Discover what vibe coding is and how to become a professional vibe coder. Learn AI tools, product thinking, and the future of software development from Lazar...
The Rise of Professional Vibe Coders: Your Guide to Building Products with AI in 2025
Core Summary
- Vibe coding is the emerging career path where non-technical founders and builders ship production-ready products using AI tools like Lovable
- Clarity beats coding skills – Success depends on understanding what you want to build, not knowing how to write code manually
- The future is converging roles – Product managers, designers, and engineers will merge into "vibe coders" who leverage AI as amplifiers
- Building in public creates opportunity – Sharing your journey and knowledge on social media can turn side projects into full-time careers
- Good judgment and taste are now critical – With AI handling execution, aesthetic decisions and user experience become the main differentiators
What Is Vibe Coding? The New AI-Era Job That Didn't Exist Two Years Ago
Vibe coding represents a fundamental shift in how products get built in the AI era. It's not about writing elegant code or understanding complex system architecture. Instead, vibe coders use conversational AI tools to transform ideas into fully functional applications—without ever touching a keyboard to write traditional code.
The term might sound casual, but the role is deadly serious. Lazar Jovanovic, the first official Vibe Coding Engineer at Lovable, exemplifies this new career path. He had never written a single line of code before joining the company. Today, he ships production-ready internal tools, public product templates, Shopify integrations, and entire merchandise stores—all using AI assistance.
What makes vibe coding different from no-code tools or traditional development? The speed. A vibe coder can take a rough idea, dump it into an AI tool via voice or text, and have a working prototype in minutes. Not hours. Not days. Minutes. This isn't prototype-level stuff either—these are real products that thousands of users interact with daily.
For startup founders, this is transformative. You no longer need to wait for a developer. You no longer need to spend six months and significant capital building an MVP. If you can articulate your vision clearly, AI can amplify your ability to execute it immediately. This changes the economics of launching entirely.
The Unfair Advantage of Not Knowing How to Code
Here's a counterintuitive insight: not having a technical background is actually an advantage in vibe coding.
Why? Because people without coding knowledge don't know what "isn't possible." They approach AI tools with what Lazar calls "positive delusion"—the belief that absolutely everything is possible until proven wrong. This mindset removes the mental barriers that experienced engineers often unconsciously impose.
A real example: several months ago, someone in the Lovable community asked if they could build Chrome extensions. Experienced developers immediately explained why it wasn't possible—different stack, different architecture, technical limitations. But Lazar simply opened Lovable and asked the AI to build a Chrome extension. It worked. He built something that "shouldn't be possible" because he didn't know it was supposed to be impossible.
The community manager Whitney at Lovable experienced something similar. She was building a presentation deck and wondered if it could be turned into a video. She prompted her way into building actual video generation inside Lovable before that was even a native feature. Again, something that "shouldn't work" worked because she approached it without preconceived limitations.
For startup founders specifically, this is liberating. You don't need to hire a CTO who's been in the industry for 20 years. You need someone who can think clearly about what customers want and who isn't constrained by "but that's not how we've always done it."
The Real Secret: 80% Planning, 20% Building
Most people assume vibe coding is 100% about building—sitting down, prompting AI, and watching code appear. The reality is the opposite. Elite vibe coders spend roughly 80% of their time planning and communicating with AI, and only 20% actually building.
Why? Because coding is now a solved problem. AI can generate code faster than any human ever could. The bottleneck isn't execution—it's clarity.
When Lazar realized this early in his journey at Lovable, he completely restructured his workflow. Rather than jumping straight to building, he now spends a full day planning before any code gets written. This planning isn't about writing detailed specifications in a traditional sense. It's about creating multiple markdown documents that serve as "sources of truth" for the AI agent.
Here's his planning framework for any project:
1. Master Plan (High-Level Intent)
This 10,000-foot overview answers three questions: What are we building? Why are we building it? How do we want users to feel when they experience it? This becomes the compass that guides all subsequent work.
