I'm Testing "Vibe Marketing" with Claude Code (Building a Complete Funnel in One Sitting)

Hey, it's G.

I came across this concept called "vibe marketing" from James Dickerson (spent $100M+ on ads, knows his stuff). The idea: use Claude Code to build entire marketing systems - ads, landing pages, lead magnets, video creatives - in a single sitting.

Sounds ambitious. Maybe too ambitious.

But I'm curious enough to try it for AI For Pinoys and document what happens.

Full transparency: I've never built a "proper" marketing funnel before. I've done landing pages (mostly with Lovable), written some blog posts, posted on social media. But a complete, integrated system? First time.

So this is me learning in public. Again.

What is "Vibe Marketing" Anyway?

From what I understand, it's the marketing equivalent of vibe coding.

Vibe coding: Documentation → Vocabulary → Tools → Ship products

Vibe marketing: Research → Skills → Build → Get customers

Same terminal-based workflow. Same AI-powered approach. Different output.

Instead of building apps, you're building:

Landing pages with conversion-focused copy
Lead magnets that actually provide value
Ad creatives (static and video)
SEO content pages
Email sequences

All from Claude Code. All in one sitting (theoretically).

Why I'm trying this:

Right now, my AI For Pinoys marketing is... scattered. I post when I remember. I write when inspired. I have no system.

If I can build a proper funnel and test it, maybe I'll finally have a repeatable way to grow the community instead of hoping people stumble upon it.

The Three-Layer System I'll Be Testing

According to James, most people skip straight to "build stuff" and wonder why AI gives them generic garbage.

The actual workflow has three layers:

Layer 1: Research (Before You Create Anything)

What this means:

Deep market research using Perplexity MCP
Competitor analysis (what's working for others)
Website crawling with Playwright (pulling real inspiration)
Finding angles others are missing

Why this matters: More context upfront = better output. If you feed Claude vague instructions, you get vague results. If you feed it comprehensive research, it has something real to work with.

What I'll be researching:

Other AI education communities (how they position, what they offer)
Filipino tech communities (how they market, what language they use)
Successful course creators (their funnel structures)
DTC brands (ad creative frameworks that convert)

Layer 2: Skills (Custom Instruction Manuals)

What these are: Reusable frameworks that train Claude Code on specific marketing tasks.

Think of them like the CLAUDE.md files from vibe coding, but for marketing disciplines:

Direct response copywriting skill
Ad ideation skill
Long-form content skill
Brand positioning skill
Lead magnet creation skill

James has built 17 of these. I'll start with 3-5 and see if I can actually use them.

How to build a skill (the process I'll test):

Pick a marketing discipline (e.g., landing page copywriting)
Research it deeply with Perplexity MCP
Feed that research to Claude Code
Have it generate the skill file
Refine by testing against real outputs

What I'm testing: Do custom skills actually produce better output than generic prompts? Or is this just extra work for marginal gains?

Layer 3: Build (Stack Research + Skills = Assets)

The actual creation:

Landing pages with conversion-focused copy
Lead magnets (guides, templates, resources)
Ad creatives (static images, video ads)
SEO content pages
Email sequences

All built from Claude Code in the terminal.

The stack: Research layer feeds into Skills layer feeds into Build layer.

More like an assembly line than random prompting.

My Experiment: Building a Full Funnel for AI For Pinoys

I'm going to attempt what James demonstrated: build a complete marketing system in one sitting.

The goal: A working funnel that captures leads and converts them to community members.

What "one sitting" actually means: Probably 4-6 hours of focused work. I'll time it.

Step 1: Market Research (Using Perplexity MCP)

What I'll research:

Competitive landscape (other AI communities for Filipinos)
Pricing models (free vs. paid, what works)
Market gaps (what's missing that I can fill)
Messaging that resonates with Filipino entrepreneurs

Questions I'm asking Perplexity:

Who are the top AI education communities targeting Filipinos?
What pain points do Filipino solopreneurs have with AI tools?
What lead magnets work best for tech education communities?
What ad angles are successful AI course creators using?

What I'm looking for: Specific frameworks, actual examples, real data. Not generic advice.

Step 2: Brand Positioning (Claude Code + Skills)

What I'll generate:

Full positioning framework
Ideal customer profiles (who I'm actually serving)
Messaging pillars (core themes)
Voice guidelines (how AI For Pinoys should sound)

The skill I need to build first: Brand positioning skill trained on successful community brands.

Why this matters: Right now, AI For Pinoys positioning is vague. "AI education for Filipinos" could mean anything. I need clarity on exactly who I serve and why they should care.

Step 3: Lead Magnet Creation

What I'm making: A practical guide or template that solves a specific problem for my audience.

James's approach that I'm stealing:

Record a working session (me actually building something or teaching a concept)
Transcribe it
Have Claude Code extract the frameworks into a polished guide

My topic (probably): "The Complete Claude Code Setup for Filipino Entrepreneurs" - covers installation, configuration, first projects, common issues.

Why this format: Not an AI-generated ebook that sounds generic. Real thinking from my actual experience, extracted and formatted properly.

