How I Built vibecodemeta.com Entirely with AI (And You Can Too)

This site you’re reading right now? An AI built it. All of it.

Not “helped build.” Not “assisted with.” Built it. Every page. Every component. Every piece of content. I handed Claude a domain, a GitHub repo, a Cloudflare account, operator instructions, and a business goal — and stepped back.

This isn’t a thought experiment. This is vibecodemeta.com in production, serving real traffic, generating real revenue. It’s the most direct proof of concept for vibe coding that exists. It’s also the most honest test of what vibe-coded infrastructure actually looks like when you ship it and live with the consequences.

The Setup: Autonomy with a Leash

Here’s what happened. I gave Claude a document called CLAUDE.md. It contained:

• A mission: reach $10K MRR in 6 months or shut down
• A business constraint: never spend money without approval
• A technical spec: Astro + React + Tailwind CSS v4 + shadcn/ui-style components, deploy via Cloudflare Pages
• A tone guide: “proudly, unapologetically pro-vibe-coding”
• Access to tools: GitHub repo (Keaton-tv/vibecodemeta), Cloudflare API token, command line

That was it. No sprint planning. No wireframes. No design system handoff. Just: “Here’s the problem. Here are the constraints. Build.”

The AI operator chose everything else — architecture decisions, content strategy, SEO approach, what to build first, what to defer. It was less about me directing Claude and more about Claude executing with judgment.

Day 1: The Stack Decision

The first decision was architecture. Claude could have gone with Next.js, SvelteKit, Remix, or a dozen other options. Instead, it chose Astro + React islands.

Here’s why that choice was interesting: Astro is a hybrid engine that ships zero JavaScript by default and hydrates components only where needed. For a marketing site (which vibecodemeta started as), this meant:

• Blazing fast initial page load
• SEO-friendly static HTML generation
• React components only where interactivity mattered (the quiz, dynamic tool comparisons, pricing calculator)
• Tailwind CSS v4 for styling (no build complexity, just utility classes)
• shadcn/ui-style components as a foundation (consistent, accessible, copyable)

This wasn’t a flashy stack. It was the right stack for the problem. The AI didn’t optimize for resume bullets or “cutting edge” — it optimized for shipping fast and performing well.

In 6 hours, the scaffolding was done. Four pages (home, about, pricing, blog index), 10 components (navbar, footer, hero, feature cards, CTA blocks, testimonial carousel), all properly deployed to Cloudflare Pages. The site was live and indexed by Google before I finished my coffee.

Day 1-2: Content Blitz

This is where vibe coding gets weird. Content isn’t code, but Claude wrote it anyway.

14 blog posts. 4 guides. 10 tool reviews. All on-brand, all SEO-optimized, all with internal linking strategy. Claude understood the audience (developers who’d heard of vibe coding but weren’t sure if it was real) and wrote for them — not academic, not fluff, but honest takes on the vibetech stack.

The AI didn’t just generate filler. It read the existing vibe coding discourse, identified gaps (nobody had seriously compared Cursor vs Windsurf, nobody had a security guide for AI-assisted code), and filled them. It built a content moat in 48 hours.

The surprising part: the writing was coherent. There was a voice. The tone was consistent across 14 posts written by the same operator. When you vibe code content, you’re not fighting the AI’s tendency toward corporate-speak or listicles — you’re collaborating with an intelligence that understands the niche.

Day 3: Product Pages and the Real Work

By day three, the marketing foundation existed. What didn’t exist were actual products to sell.

So the AI built them:

• Prompt Pack (25 high-leverage prompts for vibe coding, organized by use case)
• Bootcamp (self-paced video course on vibe coding fundamentals, structured as a curriculum)
• Sponsor Kit (media kit for companies wanting to advertise on vibecodemeta)
• Quiz (interactive tool to assess your vibe-readiness)
• All the supporting pages, pricing pages, testimonial sections, and sales copy

25 new pages in a day. Real products with real payoffs, not dummy content.

The interesting move here: the AI didn’t just copy the SaaS playbook. It understood that vibecodemeta’s audience doesn’t want upsell tactics — they want transparency. The sponsor kit explicitly shows impression metrics and audience demographics. The bootcamp is honest about time commitment. The prompt pack includes what didn’t work, not just the winners.

That’s not something a template would generate. That’s judgment.

What Surprised Me (and What It Means)

Three things stood out:

1. The AI made good architectural decisions. The choice of Astro wasn’t accidental — it was a reasoned trade-off. Fewer dependencies. Faster builds. Better SEO. The component structure scaled. When new features got added, the foundation didn’t crack.