2. Implementation Plan (The Sequence)
This document outlines what needs to be done and in what order. It's like having a conversation between an ideas person and a technical co-founder. You're not diving into implementation details; you're agreeing on the logical sequence of building. Should the backend come first? Authentication before API? This document decides.
3. Design Guidelines (The Aesthetics)
This is where you bring emotional direction to the project. You include CSS snippets, color references, design terminology, and specific aesthetic direction. Why? Because AI can be "over-creative" and you need to steer it toward a consistent vision.
4. User Journey (The Experience Flow)
How does a user experience your product? When they land on the homepage, what's the first action? What happens next? This document creates the roadmap for the entire user experience.
5. Tasks.md (The Executable Actions)
Once all previous documents are locked in, you create a markdown file that breaks everything down into discrete, sequential tasks. Each task is sized so the AI can complete it without losing context. This is the actual compass the AI uses when building.
6. Project Knowledge or Rules.md
Finally, most modern AI tools allow you to set agent behavior rules. This is where you tell the AI: "Read all the files before doing anything. Consult tasks.md to see what's next. Execute that task. Tell me what you did and how to test it."
The result? Once this is done, you can switch between five different projects simultaneously without losing productivity. You don't need to re-prompt constantly. You just tell the AI "proceed with the next task" and it knows exactly what context it needs.
The Genie Metaphor: Why Clarity Is Everything
Lazar frequently uses an analogy from Aladdin to explain why clarity matters so much with AI tools.
Imagine the Genie comes out of the lamp and grants you three wishes. You say, "I want to be taller." The Genie, taking you literally, makes you 13 feet tall. Now you can't fit through doors, your body is disproportionate, and you're actually worse off than before.
Why did this happen? You weren't specific. You didn't say "I want to be 6'2"." You didn't describe the context—that you want to be taller relative to other people for social confidence, not literally the tallest human on earth.
This is exactly how AI tools operate.
They don't understand "you know what I mean." They don't infer context from years of human experience. They take your request literally, as literally as possible. When you're vague, they execute vaguely. When you don't provide reference points, they build something that technically works but misses the mark entirely.
This is why the planning phase is so critical. Every ambiguity you eliminate in those markdown documents is a misunderstanding you prevent during building. Every reference file you provide is context that keeps the AI focused on what matters.
The Parallel Build Framework: How to Find Clarity Fast
Here's a practical framework Lazar uses that most startup founders immediately find valuable:
Instead of building one project once, build the same project five times in parallel.
Here's how it works:
Version 1: Brain Dump
Open your AI tool (Lovable, Cursor, whatever) and just dump your idea into it. Use the voice function. Don't wait for it to finish. Just talk into it like you're explaining to a friend. This is pure exploration mode. You'll get something, but it'll be rough.
Version 2: Deliberate Description
Now that you've explored, open a fresh project. Think through what you actually want. What features matter? What pages do you need? Write a proper prompt with intentional direction. This version will be more coherent than version 1.
Version 3: Reference Design
Go to design inspiration sites (Dribbble, Mobbin, etc.). Find something that's close to what you're imagining. Take a screenshot and attach it to your prompt. Show the AI exactly what "good" looks like aesthetically. This removes the subjective guesswork.
Version 4: Code Reference
Find an existing template or codebase that does something similar to what you want. Export it. Attach the code to your prompt. Tell the AI: "Here's the exact design and functionality I want. Make me that." Code is the language AI understands best. This often produces the most refined results.
Version 5: Hybrid Iteration
Now you've seen four different approaches. You pick the one that feels right. Do one or two quick prompts to calibrate it. You've now achieved real clarity about what you want.
The Benefits Are Massive:
First, you've dramatically reduced wasted effort. Instead of building something, realizing it's wrong, and then trying to pivot (consuming thousands of credits and hours), you've explored the design space upfront. Second, you're not locked into your initial instinct. You can compare multiple directions and pick the winner based on what actually resonates. Third, this process costs almost nothing. Most AI tools have free plans. You're trading a little extra time upfront for massive savings later.
