Beginner notes from the book AI Engineering by Chip Huyen

I am still learning AI engineering, so this is not expert advice. If I got something wrong, please tell me.

The question

If I build an AI product and people like it, what stops someone else from copying it? The book calls this defensibility.

Think of a castle with a ditch of water around it. That ditch is called a moat. A strong moat means your product is hard to copy.

Why this is a problem for AI

Today it is very easy to build an AI product. You can use a ready-made model, write a prompt, and have a working demo in one weekend. That is great for beginners like me.

But if it is easy for me, it is easy for everyone. Even a big company can copy a small tool and put it inside an app that millions of people already use.

Three ways to build a moat

The book talks about three types of advantage:

1. Technology. Most people use the same few models, so this alone does not help much.
2. Distribution. This means how easily you can reach users. Big companies like Google and Microsoft already have billions of users, so they win here.
3. Data. This is where a small team still has a chance. If you reach the market first, you collect real usage data. You use it to improve the product, and a better product brings more users. This cycle is called a data flywheel.

What I understood

• Building the product is only the first step.
• Being early and collecting feedback from users seems very important.
• Solving one small problem well may be smarter than fighting big companies.

What I am still thinking about

For my own projects, what will make people keep using them even after someone copies the idea? I do not know yet, but I will keep learning and share more soon.