Step 4: Landing Page (Built in Terminal)

What Claude Code will generate:

Headline variations (testing different angles)
Benefit stacks (why people should download the lead magnet)
Social proof sections (testimonials from community)
Form and CTA copy

The skill stack: Brand positioning skill → Copywriting skill → Landing page skill

Each layer informs the next.

What I'll test: Does terminal-built copy actually convert better than what I'd write myself? Or do I still need to heavily edit?

Step 5: Ad Creative (Static + Video)

Static ads via Glyph MCP + Nano Banana Pro:

Multiple format variations (square, vertical, landscape)
Different hooks and angles
Testing messaging from the research phase

Video ads via Remotion:

Animated video ads generated programmatically
Claude Code writes the code that generates the video
Iterate on CSS, fonts, pacing through screenshots and feedback

Inspiration source: Facebook Ad Library studying DTC brands.

Even though I'm not selling physical products, DTC brands test more, spend more, and find hooks that convert. I'll adapt their frameworks.

Time expectation: James spent 30 minutes iterating on one video ad. I'm budgeting 1-2 hours for this entire step.

What this is: A keyword-targeted page that integrates the lead magnet and captures organic traffic over time.

Example topic: "How to Use Claude Code for Business Automation in the Philippines"

The structure:

Comprehensive guide format
Integrates the lead magnet naturally
Optimized for search
Built to rank and convert

The Tools I'll Be Using

Here's my planned stack (adapted from James's):

Core engine: Claude Code (obviously)

Research: Perplexity MCP for market research and competitor analysis

Website crawling: Playwright to pull inspiration from successful pages

Image generation: Glyph MCP calling Nano Banana Pro (or maybe just Leonardo AI if that's simpler)

Video ads: Remotion for programmatic video generation

Transcription: WhisperFlow for the lead magnet recording

Skills: 3-5 custom marketing skills I'll build first

What I might skip: Some of these tools I haven't used before. If Remotion is too complicated, I might just do static ads first and add video later.

This is an experiment, not a production system. Yet.

What I'm Actually Testing

Beyond "can I build a funnel," here are the specific questions I want answered:

Does the three-layer system actually work? Research → Skills → Build vs. just prompting and hoping

Are custom skills worth the setup time? Or is it faster to just write good prompts each time?

Can terminal-based marketing match visual builders? Claude Code vs. Canva, Unbounce, traditional tools

Does AI-generated copy convert? Or does it still need heavy human editing?

How much time does this actually save? "One sitting" sounds great, but what's the real time investment?

What's the quality vs. cost tradeoff? AI tools cost money. Is the output good enough to justify it?

Following This Experiment

I'll document everything across:

Newsletter (2 weeks from now):

Complete breakdown of the process
What I built
Time and cost analysis
What worked, what failed
Whether the funnel actually converts

YouTube (screen recording the whole thing):

Watch me build this live
See the mistakes, iterations, dead ends
Real-time problem solving
No editing out the messy parts

Dashboard (final assets if they work):

The skills I build
Templates that worked
Full funnel breakdown
What to copy, what to avoid

Timeline: Building this funnel this week. Report in 2 weeks after testing it live.

My Honest Expectations

Best case: Build a complete, working funnel in 6 hours that actually converts leads and grows AI For Pinoys organically

Realistic case: Takes 10-12 hours spread over a few days, some parts work great (landing page), some need heavy editing (ads), but I end up with a system I can refine

Worst case: The whole thing is too complicated, the tools don't play nice together, I waste a week and $200 in tools, but I document it so you know vibe marketing isn't ready for prime time yet

All three outcomes teach us something.

Want to Try This Too?

If you're building any kind of product or service and need marketing:

Pick one funnel to build (don't try to do everything)
Start with research (Perplexity or manual)
Build 1-2 custom skills for your specific needs
Build the assets with Claude Code
Test and iterate

We can compare notes. See what works for different industries and audiences.

One ask: If you try this, share your results. Let me know what broke, what worked, what I should watch out for.

Starting This Week

I'm blocking out a full day this week to attempt the "one sitting" challenge.

I'll screen record the entire session and post it uncut on YouTube. You'll see:

Every prompt I use
Every mistake I make
How long each step actually takes
The real quality of the outputs
Whether I give up halfway through or actually finish

No polish. No "here's what I built" summary without showing the work. You'll see the whole messy process.

Then in two weeks, I'll share whether the funnel actually works. Real metrics:

Landing page conversion rate
Lead magnet downloads
Ad performance (if I run them)
Time investment vs. results
Cost vs. value

Let's see if vibe marketing is real or just hype.

P.S. - If you've tried building marketing funnels with AI before, reply and tell me what works and what doesn't. I'm going in mostly blind and would love to skip the obvious mistakes.

P.P.S. - Yes, I'm aware this might be overcomplicating things. Maybe I should just write good copy manually. But I won't know until I try the system properly.

The Core Question:

Can you really compress the research-to-launch marketing loop into a single sitting using AI tools? Or is this just wishful thinking from people who've already mastered marketing and are now speeding up with AI?

I genuinely don't know. Let's find out.

Update schedule: Full breakdown in 2 weeks with all assets, costs, time investment, and honest assessment of whether this is worth it.