Skip to content
Aakash GuptaAakash Gupta

Complete Vibe Coding Tutorial: Build a Full Stack App with AI | Andy Carroll (Windsurf)

What if you could launch a product next week without writing a single line of code? Today, it's possible. Andy Carroll (a 16 year PM and vibe coding expert) shows you exactly how to vibe-code your way into building great products — by building a $1M app live, right in front of you. We're discussing: Preview – 00:00:00 What Is Vibe Coding? – 00:01:57 Vibe Coding Tutorial Begins – 00:05:18 Ad – 00:10:32 Ad – 00:11:20 Building a Full-Stack Mobile App – 00:12:16 Creating the AI Sports Reporter – 00:18:12 Developing AI Tools Live – 00:21:22 Ad – 00:30:33 AI Sports Reporter (Part 2) – 00:31:25 Reviewing the Learning Page – 00:34:18 Using Prompts to Generate a Brand Logo – 00:42:04 Designing an Infographic – 00:46:49 Why Prompt Quality Is Everything – 00:50:21 Recap of the Live Build Session – 00:53:59 How PMs Should Use These Tools – 00:58:34 The Future of Vibe Coding for PMs – 01:00:15 What Is Aisle Partners? – 01:04:08 Should PMs Consult, Stay Full-Time, or Build Their Own Product? – 01:07:24 Closing Notes – 01:11:20 Podcast transcript + infographic: https://www.news.aakashg.com/p/andy-carroll-podcast 💼 Check out our sponsors: Maven: I’ve just launched my unique curation of their top courses - http://maven.com/x/aakash Miro: The innovation workspace: your team’s canvas - http://miro.pxf.io/PO4WZX Amplitude: Try their 2-min assessment of your company’s digital maturity - https://bit.ly/4hl25RG 👀 Where to find Andy LinkedIn: https://www.linkedin.com/in/heyandycarroll Aisle Partners: https://www.aisle.partners 👨‍💻 Where to find Aakash: Twitter: https://www.twitter.com/aakashg0 LinkedIn: https://www.linkedin.com/in/aagupta/ Instagram: https://www.instagram.com/aakashg0/ 🔑 Key Takeaways 1. Vibe coding puts product creation in non-technical hands. Even after 15 years as a PM, Andy can't write code from scratch. But AI tools like Windsurf and Lovable made it possible for him to collaborate with AI to build real products without coding skills. 2. Front-load your planning to avoid rebuilding everything. He painfully learned that diving straight into coding creates expensive headaches. So, it’s best to use AI to draft strategy documents and architecture plans first, preventing the frustration of multiple false starts. 3. Set up a simple deployment pipeline immediately. GitHub for code, Netlify for deployment, and Superbase for databases is his recommended stack. Deploy frequently and early as waiting too long means facing hundreds of errors at once instead of fixing small issues. 4. Create PM deliverables in hours, not weeks. He built a detailed product roadmap and strategy documents in a single afternoon. These become living references in GitHub that team members can access anytime, eliminating version control nightmares. 5. Watch out for AI's eagerness to change your code. Windsurf has two modes: safe "chat" and powerful "write." Only toggle to write mode when you want changes made, otherwise, AI might drastically refactor your entire page when you just want a font color change. 6. Switch AI models when you hit roadblocks. Different models have different strengths. He uses Claude 3.7 for brainstorming, DeepSeek for specific tasks, and switches to GPT-4 when stuck. A fresh model may solve problems that the first one couldn't. 7. Validate ideas faster than ever before. Skip weeks of perfecting logos, brands, and pixel-perfect designs. Build something "good enough" quickly, get real feedback, and iterate based on actual user responses rather than internal debates. 8. Target your creativity where it matters most. AI tools eliminate 90% of implementation busy work. Use templates for standard elements like landing pages, then focus your team's energy exclusively on the features that truly differentiate your product. 9. Automate status reports and presentations. He generates comprehensive project updates directly from his roadmap progress, feeding them into presentation tools like Gamma. This eliminates hours spent creating the same PowerPoint slides teams have made for decades. 10. People are launching profitable side hustles with vibe coding. Entrepreneurs like Peter Levels and John Rush build and monetize micro-SaaS products through vibe coding. The approach allows testing ideas quickly and pivoting without emotion when something doesn't work. #AI #startup #vibecoding 🧠 About Product Growth: The world's largest podcast focused solely on product + growth, with over 170K listeners. Hosted by Aakash Gupta, who spent 16 years in PM, rising to VP of product, this 2x/ week show covers product and growth topics in depth. 🔔 Subscribe and like the video to support our content! And turn on the bell for notifications.

