Vibecoding for PMs

How I Learned to Stop Stressing About "How" and Trust the Vibes

You know how tech wasn't already filled with enough meaningless buzzwords? Well, good news! We've got another one: vibecoding. Except this one might actually be the most honest and timely term ever invented. It's exactly what it sounds like – vibing while AI does the coding. It turns out that understanding how your code works was getting in the way of productivity for most of us.

TL;DR

Vibecoding means surrendering to the vibes and letting AI-powered programming tools create, repair, and ship. Despite the troubling/perfect name (thanks, Andrej Karpathy), it's actually transforming how PMs like me work by letting us build things without needing engineering. It's ridiculous, it's awesome, and it's changing who gets to build software. Its changing the speed at which an idea becomes real.

---

What the Hell is Vibecoding Anyway?

Vibecoding is what happens when you combine AI coding tools with the human desire to just get shit done without learning another JavaScript framework (or much really). Coined by Andrej Karpathy (OpenAI co-founder and former Tesla AI leader) in February 2025, it's the practice of describing what you want in plain language and letting AI generate the code while you blissfully ignore the implementation details.

Traditional coding requires developers to understand what they're writing. They understand what they want and approach it in a structured way. Vibecoding asks: "But what if we just... didn't?" What if we just say what we want? Turns out, it works for a lot of use-cases. I wouldnt recommend building a scalable web application from scratch this way, but I would recommend just about everything short of that.

Imagine if your most technically challenged relative suddenly could build a website by just describing it – that's basically vibecoding. There are so many memes to choose here, honestly im struggling to decide.

ok i picked a few

Interestingly, the "vibe" concept in AI isn't entirely new. Before the recent explosion of LLM judge and verification tools, people relied on intuitive "vibe checks" to evaluate AI outputs. Researchers even created systems like "VibeCheck" to study how different LLMs have distinctive characteristics or "vibes" in their outputs – like tone, formatting, and writing style – that users can intuitively recognize but struggle to quantify. Turns out, you can actually predict which AI wrote something with surprising accuracy just by getting a feel for its "vibe."

So in a weird way, "vibing" with AI has been part of the process since the beginning. Now we're just taking it to the next level by fully surrendering to the vibe.

Why the Name is Both Ridiculous and Awesome

Let's be real: the first time you hear "vibecoding," your brain probably does that thing where it tries to decide if this is a term worth remembering or if it should file it away with "NFT utility" and "metaverse strategy" in the cabinet labeled "New fangled words that i dont care for."

The term sounds like what happens when engineering meets astrology. "Mercury is in retrograde, so all my pull requests are getting rejected."

But here's the weird part - it's actually the perfect name. Because what else would you call it when you're just... feeling your way through code generation by chatting with an AI? I'm literally asking an AI to build me a feature, and when it doesn't work, I'm like "hmm, let's try a different way of explaining it" instead of, you know, actually learning how to fix the code myself. If that's not "vibing" with technology, I don't know what is.

Of course, we can't ignore that the term "vibe" itself comes from Gen Z slang, where it evolved from simply meaning "a feeling” a much broader meaning . As a millennial writing this, I feel ancient just trying to explain it – like I'm some digital anthropologist documenting the linguistic habits of a younger, cooler civilization (except im only digging through the past few years…).

The fact that this slang has now infiltrated technical discussions about AI and software development just proves we've come full circle – the children are now teaching the adults, and the adults are pretending we've been saying "vibe" all along. At least Karpathy had the courage to embrace it fully. I think its also a great indicator of how much this technology is being adopted and leveraged by the younger end of the workforce (a different post to be sure).

My Experience: How Vibecoding Changed My PM Life

As a Product Manager, vibecoding has fundamentally altered my approach. Here's how:

1. Brainstorming and Concept synthesis

Before vibecoding: I was very thoughtful in how i iterated on ideas, and also when to bring in designers and engineers (I follow the golden rule: Treat others time as more valuable than your own). Now, i can get to something that can illicit a strong reaction in no time at all. It cuts the corner to more powerful feedback.

I have a tendency to think out loud, and use complicated language, in an environment with multiple languages, in particular, this can be an issue. I’ve always worked to manage this and ensure i simplify my thoughts in advance of meeting with people. However, now i can vibecode my way to something more visceral - it is an absolute game changer for me as a PM.

I now spend my time sharpening my vague feature ideas to AI tools and making something that evokes reactions. There is art here, there always was, but if a picture is worth a 1000 words, at modern conversion rates, a working prototype has to be at least 10,000 words. Thats more.