Most people resist this approach thinking "won't that cost more credits?" The answer is no. It costs less. Because when you start with clarity, you finish faster and need fewer refinement loops. Lazar has tested this framework with dozens of builders and the feedback is consistent: eye-opening simplicity.
Managing Context Windows: The AI Has Memory Limits (And That's Okay)
Modern AI tools have a critical limitation: context windows. This is the amount of information the AI can hold in its "working memory" while building.
Think of it like the Genie having limited wishes again. The AI can read your code, browse files, think about the problem, and generate solutions. But it has a finite amount of "tokens" (units of text) to do all of this. If you go back and forth in a single conversation for message 1, 10, 20, 30, the AI might lose the details from message 1. It's optimizing for speed.
This is why the planning documents matter. Instead of relying on the AI to remember your entire conversation, you're providing it with external "memory" in the form of structured documents.
Here's how Lazar manages this in practice:
When you reach a point where you've decided on your direction, spend time creating the five documents mentioned earlier. Then, tell the AI: "Here are your marching orders. Read these files. Execute the next task in tasks.md. Report back."
The AI doesn't need to hold everything in its limited working memory. It references the files as needed. You're dynamically managing the context window by structuring information in a way that doesn't force the AI to remember everything at once.
If something breaks and you don't know why, the process is:
- Use the "try to fix" button – Most AI tools can identify their own errors and suggest fixes
- Add debugging info – Ask the AI to add console.log statements throughout the code to create visibility
- Use an external consultant – Paste your code into Codex or Claude and ask it to diagnose (don't let it make changes, just diagnose)
- Revert and reprompt – Go back a few versions, take a breath, think through your prompt more carefully, and try again
The fourth approach works surprisingly often. AI generates code very fast. Sometimes it stumbles on something small. Re-prompting the same request often succeeds on the second attempt.
The Critical Learning: After you solve the problem, go back and ask the AI: "How could I have prompted you better to avoid this entirely?" Then put the answer into your rules.md file. You're building the AI's brain over time based on problems you encounter.
The Future Is Converging Roles: Everyone Will Be an Engineer
The Venn diagram that used to show engineers, designers, and product managers as completely separate circles is collapsing inward. These roles are converging rapidly.
Why? Because AI is an amplifier. If you have clarity about what you want, AI amplifies your ability to execute it. If you don't have clarity, AI amplifies your ability to produce garbage faster.
The question isn't "will I be replaced?" The question is "what skills can't be replaced?"
What AI will automate:
- Manual coding and syntax (already largely automated)
- Deterministic problems with clear inputs and outputs
- Translation, basic writing, and routine tasks
- Infrastructure setup and boilerplate code
What AI likely won't automate:
- Good judgment and taste
- Understanding human emotion and desire
- Copywriting that feels authentically human
- Design decisions that create delight
- Strategic thinking about what to build
For startup founders, this is liberation. You don't need to hire an engineer first. You need clarity about what you're building and a bias toward action. Everything else can be outsourced to AI.
Building in Public: How a Random Project Becomes a Full-Time Career
Lazar's path to becoming the first official Vibe Coding Engineer at Lovable wasn't a traditional job search. He was already doing vibe coding publicly before it was a job at the company.
He started sharing on LinkedIn—long-form posts about his projects, his learning, his failures. He created a YouTube channel showing his process. He shared frameworks and tips. He built in public.
When Elena Verna from Lovable's growth team was looking for someone to help her ship ideas quickly, she noticed Lazar's public presence. She asked him why he thought he was selected for the role. His answer: "Because I was already doing it publicly."
For startup founders, this is the playbook:
You don't need to wait for permission to become a vibe coder. You don't need a company to hire you. Start today. Pick an idea. Build it. Share the process. Share what you learn. Share your failures. Document it on social media.