Aakash GuptahostAndy Carrollguest
May 15, 20251h 12mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Vibe coding with Windsurf: plan well, ship fast, deploy safely.

  1. Vibe coding is framed as humans and AI tools “jamming” together to create products quickly, with exploration enabled by modern AI IDEs like Windsurf, Cursor, Replit, v0, and Lovable.
  2. Andy shows how Windsurf’s Cascade copilot (chat vs write modes) can edit code, run terminal commands, and speed up iteration—while requiring care to avoid overzealous refactors and unintended breakage.
  3. A central theme is that strong upfront context and lightweight PM artifacts (strategy, architecture, roadmap, prompt library) dramatically reduce rework and false starts when building with AI.
  4. They demo shipping real changes live—improving landing page contrast, generating/implementing an SVG logo and favicon, and adding an interactive “How it works” section—then deploying via GitHub + Netlify CI/CD.
  5. The conversation expands to team adoption, monetization (micro-SaaS, rapid experimentation), and how PMs should use AI to eliminate low-differentiation busywork and focus on “secret sauce” product value.

IDEAS WORTH REMEMBERING

5 ideas

Start with crisp context, not code.

Andy’s biggest early mistake was jumping into building (e.g., React) without planning, causing false starts and restarts; he now over-invests in clarity (strategy, architecture, roadmap) to save time later.

Use chat mode for safety; use write mode intentionally.

Cascade’s chat mode prevents code changes and is safer for exploration, while write mode can modify files directly—powerful but risky when the model decides to refactor more than you asked.

Deploy the simplest version early, then iterate in checkpoints.

Tools like v0/Lovable/Replit can “one-click deploy,” but in a fuller stack you must build a habit of frequent deploys; otherwise you’ll discover hundreds of errors only at the end and lose days.

Commit early and often to preserve momentum and rollback options.

Frequent GitHub commits (even if imperfect) create recovery points and trigger CI/CD; Andy highlights the pain of not knowing the last stable state when something breaks.

Switch models when stuck; different models excel at different tasks.

He uses broader “thinking” models (e.g., Claude Sonnet Thinking, DeepSeek-R1) for exploration and will swap to GPT-4 when debugging stalls, often unblocking progress quickly.

WORDS WORTH SAVING

5 quotes

I think it's really about humans and AI tools vibing together to create new products, new services, to bring new things to life.

— Andy Carroll

The right answer on the internet is to post the wrong answer.

— Andy Carroll

I didn't actually run through some of the planning sort of process with any kind of real methodical rigor, and that really came back to bite me.

— Andy Carroll

To me this just feels effortless, right? It's click a few buttons and, um, you do need to be really kind of careful though.

— Andy Carroll

When we start out, right, let's define what it is that we're actually trying to achieve, what is gonna make this unique, and let's focus almost all of our energy on making that part of the product really, really great.

— Andy Carroll

Definition and origins of “vibe coding” (Karpathy)AI IDE landscape: Windsurf vs Cursor/Replit/v0/LovableCascade copilot: chat mode vs write modePrompt libraries and planning frameworks (JTBD, PRFAQ, story mapping)Model switching for better results (Claude, GPT-4, DeepSeek)Deploy pipeline: GitHub commits + Netlify buildsCommon failure modes: memory/context bleed, security risk, deployment painPM impact: faster feedback loops and differentiation focusMonetization and micro-SaaS experimentation cultureFounder/consulting journey and AI training/implementation services

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome