What is the difference between learning to use AI and learning to build AI?
Using AI means operating a finished product: you type an instruction and evaluate what comes back. Building AI means creating the thing that produces the answer. You gather and clean data, choose a model, train it, measure how far its predictions sit from the truth, diagnose why, and improve it. The first is product literacy and it expires when the product changes. The second is engineering and it does not.
Do I need to be good at mathematics to build AI?
You need less than people fear and more than the marketing suggests. The working set is linear algebra for how data is represented, calculus for how a model improves itself, and statistics for knowing whether a result means anything. We teach all three inside the machine learning rather than as a prerequisite course, so you meet each idea at the moment it explains something you just built. Students who arrive convinced they are bad at mathematics usually change their mind once it stops being abstract.
What age should someone start learning AI?
We teach from age 6 to age 67, but the entry point differs. A six year old starts with Python, drawing, and the idea that a computer can learn a rule from examples instead of being told it. A teenager starts with real data and their first trained model. An adult usually starts with the problem already on their desk. What does not change is the order: understand the data, then the model, then the evaluation.
Is prompt engineering worth learning?
Yes, briefly and properly, which is roughly six weeks rather than a career. Knowing how to brief a model well and how to catch a confident wrong answer is genuinely useful. Where it goes wrong is when it is sold as an AI education. Prompting rests on the behaviour of a specific product in a specific year. Model building rests on mathematics that has not changed in decades. We teach the first as a tool and the second as the subject.
What will I actually build in these classes?
Working software at a real address, not slide decks. Students train models on published datasets from their own country, then move to their own ideas. Past students have shipped an AI nutrition coach, a chatbot that explains mathematics and coding, an online safety assistant and a weather forecasting application, all live on the public web where anyone can open and inspect them.
How much does it cost and where do you teach?
Group batches of five to eight students are USD 40 a month and one to one teaching is USD 100 a month, shown in local currency on each country page. We teach live online to students in more than 25 countries, with class times set to local evenings rather than to our own convenience. The first class is free and does not require a card.
Are the classes live or recorded?
Live, always, with the same teacher week after week. Screens are shared in both directions so the teacher watches a student write code and interrupts at the moment the mistake happens, rather than marking it days later. Nothing on this site is a recorded course sold as a class, and nothing is auto-graded.
Can this replace a computer science degree?
No, and we will not claim otherwise. A degree gives you theory depth, a credential and time. What this gives you is the practical half taught properly and early: real code, trained models, an evidence portfolio, and the mathematics that university machine learning courses assume you already have. Most of our university-age students take it alongside a degree, not instead of one.
