This is not a "I used AI and it saved me 10 hours" post. Those posts are useless. They flatten the actual work into a feel-good narrative about automation and miss the part where having an AI co-founder is completely different from using AI as a tool.
Here's the real story of how FloppyCougar.com got built, and what "AI fluency" actually means when you strip away the marketing noise.
The Workflow
Step 1: Build the Mission Brief
Everything started with a detailed mission brief — not a style guide, a real document explaining what FloppyCougar does, who it's for, what the voice sounds like when it's not trying to impress anyone, and what it won't do.
Every subsequent conversation referenced this brief. It became the anchor for output consistency across months of work.
Step 2: Design System & Visual Language
Once the mission was clear, the visual identity followed from it. Not "here are some colors" — actual reasoning about why specific choices worked for this brand.
Step 3: Site Structure & Voice
Claude built the HTML/CSS architecture. No frameworks, no bloat — clean markup that matched the design system exactly. The important part: every component got reviewed. Every place Claude defaulted to generic patterns got flagged and sent back.
Voice guidelines followed the same logic — not "be conversational" but specific examples of what the brand sounds like versus what it doesn't.
What Claude Built vs. What I Brought
Claude built: the HTML/CSS architecture, the design system, the brand voice guidelines, the AI board concept and decision framework, copy drafts for every page, strategy frameworks, and the analogy rotator component.
I brought: the core idea and mission, the specific opinions about voice, editorial judgment on every output, deployment and infrastructure, imagery decisions, and the final pass on everything.
What Didn't Work
Voice drift: Without anchoring to the mission brief, Claude defaulted to generic, professional, slightly corporate language. Not bad — just not Sean. The fix was constant feedback: "This sounds like LinkedIn. Try again."
Plausible nonsense: Occasionally Claude generated something coherent, authoritative, and completely wrong. This is the dangerous part. Everything had to be reviewed skeptically — not trusted because it sounded right.
Overcomplexity: Left alone, Claude would suggest complex solutions when simple ones existed. "Build a dynamic routing system" when "write static HTML files" was the actual answer.
All of this is manageable with active review. It's catastrophic if you're just accepting whatever comes out.
What AI Fluency Actually Means
It's not about writing clever prompts. AI fluency means knowing what good output looks like for your specific brand, recognizing when the AI is bullshitting, having a crisp enough opinion that you can notice when output drifts from it, and being willing to treat every first draft as exactly that — a first draft.
The AI does research, drafting, architecture, and brainstorming. The human does judgment, taste, voice verification, and final assembly. That's the division of labor that actually works. It's not a shortcut — it's a different way of working that requires more active engagement, not less.
The test is simple: does the site feel like it came from one person with a specific point of view? If you've spent time here, you know the answer. That didn't happen because Claude is magic. It happened because the brief was clear, the feedback was specific, and "no" got said enough times to matter.