A Complete Curriculum on Collaborative AI Development
Welcome to Introduction to Vibe Coding! This curriculum is based on Rozar's transformative workshop on building with AI as a true collaborative partner.
"I want to transform the developers, the audience, people participating from treating AI like a magic wand to understanding it as a collaborative partner, building apps they never thought possible and maintaining understanding, control and quality judgment that matters."
Rozar has been "a dev on this chain for about four years now" and won "the last builder battle hackathon with a prediction market and the blaze protocol." He's built the XMI protocol ecosystem—"at this point, like ten different apps"—including many developer tools and resources.
By the end of this course, you will be able to:
"Most people treat AI like a magic wand expecting it just to do the thing with other things happening. Bad things happen there."
Many developers approach AI with a "do this for me" mentality, which leads to:
"It has to be a collaborative partner. I know I said this again, but it's so important. You get to treat it like a collaborative partner. Yes, it's a tool. But the better mental model for it when you're working with it is as almost like an entity, a partner that you're working with, with strengths and weaknesses."
Key Principle: AI is not just a tool—it's a partner in building.
"You're going to learn AI has strengths and weaknesses. They don't know everything. They're trained. The way they work is they are trained on a data set and then released."
You'll also learn your own strengths and weaknesses:
"You also are in the same boat. You have things you're good at, and there's things you don't know. And you need to be aware of those because there's gonna be many a times when the AI is going to try to tell you something they don't know."
"Synergy is when the whole is greater than some of the parts, like one plus one equals three sort of scenario. When you work with an AI together, and this is working well, you're often going to come up with solutions that you could just never have done alone."
"It's this partnership paradigm, and it's one where there is a bidirectional understanding between you and your AI system."
"Solutions are going to surprise you. Solutions you can't imagine alone. And you have breakthrough insights in the collaboration when you do it correctly. You might start a session and end up with something you completely did not expect. And that's good. That's supposed to happen."
"The AI will explicitly trust you by default. It won't question your assumptions or pushback on any confusion. This creates a dangerous dynamic if you're not aware of it."
This is the core challenge of vibe coding: the AI trusts you completely, even when you're wrong.
"You must approach each interaction knowing there's a possibility you are entirely wrong about the solution. Like you're trying to tell it something that just does not make sense."
If you keep telling the AI something that's impossible, "it's going to try and find a way to make it possible."
The solution is developing "intrapersonal awareness"—knowing your own mind:
"You need to know what you don't know"
Be aware of your preferences and avoidances
Move from building mode to learning/research mode
"You need to take responsibility for the quality of the context you provide"
"Constant awareness is very important."
"I call them games because it's a collaborative exercise you're doing with this AI system. And that's the way I think about it. You're playing a game with them. Because there's rules and you're trying to find this win-win together and you're aiming to achieve synergy in the process."
Based on: Habit 1 from "The 7 Habits of Highly Effective People"
"You use this game when you're stuck, avoiding something, or spinning on a problem."
"It's based on the most fundamental habit which is proactivity. You got to take responsibility for what you need to do. You got to solve blockers. You got to keep moving forward. You can't be reactive in life."
Purpose: Idea exploration—figuring out what's stopping you from moving forward.
"Often it's what you resist most that matters most."
Based on: Habit 2 - "Begin with the End in Mind"
Before building anything!
"We need to begin with the end in mind. We need to have a vision for what we're actually making before we make it. That's the most important thing about building stuff. You need to have a clear vision of what you're actually doing before you do it."
"Otherwise you're just kind of doing things with no real direction. It's like trying to go somewhere without knowing where you want to go. You're going to wander."
"Having a strong vision is very important. It's fundamental to anything you do because things always exist first in your mind before they exist in reality. Like something existing in your mind is your first creation. Only then after that can you make it manifest in the real world. And the clearer it is in your mind, the clearer it's going to become when you build it."
Based on: "First Things First" and "The One Thing"
"This is arguably the best one, if I may say so myself. It's powerful. It's one that blew my mind when I made it."
