Each week we find a new topic for our readers to learn about in our AI Education column.
When you reach a certain level of wealth, do-it-yourself definitely starts losing its charm.
We’re sure it happens at a certain age, too, where you just look at a job and think that it’s best someone else spend their time doing it instead of yourself. There’s a quiet kind of arrogance in thinking that our time is too valuable to learn how to perform and complete a task.
In our experience, when we stopped trying to save money by avoiding hiring someone or something to do the work for us, we also ended up cutting off a valuable channel for education and experience. This was a hard life lesson, but we think we’ve learned it. The reason AI Education isn’t written by generative AI is that we believe there is educational value for ourselves and for our readers in writing it ourselves.
So whenever we can, we try to do it ourselves.
But DIY takes time, and time is money. Right?
Well, maybe not, thanks to a trend that we’re discussing in this week’s AI Education column called vibe coding. Vibe coding is closely related to another DIY software concept we’ve discussed here, low- and no-code development. Before we formally define vibe coding (you should already know that is has something to do with DIY), we’re going to tell you how we got here in hopes that it will clue you in a bit. We’ve encountered a few scraps of research and opinion that suggest financial institutions, wealth management advisors in particular, are starting to build their own fintech and wealthtech tools not by going through a vendor, but by through so-called “vibe coding.” With that in mind, when we learned what vibe coding actually entails, we thought it deserved its own write-up.
So, What Is Vibe Coding?
Recall that in low-code development, software can be developed largely through a what-you-see-is-what-you-get (WYSIWYG) interface where users drag and drop and edit visual elements with a mouse and keyboard, and then someone with a little coding knowledge comes in and makes final adjustments, if needed, to the outcome. Similarly, in no-code development, someone with no knowledge of programming languages can create software using a drag-and-drop interface, with no need to write any of the underlying code—the development space does it for you. Well, vibe coding similarly allows someone with little or no knowledge of computer programming languages to create their own software, using the power of generative AI.
Vibe coding is a loose, informal term for the process of prompting a generative AI with natural language to write code for us. Basically, we, the user, tell an AI chatbot through our own typed or spoken words what we want, and the generative AI creates it for us, writing the code automatically as it would if we were using a no-code platform, but instead of dragging and dropping elements on a computer screen, we’re just typing and talking. Most of us can already access vibe coding through a number of publicly facing generative AI models.
That’s right, folks, you can create your own software. Right now. Just by talking to a chatbot.
Anyone can do the talking, but, thus far, the people prompting computers to make software are mostly the same kind of people who have been learning how to code for years on end. They’re blending their expertise with the flexibility and efficiency of AI-enabled automation to help create a more productive software industry—albeit with a reduced headcount.
A Change in Development Philosophy
Coding is a lot like prose writing. There’s good, elegantly written clean code—efficient and well-documented and to the computer literate, nice to read—but there’s also all kinds of weird workarounds and markups that make for messy, inefficient, ugly code. Both good and messy code can get their jobs done for their users, but the former is usually faster and more reliable than the latter, so traditional programmers have usually put a lot of weight into the quality of the code that they were writing itself. They edit and revise their work, many attempt to optimize or perfect their code. Vibe coding, on the other hand, shifts emphasis away from the quality of the coding and the skill with which it is written, and puts it fully on the achievement of the outcome sought by the person prompting the AI.
The result is that we’re writing software a lot faster, because a developer can now tell AI the specific problem they’re trying to solve and receive a scrap of code that gives them an effective approach to solutions. Vibe coding has made it a lot cheaper and easier to test out ideas for software and computer programs. For small tasks, vibe coding might already be able to reduce intervention from IT departments to nothing more than checking and verifying that the new software is reliable and safe. Vibe coding could be connected to ambient AI to create point software solutions on the spot for financial professionals during conversations and meetings with clients or staff members. We could, theoretically, use vibe coding to create new programs, functionalities or elements of applications without even knowing we were doing so.
More importantly, however, is the move away from software development as a task undertaken while typing hundreds of characters per minute while hunched over a keyboard—instead, vibe coding enables developers to sit up, look at their monitors, and talk aloud about what they want, with the development software responding. We’ve made software development a multi-modal task, making developers’ lives and work considerably easier. Yet, ironically, many software developers hate the very idea of vibe coding, as it devalues the time they’ve spent memorizing various programming languages.
Again, value now resides less in the knowledge we learn and retain and more in the ideas we’re able to generate—and that sounds to us like a good thing.
Limitations of Vibe Coding
While traditional coders are usually at least somewhat security aware and concerned that the fruits of their labors may somehow be vulnerable to hackers and other cyber attacks, vibe coding is usually agnostic to cyber risk. It can also be difficult to maintain and update applications created using vibe coding, as it’s possible that no one really knows or understands what’s going on underneath the hood.
Our point in this section is that yes, vibe coding is great because to some extent it democratizes software development. It takes the power out of the hands of a specialized class of translators who know how to transform ideas into code, and puts it back into the hands of the people who actually have the creative ideas—a concept we discuss in brief as it relates to the arts at the top of this week’s AI & Finance column. But, vibe coding is not really a great solution for eliminating software development professionals altogether—yet. We shouldn’t just slop together some AI-generated code and launch an app without making efforts to verify and test that code, and doing so requires some knowledge. Yet, it sounds like some wealth managers might be doing exactly that. Yikes!
The technology we’re using to generate code from natural language prompts is still new and it is still improving. Right now, in spring 2026, most generative AI struggles to build complex applications.
