Last weekend I built an application in less than 20 hours, and I am still a little surprised by how it happened.
I started by asking ChatGPT to create a mockup of an idea I had. I took that mockup into Codex and told it to build an application around it. What came back was not just a set of screens. It built a Next.js front end with a Python Django back end, created the API endpoints connecting the two, and put in the security needed to control how data moved through the application. It also added error handling, validation, informational cards, toast notifications and other details that made it feel much more complete than I expected from such a simple starting point.
It was not perfect, which actually made the experience more interesting. At one point it put about 20 data fields on the wrong screen. I told it where I wanted them instead, and about a minute later they had been moved and were working where they belonged. Not very long ago, that kind of change could have meant touching multiple files, checking data connections, testing the interface and then discovering something completely unrelated had broken along the way.
I have used ChatGPT, and Claude for years, so I have watched these tools improve gradually. This felt different because I was no longer impressed simply by how quickly it could write code. I was impressed by how much of the application it seemed to understand without me explaining every little detail.
That is what really got me thinking about what happens next. If one person can go from an idea to a working application over a weekend, a lot more people are going to start building software. Someone who knows a particular industry well may no longer need a development team and months of work just to find out whether an idea has merit. A teacher, restaurant owner, consultant, coach or small business owner may be able to turn years of experience into a usable product in a few days.
Of course, this also means we are probably about to be flooded with apps. Some will be genuinely useful and some will be clever ideas that probably should have remained ideas. When the barrier to building something becomes this low, people are naturally going to build a lot more things simply because they can.
The part I think many people will underestimate is what happens after the app is built. Development may become inexpensive, but running software still costs money. Amazon wants to be paid. Azure wants to be paid. OpenAI wants to be paid. Databases, storage, email providers, APIs and monitoring services all eventually send a bill, and they tend to be remarkably consistent about it.
That is where the entrepreneurial side becomes much more important. It is not enough to build something people think is cool. You have to understand who needs it, why they will keep using it and, eventually, who is going to pay for it. Maybe the answer is subscriptions, advertising, transactions, sponsorships, licensing, business customers or API access. The model can be different for every product, but sooner or later the math has to work.
What I am going to find fascinating over the next few years is seeing what people actually decide is worth paying for. We already pay monthly for Netflix, Spotify, cloud storage, productivity software and plenty of other services, but there is a practical limit to how many $5, $10 and $20 subscriptions people are going to tolerate.
That makes the next phase of the app market really interesting. Some ideas that look brilliant may fail because people like them but do not value them enough to pay for them. Meanwhile, something almost embarrassingly simple could become a huge business because it solves one annoying problem so well that people happily keep pulling out their credit card.
I think AI is going to create an extraordinary amount of experimentation because the cost of turning an idea into something real is falling so quickly. What it cannot do is guarantee that anyone wants the product, understands it, trusts it or considers it valuable enough to pay for.
That is what I am most interested in watching. We are getting very close to a point where building the application may be the easy part. Figuring out what deserves to be built, who really wants it and whether it can survive as a business may become the much harder part.
And that is probably exactly how it should be. The technology can build faster than ever, but the market still gets to decide what matters.
That is also why I am having so much fun working on Hey Benson right now. Some of the concepts we are building are becoming genuinely interesting, not because AI can build them, but because I think people may actually find them useful enough to pay for. I will have more to share on that soon.