> The Audit Step Matters More Than the Build: One Founder's All-AI Shopify Experiment

A solo founder just shipped a fully-functional 12-section Shopify hub page—the DDS Vibe Academy at ddsboston.com—without writing a single line of code, uploading a file via FTP, or touching the theme editor. The entire pipeline ran through chained AI agents: one wrote every Liquid section, another deployed them to Shopify, and a third audited everything after deployment. The result? A 13-section page with tabbed navigation across eight content areas, client-side filtering for 49 curriculum nodes, FAQ accordion with proper schema markup, and 93 cataloged external tools. All generated by machines under human architectural direction.

The Pipeline Architecture

The academy lives at /pages/dds-vibe-academy using Shopify's OS 2.0 JSON template system—one page.dds-vibe-academy.json file that registers all sections in a defined sequence. A sticky navigation bar controls eight content tabs: What's New, Curriculum, Evolution, Archive, Blog Feed, Useful Links, Manifesto, and FAQ. Each tab is its own Liquid section with isolated CSS scope, embedded JavaScript, and structured data. The CSS namespaces follow the pattern dva-[section-abbreviation]__ (e.g., dva-cm__ for curriculum), scoped under #{{ section_id }} using Liquid's section ID system. All colors use oklch(), the perceptually uniform color space with baseline browser support since 2024, with DDS palette tokens defined as CSS custom properties per-section.

Authoring and Deployment: Claude Does the Work

Claude Opus 4.7 authored every section file—12 Liquid sections plus 1 JSON template, approximately 6,400 lines of code across roughly 500 KB. The authoring phase followed strict scoping patterns to prevent style bleed between tabs. Google Antigravity then deployed every file to Shopify through the Shopify MCP (Model Context Protocol), running Claude Opus 4.6 as its agent model inside Google's agentic IDE. The deploy flow read local .liquid files and pushed them directly to Shopify's theme asset API via MCP—no CLI, no admin panel, no manual save clicks. Fourteen files total: twelve sections, one template, and one OG image, all shipped autonomously.

Cowork Changed Everything

Here's where it gets interesting. After the initial deploy, Anthropic's Cowork ran an autonomous browser audit—navigating every tab, checking every link across multiple viewports, and producing a structured findings report. The results were humbling: 30 broken URLs identified that had accumulated silently across classes over months, a z-index collision between the sticky navigation and anchor-positioned popovers, and three FAQPage JSON-LD schema corrections needed for proper search visibility. Every fix was applied in a second deploy cycle—again, by AI agents, not by hand. The founder's own assessment cuts to the core: "The biggest surprise—the audit step matters more than the authoring step. Cowork found things no human code review would catch."

Methodology: DDS Vibe Coding

The build process is documented as a methodology called DDS Vibe Coding with five pillars: constraint-first design (define what the system should do, let the model figure out implementation), progressive disclosure (ship simplest version first, add complexity only when earned), deterministic rails (CSS scoping, naming conventions, and schema rules that prevent drift), context hygiene (keep the model's context window clean by running one section per conversation), and sovereign fallback (every workflow must have a path that doesn't depend on someone else's uptime). The Build Manifest section between hero and tab nav serves as a deployment receipt, naming every AI agent involved and including CreativeWork structured data for AI search visibility. This transparency about the build process itself is increasingly important as AI-generated content proliferates across platforms.

The Curriculum Grid and Tooling Details

The curriculum tab renders the DVA Atlas Grid—a scalable 2-axis responsive CSS Grid with client-side filtering allowing users to group 49 nodes by Ring (Foundation, Development, Application, Mastery) or Lane (the 10-lane taxonomy including Claude, Gemini, Antigravity, and others). The FAQ tab contains 14 entries with proper FAQPage JSON-LD schema markup, native