Today I learned about “Vibe Marketing” — a concept from Greg Isenberg (Silicon Valley entrepreneur). His point: while you’re still Vibe Coding, you should already be Vibe Marketing. Most founders get stuck at step 1 (building the product) and never move on.
The 4-Step Framework
Greg breaks the AI-native product launch into four steps:
Step 1: Build with AI (where most people stop)
Use AI to write code and ship the product. Backend, auth, infrastructure — get the core functionality done.
This is “Vibe Coding” — the part everyone’s already doing.
Step 2: Design a Marketing Playbook with AI
Use AI to create your entire marketing strategy:
- Content format — what types of posts work for your audience
- Hook copywriting — attention-grabbing openers
- Lead-gen product — free tools or content that drive signups
- Reply rules — how to engage with comments and DMs
- Brand tone — consistent voice across all channels
- Weekly experiments — what to test each week
This is the step most founders skip. The product is done, but there’s no system for getting it in front of people.
Step 3: Let AI Agents Execute 24/7
Hand the playbook to AI agents that run it around the clock:
- Auto-post content on schedule
- Auto-run ad campaigns
- Auto-reply to comments and DMs
- Auto-follow up with interested users
- Auto-repurpose content for secondary distribution
The key shift: marketing goes from a manual task you dread to an automated system you monitor.
Step 4: Read the Data, Double Down
Look at the metrics and decide where to invest more:
- Bookmarks, shares, replies → content resonance
- Clicks, signups → conversion effectiveness
- These indicators tell you what’s working and what to cut
The Core Insight
Most founders finish building and then ask: “Why is nobody using this?”
The answer is they stopped at step 1. Building the product is only the beginning — without a marketing system, nobody knows it exists. Vibe Marketing means applying the same AI-first mindset you used for coding to the entire go-to-market process.
How This Connects to Vibe Coding
| Â | Vibe Coding | Vibe Marketing |
|---|---|---|
| What | Build the product with AI | Market the product with AI |
| AI role | Writes code, handles infrastructure | Writes copy, runs campaigns, engages users |
| Human role | Direct the vision, review output | Set strategy, review metrics |
| Mistake | Over-polishing before launching | Never starting at all |
The pattern is the same: describe what you want, let AI do the heavy lifting, review and iterate. The only difference is what you’re building — code vs. distribution.
Claude as a Strategic Thinking Partner (Not Faster Google)
A YC founder shared how they compressed 1 month of market research into 3 hours — not by asking Claude to “research this market,” but by feeding it raw data and asking strategic questions:
What they fed Claude:
- 8 competitor landing pages
- 3 earnings call transcripts
- 12 customer reviews
- 1 Reddit complaint thread
The questions that made it work:
| Question | What It Reveals |
|---|---|
| “What does every successful player in this market understand that their customers never say out loud?” | The unspoken insights — things everyone knows but nobody articulates |
| “Show me the 3 assumptions this entire market is built on, and what would have to be true for each one to be wrong.” | The market’s attack surface — blind spots, shared assumptions, uncontested gaps |
| “Write the 5 questions a top investor would ask to challenge this business model, and answer each one using only evidence from these documents.” | Stress-tests each assumption — every weak answer generates follow-up investigation |
| “What is the strongest version of this argument, and where does it still fail?” | Steelmanning — forces the AI to find real weaknesses, not strawmen |
Result: In 3 hours, a complete strategy document that looked like it came from a 10-year industry veteran. The tools didn’t change. The questions changed everything.
Key insight: Most people use Claude as a faster Google. The real unlock is using it as a thinking partner that challenges assumptions and finds blind spots. You don’t ask “research this market” — you ask “break this market’s assumptions.”
Product Management on the AI Exponential
Cat Wu (Head of Product for Claude Code at Anthropic) wrote the definitive piece on how PM practices are evolving:
The Core Problem
The traditional product management playbook assumes what’s technologically possible at the start of a project is roughly what’s possible at the end. Exponentially improving models break that assumption.
What Changes for PMs
| Old PM Playbook | New PM Playbook |
|---|---|
| Write long PRDs | Build working prototypes with Claude Code |
| Months of planning | Rapid experimentation, consistent shipping |
| Features locked at project start | Re-evaluate feature list every model release |
| Complex system design | Find the simplest approach that works |
| Document-first | Demo-first |
Practical Advice
- Less documents, more demos — With Claude Code + Opus 4.6, build a working prototype to show your idea instead of writing about it
- Every new model release → revisit your feature list — Things that were impossible 3 months ago might be trivial now. Remove workaround code you added to compensate for model limitations
- Keep it simple — Complex AI systems crash more. Find the simplest approach that works
- Track two things simultaneously — How AI changes the way you work, AND how it changes what’s possible in your product
| *Source: Product Management on the AI Exponential | The Vibe Marketing Manifesto | Vibe Marketing Skills for Claude | PM Skills Framework* |
Real-World Use Cases
- Startup launch — Build the marketing playbook while the AI-built product is still being developed.
- Market research — Feed Claude competitor pages, reviews, transcripts, and community threads to uncover positioning gaps.
- Product management — Use demos, prototypes, and model-release reviews to shorten planning cycles.
How LearnAI Team Could Use This
These examples demonstrate a crucial distinction for students: AI as tool vs AI as thinking partner.
- Tool mode: “Claude, research competitor X” → gets a summary (useful but shallow)
- Partner mode: “Claude, here are 8 competitor pages. What do successful players understand that customers never say out loud?” → gets strategic insight
Teaching students to ask the right questions — not just use the right tools — is exactly what LAI research should measure. The YC founder’s example could be a case study in an AI literacy course.