Catching The Right Vibes
It’s now possible to build entire apps, to exacting specifications, without even touching code, using artificial intelligence. But is vibe coding worth the risks?
For those not familiar with the term, “vibe coding” is a new AI-centric approach to app development that leans into large language models (LLMs) to, according to an AI-generated, self-referential description: “generate and refine code based upon natural language prompts.” The approach is both iterative and experimental, and vibe coders often describe the experience as one in which maintaining a creative “flow” during the coding process is desirable.
Instead of writing code directly, developers describe what they want the software to do, and the AI handles the implementation details. “This approach emphasizes a more conversational and improvisational style, similar to pair programming with an AI partner.”
These tools were originally intended as coding aids, but are effective enough that even programming Luddites can figure it out. Over the past six months, it’s taken the tech world by storm. The term was coined less than a year ago, by OpenAI cofounder Andrej Karpathy.
Why should companies care? Simply put, it’s an effective way to workshop new ideas, build custom tools, or speed up the development of websites. Vibe coding could be used in different industries to:
• Develop lead tracking tools for sales teams
• Automate repetitive tasks
• Create landing pages or micro sites for new products
• Develop internal tools to streamline workflows
• Rapid prototyping and idea validating
Startups can stress-test basic concepts before investing in a full product. Project managers can inexpensively develop internal apps to simplify time-consuming workflow problems. Marketing teams can make microsites and landing pages without a ton of hassle—or added cost.
Panel at Digital Builders Event
And it’s so easy, even a CEO can do it.
“The time, from iterations to prototype, to get to an MVP (minimum viable product) launch, and to test with users? That cycle has been shortened dramatically now,” says Emanuel Perez, the founder of the creative lab Explay, and a key part of a rising community of local vibe coding enthusiasts.
The enthusiasm around LLMs, despite their clear flaws, is real—as highlighted by the unprecedented emergence of tools like Lovable, a months-old company that recently topped $100 million in annual recurring revenue. Locally, communities like the Hampton Roads chapter of the AI Collective, where Perez recently spoke, seem to suggest that the excitement goes beyond tweets.
And while old-school coders may be skeptical, it’s adoption is widespread. A GitHub study found that 97% of professional programmers have tried it. Emerging software engineers, like Lionel Sapp, have no preconceptions about Claude and OpenAI.
“I was just the perfect time to where I didn’t have any loyalty to writing code myself,” the self-taught programmer says in an interview. “So I came in as an ally with AI.”
Culture & Code
Sapp, like Perez, is no mere LLM user. He heavily promotes the vibe coding concept through Digital Builders 757, a Norfolk-based education and events community. The company’s events, frequently at the Norfolk coworking space Assembly, emphasize culture just as much as code, integrating business, art, music, even fashion around a climate of hackathons—something on display at the firm’s June Ctrl Alt Play event.
“When I came into the tech world, I saw that the events were just lacking a lot of the creativity that I was used to from the performing arts industry,” the former videographer says.
This South-by-Southwest-like fusion of disciplines, Sapp says, points to how vibe coding enables creatives without programming backgrounds to test the waters. “If someone is not technical, it 1,000% lowers the barrier of entry,” Sapp adds. “And the only thing that you need is just a will to create.”
Vibe Coding: The Good & The Bad
Vibe coding’s lower barrier to entry can be felt in numerous ways, including in the creation of personal productivity tools. Perez, for example, says he built a custom app to manage his ADHD. “It’s something silly and simple that I made, but it works for me,” he says.
It’s flexible, too. Perez points to the fact that, beyond typing, it is possible to use voice recognition and even drawing to interact with many vibe-coding apps. “It’s like using the internet for the first time, but on this next new level,” he says.
But it’s not all perfect. Complex tasks can often confuse LLMs, a concept known as “hallucination.” Plus, there is a real risk that an amateur vibe-coded project can have a glaring security hole. Researchers at Virginia Tech contributed to a recent academic study that warned that AI tools often made up open-source package names, creating a malware risk known as “slopsquatting.”
And then there’s the cost. While free tools exist, many AI platforms charge by the number of “tokens” you use. With added complexity, the costs can add up fast.
Finally, there are the occasional horror stories, such as that of tech founder Jason Lemkin, who fell hard for an AI coding tool called Replit, only for the tool to delete his company’s database.
“There is no way to enforce a code freeze in vibe coding apps like Replit. There just isn’t,” Lemkin wrote.
While that’s an extreme case, Sapp emphasizes that this is an important sign that guardrails are necessary—including frequent backups and the use of sandboxes for your projects.
“Treat AI like a junior developer with superpowers,” he says. “Give it specific tasks, but always review its work, and don’t hand over admin rights unless you’re truly ready for anything.”
For many people, “a junior developer with superpowers” is still a huge upgrade from their nonexistent programming skills. And in a business context, that can be a huge shot in the arm in cases where custom tooling wasn’t quite so cost-effective before.
“I think the magic in vibe coding is: If you have a process or a task that takes a numerous amount of steps, you can cut those steps down considerably,” Perez says.
Given the right context, it could potentially clear up the nagging annoyances that often slow businesses down…just make sure you test your code first.
Vibe Coding Tools Worth Trying
v0
v0.dev
Both Sapp and Perez recommended this Vercel-developed tool, which shines with front-end development, making it easy to build designs into working demos.
Lovable
lovable.dev
This general-purpose LLM offering makes it possible to program internal tools or other personal projects without much fuss.
Claude Code
anthropic.com/claude-code
Willing to dive into a terminal? Claude Code makes it easy to develop a project via direct access to your code base. The regular Claude (https://claude.ai) is no slouch, either.