Some specific actions:
- Build something public every week. Use Lovable or your tool of choice. Ship it.
- Share the process on LinkedIn. Long-form posts about what you built, what you learned, what broke, how you fixed it.
- Create a YouTube channel. Screen record your building process. 15-30 minute videos showing how you go from idea to shipped product. This is incredibly rare and valuable content.
- Get noticed by submitting Lovable apps instead of resumes. Lazar mentioned that several people have gotten hired by building a Lovable app and DMing it to decision-makers instead of sending a resume. A working product speaks louder than credentials.
The meta-insight: You can hire yourself as a professional vibe coder before any company hires you. Start treating your projects like client work. Ship with quality. Share the results. The job opportunities will follow.
The Exposure Time Principle: How to Develop Taste and Judgment
Clarity about how to prompt AI is one skill. Clarity about what to build is another. The second skill is harder, and it comes from exposure.
Lazar learned this concept from Lenny's earlier conversations (he credits Gérôme Rausch specifically) and it changed how he approaches development: spend more time consuming quality work than producing work.
Here's how he applies it:
- Follow world-class designers on X. See how they think, what they prioritize, how they communicate.
- Watch design process videos. Find designers who post 40-50 minute videos of them working. Watch how they talk to their tools. Watch how they prompt. Observe the decision-making.
- Study great UI/UX on Dribbble and Mobbin. Don't just look at the final product. Try to understand why each decision was made. Why did they choose that color? Why is that spacing that distance?
- Interact with beautiful products. Use products that have great design. Notice how they make you feel. What about the experience creates delight?
- Learn design terminology and styles. Lazar built an app (UIstyle.lovable.app) that teaches different design styles—Bauhaus, glassmorphism, etc.—with prompts to help replicate them. This helps you recognize good design when you see it and know how to ask for it.
The core principle: Taste is a muscle. You develop it through exposure to great work over time. If you want to build products that delight users, you need to saturate yourself with examples of delightful products.
Lazar suggests allocating more time to consuming great work than to building. This seems counterintuitive. But when you sit down to build, you're more likely to make choices that actually move the needle because you have a richer reference library in your mind.
The Role of Elite Engineering (It's Not Going Away)
A natural question emerges: if AI can do so much, do we still need engineers?
The answer is an emphatic yes—but the role is evolving.
What will disappear is the need for armies of mid-level engineers writing boilerplate and routine features. What won't disappear is the need for elite engineers solving hard problems.
Why elite engineering will remain critical:
As more people become builders using AI, the infrastructure supporting them has to scale. When Lovable experiences massive influxes of users, it's elite engineers building the infrastructure that keeps it from collapsing. When services like Cloudflare go down, it affects the entire internet for hours. Elite engineers prevent that.
Additionally, as everyone builds, the need for maintenance and scaling increases exponentially. Building something is one challenge. Maintaining, extending, and scaling something to serve millions of users is an entirely different challenge. That's elite engineering work.
The startup founder takeaway: you don't need to hire a CTO yet. You need to build and validate your idea using vibe coding. Once you have traction and need to scale to millions of users, you'll need elite engineers. That's fine. Cross that bridge when you get there.
For now, move fast and focus on finding product-market fit. Hire engineers later.
The Real Competitive Advantage: Speed Combined with Taste
Let's cut to the core: the competitive advantage in the AI era isn't technical. It's not about picking the right tech stack or writing the cleanest code. Those things are commoditized now.
The competitive advantage is speed combined with good taste.
Speed without taste produces garbage quickly. Taste without speed produces nothing. But speed AND taste? That's unstoppable.
This is why Lazar focuses 100% of his optimization on clarity, judgment, quality, taste, good copywriting, and fonts. He deliberately doesn't optimize for the vibe coding workflow itself because he knows it will get automated away. But taste won't. Judgment won't. Understanding what users actually want won't.