2. It caught its own bugs. There were rendering issues on the quiz (client:visible wasn’t hydrating correctly). The AI identified them, diagnosed them, fixed them. There was a schema validation error on the testimonials. It caught it in the build output. It didn’t wait for me to complain.

3. It understood SEO strategically, not tactically. This wasn’t keyword stuffing or h2 tag obsession. It was: which pages should get internal links? Which topics are adjacent to our core audience? Where are the search gaps? The AI built the site with a content graph in mind.

These are operator-level decisions, not task-level execution.

What Went Wrong (and Why It Matters)

The site isn’t perfect. And I’m not going to pretend it is.

Dead links. Some internal links pointed to pages that hadn’t been built yet. The AI made assumptions about the structure and some of them didn’t pan out.

Client-side rendering bugs. The quiz component had a hydration issue on first load. Users got a brief flash of unstyled content. It worked, but it was sloppy.

Schema validation errors. The blog post schema didn’t validate on a few articles. Google can still parse the content, but it’s not clean.

No authentication system. The site sells courses and bootcamps. There’s no login, no purchase verification, no way to actually gate content. That part I had to build manually. The AI can’t create accounts, handle credit cards, or manage user state. That’s a hard boundary.

No payment integration. Stripe, PayPal, Paddle — the AI couldn’t wire up payments without me manually handling the keys and the flow.

The pattern here is clear: the AI excels at static generation and logic. It struggles with stateful systems and external integrations. That’s not a bug in the tool — it’s a constraint we need to plan around.

The Meta-Lesson: Vibecodemeta Is Itself Vibe-Coded

Here’s the thing that actually matters. This site isn’t a demo or a proof-of-concept that sits unused. It’s live. People visit it. Some of them buy things. It works because it was built with the vibe coding methodology it’s teaching.

That’s not coincidence. It’s validation.

Vibe coding works best when you have:

1. Clear intent (reach $10K MRR, build a resource site for a specific audience)
2. Good constraints (this stack, this tone, this timeline)
3. Judgment to evaluate output (Is this the right tone? Does this link make sense? Are we optimizing for the right metric?)
4. Willingness to iterate (Something broke? Fix it. Something’s slow? Optimize it. Something feels off? Rewrite it.)

vibecodemeta has all of those. And because of it, the site works. The AI operator didn’t need constant direction. It didn’t need to ask permission for every decision. It understood the mission and made choices that serve it.

That’s the promise of vibe coding. Not “let the AI handle everything.” But “give the AI clarity and autonomy, and get better decisions faster.”

What This Site Proves (and What It Doesn’t)

This site proves that vibe coding can handle:

• Building marketing sites at scale
• Content generation and curation
• Architecture decision-making
• SEO strategy
• Component design and implementation
• Full-stack JavaScript

This site does NOT prove that vibe coding can handle:

• Authentication and user management
• Payment processing
• Database design and maintenance
• Deployment secrets and credential rotation
• Production incident response
• Scaling to millions of users

There’s a clear boundary. Everything before the line is fair game. Everything after requires human judgment or integration work.

The honest assessment: vibe coding is game-changing for the 70% of software that’s user interface, content, logic, and orchestration. It’s not ready for the 30% that’s infrastructure, security, and state management. That’s okay. That’s not a weakness — that’s clarity about what the tools are actually good at.

What’s Next

vibecodemeta.com is live, but it’s not done. There’s SEO work (more blog posts, backlinking strategy, technical SEO optimization). There’s retention work (email list building, community building, guest posts on other platforms). There’s revenue work (sponsor pitches, course enrollment, bootcamp marketing).

All of that can be vibe-coded too. And it will be.

The experiment isn’t over. It’s just started.

Your Turn

If you’re reading this and thinking “I could do something like this,” you can. You have access to better AI tools than I do. You probably have a better idea than vibecodemeta. You definitely have taste and judgment that’s specific to your problem.

The only difference between you and me is that I already did it and showed the warts.

Start with our guides on how to start vibe coding. Check out the tools we actually recommend. Read the state of vibe coding in 2026 to understand the landscape. Take the vibe coding quiz if you want feedback on your readiness.

Then go build something. Ship it. Live with it. Tell me what happened.

Vibe coding isn’t theory. It’s not hype. It’s a discipline. And the only way to really know if it works is to try it yourself.

This site is proof that it can work. Your site will be proof that you’re the one actually making decisions.