There is a huge debate going on right now about vibe coding and fintech, and I have a dog in this fight. So let me tell you where I stand, how I got there, and why I think most of the people shouting "no" are arguing about the wrong thing.
I failed at this three times
Truth be told, my first attempt was a disaster. Sometime in the summer of 2025, maybe September, I sat down to vibe code for the first time. I am technically inclined, so I assumed this would be easy. It was not. I spent far more time fixing errors than building anything. And, to be fair to the tools, I was doing it wrong. I gave up.
I tried again in October. Gave up again.
Then, because everything I kept reading said the models were improving every few weeks, I tried a third time in December. The results were phenomenal. Since January 2026 it has only gotten better. Ideas that were sitting in the archive at the back of my head, the ones I would visit every now and then and sigh at, can now be deployed literally overnight.
That is the part the naysayers keep missing.
What the naysayers say, and why they are half right
Ask anyone in the commercial software industry whether vibe coding can produce a financial services website and you will get a fast answer: "No. It's not secure. Look what happened to this company. Look what happened to that one. This site got hacked. That one leaked data."
All true. I am not denying any of it. If you take an AI-generated codebase, point it at the public internet with real customer money behind it, and never let anyone competent look at it, you deserve what comes next.
But notice what that argument is actually about. It is about carelessness. It is not about vibe coding.
What vibe coding actually gives you
Allow me to digress for a moment. Think about an architect. Before anyone pours concrete, the architect builds a cardboard model. Nobody lives in the cardboard model. Nobody worries about whether the cardboard model is earthquake-proof. The model exists so the client can walk around it, point at things, and say "this room is too small" before it costs a million dollars to fix.
That is what vibe coding is for me.
I have an idea. I build it. Now I have a working prototype, a real minimum viable product, and I can put it in front of the people who matter: investors, partners, potential customers, potential suppliers, the operators who would actually run this thing. They can click on it. They can break it. They can tell me what is wrong with it.
And more importantly, I get to pause. I get to look at the thing I imagined, now sitting in front of me, and say: "Ah. Okay. This is what I could have done better. Maybe this flow should go there. Maybe this screen does not need to exist at all."
Just one year ago, none of this was possible. Not for a solo operator. Not at this speed. Not at this cost.
So what about security?
Here is where I will take a position most consultants would hedge on.
Yes, a vibe-coded fintech site can go commercial. With a condition. You lock it down to a reasonable extent, you do not exhibit carelessness, and you accept that you are operating within the known risks of this approach and doing a good job inside those limits.
There are now commercial tools and companies whose entire business is reviewing and hardening AI-generated code. There are better models every month. And I am a firm believer that very soon security will stop being a separate step you remember to do at the end. It will be part and parcel of the build process itself, baked into the review loop, the same way spell-check stopped being a separate program you ran on your document.
Careless is the problem. Vibe coding is not the problem.
What I have actually shipped
I have not yet put my commercial fintech products, the ones with licensing and money movement behind them, through this process. Those stay on the cautious path for now. But I have launched three things, and you can go look at all of them.
DealHarbor. As far as I am aware, this is the world's first marketplace for regulated financial services deal flow. I say "as far as I am aware" deliberately. There are Telegram groups, yes, but those are all one-sided. Everyone is advertising something they sell, or represent, or have an interest in. Nobody is matching buyers to sellers in the open. DealHarbor does. Go check it out.
AskMint.app. This is my Quora equivalent. I have written more than 5,000 answers on Quora over the years, but not all of them belong online. Some were not about finance, fintech, banking, or payments, so they stay off. Right now AskMint holds about 2,800 of my answers, and I keep adding. By the end of this year I expect it to sit around 3,100 to 3,200. I plan to open it up to the wider public sometime next year.
Money Wiki. A Wikipedia for money, banking, and payments. Simple as that.
Behind these, I have more in the pipeline: an exchange rate tool, a smart contracts project, a directory of companies in financial services, a regulatory alert mechanism, a user aggregation play for the industry, and a few others. I was working on all of these. Then I was overwhelmed by the sheer amount of work, so I slowed down.
What changed is that I have now figured out how to use AI agents properly. I am running Grok's agent and ChatGPT's agent, and testing a few more, and they are starting to handle entire workstreams end to end. The overwhelm is going away.
If you have an idea
Here is my advice, and I mean it.
Do not let the naysayers scare you. You have to understand who is doing the shouting. Someone who is going into obsolescence will always try to frighten you out of the thing that is making them obsolete. At some point in history there was a man who made the best buggy whips in the world. Then the internal combustion engine arrived, the horse was under threat, and of course he had negative things to say about automobiles. Of course he did.
Do not be disheartened by the buggy whip maker.
If you have an idea, vibe code it. Bring it to life. Put it in front of real people. You are going to love that journey, and I say that as someone who gave up on it twice.
And if you have already vibe coded something, get in touch with me. I would love to see it and give you some feedback.
Interesting times we live in.
Quick answers
Is vibe coding safe for fintech?
Safe enough for prototypes, MVPs, and investor demos, yes. For live money movement, only after a proper security review and hardening by someone who knows what they are doing. The risk is carelessness, not the tool.
Can I show a vibe-coded MVP to investors?
Yes, and you should. A working prototype beats a slide deck every time. Just be clear with them about what it is: a model, not the finished building.
How long before vibe coding handles security on its own?
IMHO, not long. The tools are already moving review and hardening into the build loop. My guess is that within a year or two it will be unusual to ship without it.