Vibe Coding Explained with Andrea Abbondanza | James Dooley

James Dooley: Vibe coding. The amount of people that reach out to me and ask, what is vibe coding? Today I am joined with Andrea Abbondanza, who has been doing a lot of vibe coding recently. He is a specialist in SEO for restaurants. So for anyone who is watching this, Andrea Abbondanza, what is vibe coding to start with?

Andrea Abbondanza: It is coding with vibes. That is what it is. I feel like I am a full stack senior developer. I am using Claude Code and recently I have been implementing it inside Anti Gravity. I mainly use Claude Code for creating the whole process, building agents and building orchestrations. Then I use Google mainly for Flash. Now we have got Flash 1 for debugging. I have not slept the last three days, James Dooley. I got so excited. It is great. With vibe coding, I was actually able to create projects that I had been waiting weeks or months to build because now I can do it myself. The tool basically helps you to build stuff, but it does not work like a normal LLM where it tells you what to do and then you copy and paste. It pushes things live straight away and then it debugs. The main problem for coders was debugging. Now you do not have that problem anymore. You can integrate GitHub. You can integrate a lot of technologies. It is crazy. I was able to create a full online AI avatar yesterday. The hardest part was integrating Facebook and YouTube APIs because Facebook is a pain to deal with. But the vibe coding has been great. I am so excited. It feels like when I was playing games when I was 13 and getting that same adrenaline and dopamine. I have been coding for three days in a row with my friends. I am not even sleeping. I am in Bangkok and I am not even going out. I am just vibe coding.

James Dooley: That is crazy. So what is it that you are actually vibe coding then? Are you vibe coding any apps? Are you vibe coding just for websites and programmatic SEO? What are you doing with regard to vibe coding?

Andrea Abbondanza: Yes. The first thing I want to do is build AI avatars. You have probably seen a lot of those getting traction lately. Usually it is based around a persona. You can use tools like Gen for visuals and ElevenLabs for voice, and tools like image generators to create assets. Then I am testing video workflows as well. The whole process can be created with vibe coding. That is the main project I have been working on. The other part is memory. LLMs hallucinate when you give them too much data, so you need to build systems properly. Because the memory is short, you need to create agents that handle different processes. For example, you would have one main agent that acts like a marketing manager and handles operations. Then you would have a social media manager agent that handles content creation and publishing. Then another agent that handles publishing across platforms like YouTube and Instagram. Then you need to connect all the APIs to make them work. When each agent is specialised, you then use an orchestrator. The orchestrator is another agent that manages everything and works like a supervisor. When you combine all of this, you get a full workflow that actually works. That is the project I am working on. I am here with my friend Arie, who is an engineer, so we are sharing knowledge. He was able to replicate something similar to Whisper, build an app and a website, and it can be done at a fraction of the cost. Big companies spend millions building this stuff, but now it can be done in days. That is a game changer.

James Dooley: Yes, it is crazy. So with regard to the different LLMs for vibe coding, what do you prefer? Is it GPT or Claude Code? A lot of people are saying Claude Code seems to be the best. What are you preferring overall?

Andrea Abbondanza: I am mainly using Claude Code. I am running a workshop at SEO Mastery Summit, and I want to make it actionable. I want people to download Claude Code and Cloudflare so they can push things live. I have also been testing Anti Gravity, which is Google’s platform that works with different LLMs. You can use OpenAI, Claude, and Google Flash. Google Flash is very fast but runs out of credits quickly. The best approach in my opinion is using Claude Code for building and structure, and then using Google Gemini or Flash for debugging. Claude pushes things live, which is why I prefer it. OpenAI I use less. Usually only when I run out of credits elsewhere. It can get expensive, but if you are saving time on development, the cost is worth it.

James Dooley: Yes, for me, when we tested GPT against others, the context window was not strong enough, so it hallucinated more. Claude Code handled context much better. So are you using Perplexity with Claude Code or separately?

Andrea Abbondanza: I have not tested Perplexity heavily for building. I use it more for research. For example, if you want an agent to pull real-time data like news or updates, you use Perplexity to feed that information into your system. What works best is using different LLMs for different tasks. For example, if GPT is better for writing, you use it for content. Then you use another model for reviewing, and another for structuring. You define the outcome clearly and assign roles to each agent. Another important thing is when you get an output you like, ask the LLM to show you the prompt that created it. That way, next time you can reuse the prompt and reduce hallucination.

James Dooley: Yes, that is smart. You are basically getting AI to improve your prompting so you do not need to do it manually.

Andrea Abbondanza: Exactly. You build the agent once, and then the agent handles the prompting for you. You just review the outputs.

James Dooley: That is quality. Anyone watching this, are you vibe coding yourself? What are you building? Leave a comment and let me know. If you want to come on and talk about your setup, your agents, and your preferred LLMs, I would love to dive deeper into vibe coding. It is moving fast and hard to keep up. Andrea Abbondanza has been doing it non-stop for the last 72 hours in Bangkok. It has been a pleasure having you on.

Andrea Abbondanza: Thank you so much.