video. Written for people who learn by reading.
- A sister blog at /blogs/vibe-code-academy publishes Monday through Friday.
- Time-series learning: old classes stay alongside new ones, dated, in context. You can read them in order and watch the field move.
The methodology
I call it DDS Vibe Coding. Five pillars: constraint-first design, progressive disclosure, deterministic rails, context hygiene, sovereign fallback. The Manifesto tab on the academy hub lays this out in detail.
The core idea: you don't learn AI. You become the architect. You define what the system should do. The model figures out how.
