It’s already part of your day
AI writes drafts, answers questions, recommends what to watch and sorts your inbox. Knowing how it works helps you use it well.
New courses on the way
Learn how AI works by playing with it.
Short, interactive courses on how AI learns, the maths underneath, and what really happens inside a chatbot. Every number comes from real data or a real model. No maths background needed, nothing to install.
- 3 courses
- 21 interactive lessons
- Real data in every lesson
- Free to start
Hi, I’m Spark! Click the chart to confuse my friend the network.
Training round 0Training accuracy 57%
This network is learning right now, in your browser, on real data. Click the plot to plant a contrarian point and watch it adapt.
Data source: GeoNames cities15000 CC BY 4.0
Why learn how AI works?
You don’t need to build AI to benefit from understanding it.
Know when to trust it
A language model predicts likely words. It doesn’t look facts up. Once you see that, you can spot confident nonsense and check what matters.
See past the hype and the fear
At work and in the news, decisions about AI are being made right now. Understanding the basics lets you take part instead of guessing.
Ideas that outlast the tools
Products change every few months. Data, loss, gradients, vectors and attention don’t. Learn them once and every new tool makes more sense.
Courses
Every course starts with a free module, and new courses are on the way.
AI fundamentals
From “what is a model?” to how a chatbot picks its next word — nine hands-on lessons, no maths background needed.
- What learning means
- Getting better, step by step
- Neural networks
- How language models work
How AI assistants work
From next-word prediction to the chatbots you use every day: how they are trained, why they make things up, and how to get reliable answers from them.
- From text to prediction
- From predictor to assistant
- Using assistants well
The maths behind AI
Just enough maths to see inside a model: vectors, matrices, derivatives, the chain rule and probability — every idea shown with live, exact calculations.
- Vectors and matrices
- Change and optimisation
- Probability
More courses on the way
We’re writing new courses, built the same way: interactive, bite-sized and based on real data. Pro members get every new course at no extra cost.
How we teach
Reading about AI only gets you so far. Here you poke at it until it clicks.
Learn by doing
Each lesson is built around an interactive: drag the points, tune the knobs, train the network and watch what changes.
Real data, real models
Every number comes from a public dataset or the recorded outputs of a real model like GPT-2. Nothing is made up to look nice.
Sources you can check
Every lesson lists its sources. Spot something wrong? Report it in one click and an editor will look at it.
Bite-sized and friendly
Lessons take about ten minutes and end with a quick quiz. Earn XP, keep a streak, and Spark cheers you on.
Questions
Do I need to know maths or coding?
No. Lessons start from everyday ideas and build up step by step. If you want to go deeper, there’s a whole course on the maths, and it starts from the basics too.
Is it really free?
Yes. The first module of every course is free for good, and you can create an account without a card. Pro unlocks the rest.
Does it run a real AI model in my browser?
Lessons use outputs recorded from real models, like GPT-2 and Qwen, and real public datasets. Some interactives, like the network on this page, train a small network live in your browser.
How do you keep the lessons accurate?
Every figure is traced back to a dataset, a model or a published paper, and every lesson lists its sources. Lessons are reviewed, and anyone can report a mistake from any lesson.
Can I cancel Pro?
Yes, at any time from your account. You keep Pro until the end of the period you’ve paid for. Payments aren’t refundable, so try the free modules first.
Ready to see how it really works?
Your first lesson takes about 8 minutes.