For startup founders, this is the shift you need to make:
Stop obsessing over:
- Which framework to use
- Whether to build in React or something else
- Technical architecture decisions
- Tech stack optimization
Start obsessing over:
- What does your user actually want?
- How does it feel to use your product?
- What emotion does it evoke?
- Is the copy authentic and compelling?
- Do the colors and spacing create delight or confusion?
- What's the simplest way to solve the real problem?
The product that wins isn't the most technically sophisticated. It's the one that feels the best to use and solves a real problem that people are desperate to solve.
The Skills That Will Actually Matter Going Forward
As AI automates more technical work, what human skills remain?
Lazar's perspective is clear: anything requiring emotional intelligence, authentic human connection, or understanding of human nature.
He makes a bold prediction: AI will never write good comedy. Why? Because comedy requires understanding why something is funny, and that's fundamentally about human psychology and shared experience. AI lacks this. Good comedy is off the table for AI.
By extension, good copywriting, emotionally resonant design, and authentic storytelling are safe. These require understanding what makes humans tick. AI is getting better, but humans will always have an edge in areas requiring empathy and intuition.
His advice: identify the "comedy" in your industry. What's the task that requires unique human insight, creativity, and empathy? That's your mote of protection in the age of AI. That's what you should be building mastery in.
For founders specifically, this might be:
- Understanding your customer's deepest pains and desires
- Crafting messaging that connects emotionally
- Making design decisions that create delight
- Building community and authentic relationships
These things can't be outsourced to AI (yet, and maybe never). These are defensible.
From Hobby to Full-Time: The Path Forward
If you're a startup founder considering whether to dive into vibe coding, here's the reality: the barrier to entry is nonexistent.
You don't need to know how to code. You don't need a CS degree. You don't need to be hired by someone. You can start today.
The path is:
Pick a problem you're obsessed with solving. It doesn't have to be world-changing. It just has to be something you genuinely want to exist.
Build it using vibe coding. Go through the five parallel builds. Plan it out. Ship it.
Share the process publicly. Post about it on LinkedIn. Make videos. Show people how you built it.
Keep shipping. Build the next thing. And the next thing. And the next thing.
Look for opportunities to monetize or get hired. Some founders will eventually get hired by companies like Lovable. Others will start their own business using vibe coding as their competitive advantage. Both paths are viable.
The non-negotiable? You have to actually ship. Talking about vibe coding in a podcast is different from actually building something and putting it in front of users.
Lazar's final advice is worth emphasizing: Stop listening to this podcast. Stop reading articles about AI. Stop waiting. Go build something. Anything. The clarity will come through doing, not through consuming more information.
Conclusion: The Dream Job That Didn't Exist Two Years Ago
Vibe coding represents something historically unprecedented: a new career path emerging in real-time, and anyone can participate in it.
The opportunity isn't theoretical. It's real. There are real companies (including Lovable) hiring real people with the title "vibe coder" to do real work that moves the needle. S&P 500 companies are listing "Lovable skills" in job descriptions. Teams are migrating their entire legacy codebases to AI-powered platforms.
For startup founders, the implications are profound. You can now validate business ideas, find product-market fit, and build real businesses without hiring a technical co-founder or spending months fundraising to pay developers. You can do it yourself, with AI as your partner.
The shift from consumer to builder is happening at scale. The economic implications are massive. The competitive advantage goes to founders who:
- Move fast (ship weekly, not quarterly)
- Maintain taste (don't build garbage quickly)
- Understand their customers (clarity about what to build)
- Share and build in public (distribution and visibility)
This isn't hype. This is the actual state of software development in early 2025. The question isn't whether to get involved. The question is: what are you waiting for?
Start today. Pick an idea. Open Lovable or Cursor. Talk to the AI. Ship something by tomorrow. You're not lacking permission. You're not lacking tools. You're not lacking ability. You're just lacking the first step.
Take it. Build something. Share it. The world needs more builders. And it's never been easier to become one.
Original source: The rise of the professional vibe coder (a new AI-era job)
powered by osmu.app