A Developer Conversation Pre-Vibecoding:

Me: "So based on customer interviews, I think we need some kind of magical widget that runs on dreams and outputs rainbows."
Dev: "Okay, let me unpack that... so the user gives us their dreams, or do I need to get those from somewhere? you know what, no. just no."
Me: "But the customer really needs this to solve their problem."
Dev: "I'm sure they do. Lets find a whiteboard so you can prove to me you actually understand what you’re asking.
Me: silently composing slack update about 'technical blockers'

A Developer Conversation Post-Vibecoding:

Me: "Look what I built this morning!" shows functioning prototype
Dev: "You... built this? It actually does what you could barely describe?"
Me: "I’m not crazy you know…"
Dev: clicks around "This is actually interesting. It won't scale and the security is questionable, but I see what you're trying to do now. What if we built something more focused that solves the same problem but would actually work in production?"
Me: "Great. So… when are we talking? Tomorrow?"

2. Fast Iteration With Customers - Move With Urgency

Nothing has been more empowering than live iterating on a design with a customer and seeing AI tools help them better describe what they want. I can make adjustments on the fly as they explain their needs, and suddenly we're co-creating rather than me trying to interpret vague feedback days or weeks later. These delays can compound given schedules and on-going work. Showing instant progress is a powerful trust building tool as well as a clarifying and aligning method.

Product management is all about idea iteration - and vibecoding lets you move with urgency in the presence of customers. It turns abstract conversation into tangible changes they can react to in real-time. The feedback loop that used to take weeks now happens in minutes.

3. From Data to Insights - Without the Product Ops Overhead

Vibecoding isn't just changing how we build products - it's revolutionizing how we understand customer needs in the first place.

Product Managers never have enough time. Often really high value, but tedious activities get cut. I used to spend days pouring through customer feedback spreadsheets and trying to normalize it for distribution. Now I just dump hundreds of sales call transcripts, support tickets, and survey responses into an AI and ask for patterns, summaries, and even specific feature based questions. Newer reasoning models can cite sources and point directly at sharp insights that were otherwise hidden.

The entire product operations function is being transformed. Instead of maintaining complex systems to surface insights, we're creating flexible AI agents that continuously monitor customer signals and proactively identify opportunities. Being able to vibecode your way into these automations is an essential skill of a product manager now.

The best part? These insights aren't just faster - they're often better because we're processing ALL the data, not just the subset we had time to review manually. Those subtle patterns and edge cases that used to get missed are now being surfaced automatically.

You can even have it actively reach out to Account Management and Sales in cases where a customer has beta availability or some other interesting use case in near-real time.

The Vibecoding Spectrum: From Side Projects to Enterprise

There's an important nuance to vibecoding that deserves attention: its impact varies dramatically depending on where you sit in the product development ecosystem.

For bootstrapped startups, side projects, and indie hackers, vibecoding is nothing short of revolutionary. It's a 100% unlock. Y Combinator recently reported that a quarter of startups in their current cohort have codebases that are almost entirely AI-generated. When you're resource-constrained and time is your most precious commodity, being able to transform ideas into working code without hiring a full engineering team is game-changing.

For mid-tier companies, it's an acceleration tool. Product managers can prototype and validate faster, but they'll still need engineers to stabilize, secure, and scale solutions. The production code might start with a vibecoded foundation, but experienced developers are essential for hardening it for real-world use.

At enterprise scale, vibecoding is more of an ideation and communication tool than a production solution. The complexity of large-scale systems, strict security requirements, and massive customer bases means that while you might vibecode to demonstrate concepts, the journey to production-ready code is still heavily dependent on expert engineering.

So apply this with thoughtfulness - but do apply it.

End with Something: Embrace the Vibe

If you're a PM, designer, or non-technical founder who's been held back by your lack of coding skills, I challenge you to dip your toes into the vibecoding waters. Yes, you might feel slightly ridiculous saying, "I built this by vibing with AI" (and i strongly advise not saying anything like that, make it sound cooler at least). But I promise the skills you develop when working with tools like Cursor, v0, replit, lovable, bolt, etc. are worth cultivating. They are the tool set of the modern product person. Increasingly the lines between product, dev, design, biz dev, ops is blurring - and your breadth of knowledge makes you faster and more efficient. Ill follow up with a deeper dive example on these, but for now, start playing.

While the hardcore engineers might scoff, they'll eventually have to reckon with the fact that vibecoding is making their specialized knowledge increasingly optional for many projects at the small scale. People with that expertise remain essential thought partners and the only ones who can build at scale, balancing the myriad other considerations (security, cost, simplify, scale). But getting started is the fastest way to inform those considerations. As a product person, at the end of the day, what matters is whether you can bring your ideas to life, not whether you understand every line of code that makes them work.

---