The Rise of the Professional Vibe Coder: Lazar Jovanovic on Building Without Writing Code

A forestry engineer who waited tables and worked blue-collar jobs is now Lovable’s first official “vibe coding engineer,” shipping production products without touching code. The job title sounds like a meme, but Lazar Jovanovic’s workflow reveals something deeper about where software development is heading: the bottleneck has shifted from execution to clarity.

The Episode

Lenny Rachitsky interviews Lazar Jovanovic, the first professional vibe coder hired at Lovable, the AI-powered app builder. Lazar has no traditional coding background. He’s a forestry engineer by education who spent years in startups doing community management, social media, and product work. He started vibe coding in July 2024, months before Andrej Karpathy coined the term, and built enough in public to catch the attention of Lovable’s CEO Elena. The conversation covers his tactical workflows, mental models, and a surprisingly coherent philosophy about what skills matter when AI handles the building.

Clarity Is the New Skill Stack

Lazar’s core argument is that AI tools are amplifiers, not replacements for thinking. If you don’t know what you want, you’ll produce garbage faster. His entire workflow optimizes for one thing: clarity of input.

“I’m optimizing 100% of my time today on good judgment, clarity, quality, taste.”

He uses the Aladdin and the genie analogy: you get three wishes. Say “I want to be taller” and the genie makes you 13 feet tall. You weren’t specific enough. AI tools work the same way. The quality of output is bounded by the quality of the ask.

This isn’t a platitude. Lazar estimates he spends 80% of his time in chat mode, planning and learning, and only 20% actually building. Most people invert this ratio, jumping straight to prompting. His key insight: spending time talking to the AI in planning mode develops your judgment faster than prompting it to build things.

The PRD System: How to Actually Ship

Lazar’s tactical workflow centers on a system of Markdown files that serve as persistent context for AI agents:

1. PRD file: A structured product requirements document that acts as the “brain” of the project. It contains the project vision, feature specs, user flows, and current status. The AI reads this before every session.

2. Iteration log: A running record of what’s been built, what failed, and what changed. This prevents the AI from going in circles or losing context across sessions.

3. Design specs: Visual references, component libraries, and style guidelines that keep the AI aligned on aesthetics.

The key mechanic: when starting a new chat session, Lazar pastes the PRD as context. The AI then knows the full project state and can pick up where it left off. This solves the biggest pain point of vibe coding: context loss between sessions.

The 4x4 Debugging Workflow

When stuck on a bug or design issue, Lazar runs a systematic debugging process:

1. Explain the problem to the AI in detail, including what you expected vs. what happened
2. Ask the AI to reason about root causes before suggesting fixes
3. Try the fix in isolation rather than stacking changes
4. If four attempts fail, go nuclear: revert to a working state and try a completely different approach

The “four attempts” threshold is deliberate. Beyond that point, the AI is likely going in circles, and more prompting won’t help. Better to restart from a clean state than to accumulate patches on patches.

Parallel Prototyping: Five Genies, Not One

One of Lazar’s most counterintuitive practices is kicking off four or five parallel prototypes for any significant feature. Instead of iterating on one approach, he tells multiple AI sessions to build the same thing with different constraints or styles, then picks the best result.

“It’s like having five genies instead of one. You’re not married to any single approach.”

This works because AI generation is essentially free in terms of time and cost. The bottleneck isn’t building; it’s evaluating. Lazar’s judgment about what constitutes “good” is the scarce resource, and parallel prototyping gives him more options to exercise that judgment on.

The Convergence of PM, Design, and Engineering

Lazar sees the traditional Venn diagram of product manager, designer, and engineer collapsing into a single role. These used to be separate disciplines with separate skill sets. Now, with AI handling the implementation, the differentiator is the person who can think across all three domains.

“These Venn diagrams of engineer, designer, PM used to be very separate. Now they’re converging.”

This doesn’t mean everyone becomes a generalist. It means the most valuable person is the one who can make good product decisions (PM instinct), express them visually (design taste), and communicate them clearly to AI (engineering interface). The boundaries dissolve because the execution layer is abstracted away.

What AI Won’t Replace

Lazar makes bold predictions about which skills are AI-proof:

Will be replaced: Translators, average writers, middle managers who primarily relay information, any role that’s essentially deterministic (clear X input produces clear Y output).

Won’t be replaced: Comedians, elite writers, anyone whose work requires understanding human emotional dynamics. His test: if the role requires understanding what’s funny, what’s moving, or what creates genuine human connection, AI can’t do it.

“AI is never going to be able to write a good joke. Never, never, never.”

He acknowledges he’ll be wrong on 95% of his predictions within three months, but holds firm on this one. The argument isn’t about AI capability but about the irreducibly human nature of humor and emotional resonance.

Design Taste as Competitive Moat

In a world where anyone can build “good enough” with AI, the differentiator becomes taste. Lazar is obsessed with design quality, following world-class designers, watching 40-50 minute design process videos, and studying how elite creators think about visual decisions.

His argument: tech stack doesn’t matter anymore. Whether something is built in React or HTML is irrelevant. The end user just wants a stellar experience. When building is instant, the only thing left to compete on is the quality of what you build. And quality is a function of taste, which is a function of exposure and deliberate study.

“Forget about decisions on tech stack. Forget about which backend, which frontend. That doesn’t matter. Quality, taste, design. That’s all you need to optimize for.”

How to Become a Professional Vibe Coder

Lazar’s path to this role was non-linear: forestry engineering degree, waiting tables at Subway, blue-collar work, then seven years in startups doing non-technical roles. The job at Lovable came from one thing: building in public.

His advice for others:

• Build in public: Share everything you create, including failures. Lazar’s YouTube channel and LinkedIn presence caught Lovable’s attention.
• Don’t wait for permission: “You don’t need a company to hire you. You can hire yourself as a professional vibe coder first.” He was already doing the job before getting hired.
• Send apps, not resumes: Multiple Lovable hires got noticed by sending Lovable apps instead of traditional applications. Show the work, don’t describe it.
• Exposure time over building time: Deliberately allocate more time to learning (watching others design, reading AI agent output, studying what’s possible) than to building.

The Horses Analogy

Lazar references a viral observation: when steam engines led to cars, 90% of the horse population in the US was eradicated within 20 years. But horses had 20 years. The person who tweeted this, who works at Claude Code, was hired for a technical writing job and became obsolete six months later.

The implication isn’t doom. Lazar’s point is that the speed of change demands continuous evolution. The people who thrive aren’t the ones with the most skills but the ones who can redefine their role fastest.

“You should only be afraid if you’re doing nothing. If you’re doing absolutely nothing, yes, be terrified. And then take a step towards doing something about it.”

Some Thoughts

This conversation works best as a case study in career reinvention rather than a technical guide. Lazar’s specific tool workflows will be obsolete in months, as he himself acknowledges. But the meta-pattern holds:

• The shift from execution to judgment is real and accelerating. Lazar’s 80/20 planning-to-building ratio is the clearest articulation of this I’ve heard.
• “Build in public” as a career strategy has a new dimension when the building itself is AI-powered. The differentiator isn’t what you built but how you thought about building it.
• The convergence of PM/design/engineering isn’t a future prediction; it’s already the job description of a vibe coder. Companies are hiring for this role before they have a name for it.
• His insistence that taste and design are the final moat deserves serious consideration. When code is free, aesthetics become the scarce resource.