I have been reading AI Engineering by Chip Huyen, and one idea has stayed with me. Building AI applications today is much easier than it used to be.

Starting with a pre-trained brain

In traditional ML engineering, you had to collect data, design features, and train a model from zero for every single problem. It took a lot of time, data, and money.

Foundation models changed this. Now we start with a model that has already learned from a huge amount of text. It is like a pre-trained brain. Our job is to build on top of it. We guide it, connect it to our own data, and shape it into a useful application. We no longer need to build the brain ourselves.

It is not closed, it is open ended

Old ML models usually did one fixed task, like predicting a price or tagging an image. Modern LLMs are open ended. We can chat with them as long as we want, about anything we want.

LLMs can also use outside sources. With a method called RAG (retrieval augmented generation), the model looks up information from external documents and uses it to write better answers. So it does not only depend on what it was trained on.

How LLMs read text: tokens

An LLM does not read words the way we do. It breaks text into small pieces called tokens. A token can be a whole word, part of a word, or even a single character. For example, a word like “unbelievable” might be split into “un”, “believ”, and “able”. The model reads and writes everything one token at a time. This is also why AI usage is often measured and priced in tokens.

Companies are seeing results

This shift is also showing up in business. Companies are getting better returns on their AI investments over time. One reason AI engineering is growing so fast is that companies are putting a lot more money into AI. In the book, Chip Huyen lists this as one of the main factors behind the rise of AI engineering. She mentions a Goldman Sachs Research estimate that AI investment could reach about $100 billion in the US and $200 billion worldwide by 2025.

When this much money flows in, more tools, more jobs, and more AI products get built. Companies also see AI as a way to stay ahead of their competitors, so they keep investing. That is good news for anyone learning AI engineering today.

My takeaway

AI engineering today is less about building models from scratch and more about building smart applications on top of powerful ones. That makes the field easier to enter, and it makes our lives easier too.