There are four options now, not three: hire someone, learn to code, use a no-code platform, or let AI write it. My recommendation, plainly: use the cheap ones to make prototypes and to learn, and do not use them to build the product you intend to sell.

I am not saying that to protect anyone's job. I think it is amazing that vibe coding exists and that everybody can be a builder now. You just need to know what each choice actually costs you.

Hiring someone

Good when you know what you want built, you can pay for it, and you want to spend your own time on the product rather than the code.

The catch is that a developer can only build what you can describe. Bring an idea and no research and you are paying them to make your product decisions, which they will do sensibly and without ever having met your user.

What makes it work: a written scope, and asking them to size each item as small, medium, large or extra large rather than demanding dates.

Learning to code

Good when you want this skill for its own sake, or you plan to keep building things for years.

The catch is the timescale. This is not a detour of a few weeks, and the thing that decides whether your app works is not the code. I learned to code because of an app idea, and the app idea waited years while I became an engineer.

Learning enough to be dangerous, so you can read what you are given and ask decent questions, is a different and much smaller project. That one is usually worth it.

No-code platforms

Good for prototypes, internal tools and testing an idea quickly.

The catch is the ceiling, and people meet it late. They start building, get excited about how much they can do, and as they get close to launch they discover they cannot launch with it. Then it is a rebuild from scratch, which takes longer than the first build did. If you take a Bubble or FlutterFlow app to an agency and ask them to finish it, they will tell you it is a great start and that they have to start over.

I saw a company come into the agency I worked at with government grant money, having spent around $500,000 on a Bubble prototype that barely worked. When we explained the rebuild cost, they said: yes, but I did that in Bubble in five minutes.

AI writing it for you

This is what most people mean now, and it is worth separating two things. An engineer using AI to write code is one thing. Letting AI decide and write the whole app, which is what Replit, Lovable and prompting Cursor or ChatGPT to build your app amount to, is quite another.

Good for prototypes and for finding out what you want. I built a working meditation app prototype that way, because I had never used one and wanted to see what it was about.

The catch:

• It randomly adds features that do not need to be there, does not follow instructions reliably, and the code is often in bad shape.
• You have to check everything, constantly. I asked for a change to one screen and noticed it had edited the profile file, which we were nowhere near.
• It is not free. Building one small thing that way cost me $300.
• AI speeds up junior developers and tends to slow down senior ones, because of all the checking. Fast to produce is not the same as fast to finish. Something can be built in three weeks and leave you with bugs for months.

What I actually recommend

Prototypes and learning: no-code and AI, freely. Production: a person, or you after you have properly learned.

And whichever you choose, the tool is not the decision that matters. There are plenty of courses that teach you to ship something with a no-code tool, and they work in the sense that you end up with an app. Without the research, there is no way to know the product is any good, because nothing in that process validates it.

One more thing, if you go the AI route: the vibe-coding world has discovered product requirements documents and is trying to write them, which is exactly right. A lot of the guidance on how to write one is coming from people outside product, and it tends to stuff everything into the document rather than the things that belong there.

How to choose, in one pass

1. Is this a test or the real thing? Test: cheapest tool available. Real: a person, or your own trained hands. 2. Do you have a written scope? If not, no choice here saves you. 3. What happens if it works? If you cannot answer how it gets maintained and by whom, the cheap route is more expensive than it looks.

Keep reading

What Do I Need to Know Before Hiring a Developer?

Do I Need a Technical Co-Founder to Build an App?

How Much Does It Actually Cost to Build an App in 2026?

Learn this with me

This decision is class 8 of From Passion to Product, my free six-week live course, which covers no-code platforms, vibe coding, AI-assisted coding and traditional engineering side by side, including the no-code rebuild trap. It comes after you have scoped the product in class 7, which is the order that saves money.

No coding, no-code or AI skills are needed to take it.

Join the waitlist for the next free cohort, or start with the App Strategy Workbook and make the decision on your own timeline.