Vibe Coding and the Myth of the Disappearing Developer
I spend a good portion of my time developing product strategy and leading a tech team through implementations. We’ve deployed AI tools for customers to search our M&A research database. Internally, we’ve built AI workflows that drive real production efficiency. I use AI personally as well, and I’ve spent considerable time “vibe coding.”
As a result, I’ve been thinking more carefully about what AI actually means for the people who build things for a living, how AI is being used, where the real gains are, and where it still falls short.
Recent Progress
Since Anthropic released its latest Claude model earlier this year (2026), vibe coding has become all the rage. Vibe coding allows non-developers to describe what they want in plain English, and the AI model produces working code. No computer science degree required… it would seem.
It’s genuinely impressive. In fact, I built an entire productivity app for myself using vibe coding alone (more on that in a later post). Because of the ease to create with vibe coding, a curious person can now get done in an afternoon what would have seemed like science fiction just a few years ago.
But here’s an important observation: the gap between “it works on my screen” and “it works in production” is no small gap. It is the whole game.
Excellent Prototypes
At its best, vibe coding produces a working prototype at extraordinary speed.
Historically, to communicate a new project to developers, I would outline the strategy (what it is meant to accomplish), write detailed specifications, define data structures, and sketch the interface to illustrate how I envisioned it. It was time-consuming… and even then, gaps remained between what I wanted and what the developer might have assumed I wanted because I didn’t explicitly describe certain components. Fortunately for me, I have an excellent team who fills the gaps with something superior to what I would have asked for. But not everyone has a rock star developer like this.
By contrast, with vibe coding, the process compresses dramatically. I can describe what I want, and the AI model will produce an excellent, functional version within minutes. After just a few hours of iteration, I don’t just have a sketch, I have a working application running locally.
That changes the conversation entirely. Plus, it’s much easier to have something on the screen that actually works to show the developer what I want, to clarify behavior, and to surface questions about edge cases for further discussion.
If a picture is worth a thousand words, an AI vibe-coded prototype is worth about 10,000 lines of code.
To be clear, vibe-coded prototypes actually work… assuming the user behaves exactly as expected, the input data is clean, and nothing unusual happens to break it. In other words, vibe-coded apps are genuinely useful. They work… as prototypes.
The Problem
Too many people think they’ve vibe-coded a production-ready application when they’ve actually only built a lightweight prototype.
What vibe coding does not produce, without experienced human developer oversight, is software you can stake your business on.
Production-ready code requires someone to think carefully about the things that can go wrong:
- What happens when authentication fails?
- Where does the data live, and who has access to it?
- Are there safeguards for bad or unexpected data?
- Is sensitive information properly secured and encrypted?
- What’s the backup and recovery plan at 2 a.m.?
- Is it secure, or does it just appear secure?
These aren’t glamorous questions. They don’t make for good demos. But they are the difference between a useful tool and a liability.
A Parallel Worth Considering
When web hosting became inexpensive and drag-and-drop builders became widely available, many predicted web developers would be rendered obsolete. Why hire someone when anyone could build a website?
What actually happened was the opposite. The market for web developers expanded. Suddenly, every business needed a web presence… small businesses, nonprofits, solo practitioners. The more accessible the tools became, the more people discovered how much they didn’t know. The result: demand for professional developers increased.
We’re seeing a similar pattern now, just with different technology, more powerful tools, and higher stakes.
More Projects, Not Fewer Developers
Here’s the hypothesis:
AI will not reduce the demand for skilled developers. It will increase it.
The floor for building something has dropped dramatically. Projects that were previously shelved because they were too expensive, too complex, or provided too marginal a benefit are now viable. Ideas that once lived on whiteboards are becoming prototypes. Prototypes that once died in proof-of-concept are getting funded.
But every one of those projects, at some point, needs someone who understands how to make it robust. Secure. Maintainable. Something you can confidently put in front of customers.
The developers who recognize this moment for what it is, a leverage opportunity, not an existential threat, will be extraordinarily valuable. They can move faster on early-stage work and focus their effort where it matters most: robustness, scalability, and human judgment that comes from hard-won experience.
The nature of the work will shift. The skill profile will evolve.
The best developers of the next decade won’t be the ones who resist AI tools. They’ll be the ones who master them, vibe coding as a starting point for rapid iteration, and then use their expertise as the quality filter to turn prototypes into real systems.
My point is, AI will enable an explosion of software, applications, and tools that drives an unprecedented step-change in productivity… but it will require more developers, not less.
AI amplifies capability. It doesn’t (yet) replace judgment.
And the rest of us? We’ll be able to build more than we ever imagined. We’ll just need to know when to hand the keys to someone who can make sure our ideas work and scale to production-level requirements.
Related posts:
Follow Past Midway if you would like an email notification when I post something new.
You said it was not my cup of tea. In one respect, you are correct. I know absolutely nothing about, well, about anything involving computers except at the kindergarten level. But, I found this fascinating because it refutes a lot of the rhetoric about AI from the general uninformed public (which includes people with my computer skill level). I also find it wonderfully interesting that my son continues in his quest (very successfully) of learning, knowing, & doing more…& stays excited about it. Good job.