Mechanisms, not metaphors
Most free AI content stops at analogies. Here you compute an attention weight, apply a convolution kernel and tune a PID gain yourself, in Python you can run today.
Neural networks, language models, Bayesian reasoning and robotics, taught with real code and free for everyone.
No black boxes. You implement a neuron, an attention head, a spam filter and a PID controller yourself, then prove it at mastery level.

Convolution, attention, priors, control loops. Taught properly, with working code, at no cost and with nothing to sell you.
Most free AI content stops at analogies. Here you compute an attention weight, apply a convolution kernel and tune a PID gain yourself, in Python you can run today.
Every unit uses retrieval practice and mastery checks, the two methods research consistently ranks highest. You recall and apply, you do not just watch.
Every mastery check demands 90 per cent, so the certificate means you can actually do it, not that you watched a video to the end.
Start with the AI and robotics core, or build the foundations first. Every track ends with a certificate.
The same four steps in every unit, because this is the sequence that makes learning stick.
A short, plain English explanation with real worked examples. No filler and no jargon.
Retrieval questions with instant feedback that tells you why, so mistakes teach you something.
A small real task you actually do, with a checklist. This is where a skill becomes yours.
A mastery check at 90 per cent. Pass every unit and the track certificate is issued.
Not a trial, not a teaser, and not free until the useful part. Here is exactly what you get.
Everything on this site, for every learner, with no account required.
WAJD Academy covers the exam side: 11 Plus, GCSE and A Level, also free.
Everyone explains what AI does. Almost nobody free will show you the mechanism.
Every explainer told me attention lets the model focus. Actually writing the softmax and watching the weights land on one key is when it finally clicked.
Sixth form learnerInside a large language modelThe 99 per cent accurate test being wrong nine times out of ten genuinely changed how I read every statistic since. I had to work it out myself to believe it.
Adult learnerBayesian thinkingI tuned a PID controller until it stopped oscillating and finally understood why our robotics club project kept shaking itself apart.
Robotics club memberPhysical AI and roboticsBuild a neuron, train it with gradient descent, then make a network see with convolution. Four units, real code, completely free.