Is Vibe Coding Bullshit?

An exploration of transformation, built in public

I've spent the past three years obsessing over AI and the future of work, and the number one lesson I’ve learned is that most people who talk about AI have no clue what’s actually happening in real-world AI implementation.

On one side are the AI optimists. These people who claim AGI is just around the corner and that all knowledge work will be extinct by the end of the decade. Futurist Ray Kurzweil has long predicted that human-level AGI will be achieved by 2029 and a full-blown “singularity” by 2045. Other extreme scenarios portray 2027–2030 as the window during which 90% of jobs will be automated away… a claim that many AI researchers dismiss as wildly speculative.

On the other side are the AI skeptics. These are the folks who dismiss the whole thing as “internet slop” or “stochastic parrots” remixing text without understanding. They point to hallucinations and shallow outputs as proof that AI is useless.

Both camps are wrong. The optimists overestimate the tech’s capabilities. The skeptics ignore where it’s already delivering results.

The truth is somewhere in the middle… AI excels at certain tasks and struggles with others. The result isn’t a replacement for human labor, but rather an augmentation of human labor that’s leading to a paradigm shift in how we work.

This is causing a mix of excitement and panic. For first-principles thinkers adept at thinking in systems, this is a dream come true. They have the opportunity to reinvent the concept of work from the ground up. To cargo culters and best practices merchants, this is a nightmare. The old playbook is dead, the new playbook hasn’t been written, and there’s no one else to tell them what to do.

This Transformation In Action

Over the past three years, I’ve put AI through its paces in my consulting work while interviewing every executive who would take a lunch with me to understand how they’re using AI in their business. I’ve seen what AI is capable of achieving for go-to-market teams, and, equally important, what it’s not yet capable of achieving.

There are some things, like writing copy or designing simple icons, that it excels at, while there are other, seemingly equally simple tasks, like designing a PowerPoint slide, that it’s terrible at with no signs of getting better. In understanding these strengths and limitations, the shape of my work has changed.

AI can’t do everything that an experienced marketer can do, but I can augment my work with an AI assistant, custom-tailored to the specific tasks that it excels at. Over time, as both I and my army of AI copilots get better at working together, our shared productivity skyrockets. Compounded across a team, this creates a cascading effect of increased productivity.

That said, the opposite can be equally true. By forcing AI into places where it doesn’t belong, doing jobs it’s ill-suited for, you can breed the exact opposite effect. It takes longer to get your AI assistant to do its job correctly than it would have taken to simply do it yourself, and your team copying and pasting the outputs of ChatGPT back and forth turns productivity theater into a snake eating its own tail. More activity is happening, and less is getting done.

What’s The New Shape of Work For Engineering Teams?

My general thesis right now is that every piece of your organization is slowly coming to this same conclusion… AI excels in some areas, struggles in others, and the real magic lies in how we integrate it into our workflows, seamlessly merging AI and human capabilities into one fluid motion.

Seeing this in the sales and marketing world has sparked my intense interest in a different world… How are engineering teams thinking about integrating AI into their work? There are some CEOs out there who genuinely believe that their engineering team can just “Vibe Code” their app now, so production times will skyrocket, right? After all, I downloaded Replit and made a crappy version of Wordle while I was on the toilet, so that must mean that my engineering team can 10x their production with the same technology, right?

This brings us back to the question we started with… Is “Vibe Coding” real, or is it bullshit?

I’m deeply curious about both the opportunities and limitations of Vibe Coding. Like many of you, I’ve tinkered with Replit, Cursor, and Claude Code, but I don’t do this for a living, so I don’t know what I’m doing. Just as someone with no marketing experience could have ChatGPT write a marketing plan that’d be passable in an undergraduate business class, I can Vibe Code my way to 'Hello, World,' but that experience doesn’t come close to answering my question.

• What are the use cases where Vibe Coding adds disproportionate value compared to more traditional methods?

• What are the limitations of vibe coding (and don’t just say “it can’t build production-ready apps.” Be more specific!)

• What changes do we need to make to our workflow to augment Vibe Coding’s strengths and weaknesses to get the most out of it?

I’m ready to get some answers.

18 Hours To Find An Answer

This September, the RALLY Innovation Conference is doing something terribly interesting… They’re gathering some of the smartest Vibe Coders they can find and giving them 18 hours to compete for a $10,000 prize in a first-of-its-kind Vibe Coding competition. This is the exact type of challenge that can help me explore this curiosity and figure out what’s real and what’s hallucinated in the realm of Vibe Coding.

The best part? I’ve been granted exclusive access to interview the participants leading up to the event.

Over the next few weeks, I will host a series of live streams featuring some of the smartest engineering minds in the Midwest to gain a deeper understanding of their perspectives on Vibe Coding.

• How are they using it today?

• Where is it useful? Where is it a distraction?

• What’s over-hyped and what’s under-hyped?

My goal is to return from this competition with a clear-eyed understanding of where Vibe Coding fits into the broader software engineering landscape and how teams can adapt to leverage this new tool without getting swept up in the hype cycle.

We’re sitting at a critical inflection point in the history of work, and I want to discover the edge of what’s possible. The next month is an exploration of the possible, and I’d love for you to join me.

I’m curious… What questions do you have about AI and Vibe Coding? I’d love to use these questions to get you some answers. Drop them in the comments below!