"If you know the concept of bottlenecks in software engineering, there's always like one thing in a system or constraint, and maybe like one thing that's slowing everything else down."
"What's the one thing you can do such that by doing it, everything else becomes easier or unnecessary?"
How it works:
"You always look for the bottleneck in the situation. Everything else becomes easier if you can solve that problem."
Rozar demonstrates building a workshop completion app live, starting with:
"I just want to start something simple that I can build on."
Using the Envision game, Rozar transformed a basic idea into something meaningful:
"An app where the audience can collect prompts and mark that they watched the demo"
"It's not a completion tracker. It's a transformation threshold. A witness to transformation. A place where devs can cross the threshold and see others who've crossed it."
"This app went from zero to fucking awesome pretty quick."
When building, Rozar asks critical questions:
"I think the back end, because you kind of want to build bottom up when you build software apps. You want to start with the back end and work your way up."
Notice how Rozar engages with the AI:
"So this is now becoming collaborative. It's asking questions. I'm giving answers. I'm thinking."
He doesn't just tell it what to do—he has a conversation, validates assumptions, and maintains control.
"It was going to try and build the react app in my markdown file thing, and I don't want that. So let's do this again. Press escape. Stop it, redirect it. Don't let it just take you for a ride."
Lesson: You're in the driver's seat. If it's going the wrong direction, stop it immediately.
"I kind of just like to make sure my assumptions are right."
Don't let the AI proceed based on wrong assumptions. Test things. Verify outputs.
Rozar demonstrates awareness of what AI can and can't do:
"I did know that it would have no problem doing all this because it's all just react CSS, HTML, UI components, straightforward stuff. So it should be able to do it no problem all at once."
But for complex backend architecture:
"I wasn't sure about the back end, secure architecture and like party kit... So I wanted to validate those things first and make sure they work properly."
"Generally what people do here is the lazy way to debug: just paste errors in. Sometimes it works. It just depends. If it works the first time, it's fine. If you have to do it more than once, that's when you're getting into tricky territory. Means there's some sort of misalignment. You got to be more specific."
"Clarifications are important. Found a security hole and also guided it to the correct architecture."
Don't assume the AI understood everything. Ask clarifying questions and verify understanding.
Once you've validated the approach and architecture:
"Now that it really understands what we're doing here, I just trust it to do it. That was the preparation. And I can just go, I can take a break."
Watch Rozar build a complete app from scratch using vibe coding
"Prompt engineering is useful. Yes. But think of it like a collaboration. You're going to think more like what are good collaboration approaches and principles and taking responsibility for this collaborative process and the quality of it."
"It's gonna want to do things, and you got to be ready to say no. Like, that's not what I wanted. Even if it's impressive and feels exciting and works."
"More importantly: building with understanding."
"You're not going to be a prompt engineer. You'll be a developer collaborating effectively with AI, understanding it, knowing when to trust it, knowing when to question it, and how to maintain quality judgment throughout the building process."
"You can go from idea to implementation without thinking there's some kind of barrier."
"A world of potential opened up. Limitless possibilities, options, and creative freedom."
"They fundamentally changed from the time shift of knowing what they are capable of."
Identify three areas where you have knowledge gaps in development. Write them down. Practice recognizing when AI discussions enter these territories.
Choose a simple app idea. Use the Envision game approach to transform it from a basic feature list into a compelling vision. Ask yourself:
Take a current project or problem. Use the focusing question: "What's the one thing I can do such that by doing it, everything else becomes easier or unnecessary?"
Build something small with AI. Practice:
by Stephen Covey
by Gary Keller
Research on distributed cognition
Rozar's preferred tool: "You can use it in terminal, just type 'claude' and it launches"
For note-taking and planning
For full-stack React apps
For real-time features
Cross the threshold and join the community of builders
"Congratulations to you. Congratulations to me. This is very exciting."
Now go build something amazing. Remember: you're not using AI as a magic wand—you're partnering with it to create things you never thought possible while maintaining understanding, control, and judgment throughout.