Fewer shortcuts. More understanding.
Learn AI.
Then prove you know it.
A complete path from concepts to calculations, from common mistakes to decisions that matter in real systems. Read, try, fail and restart: that is how learning works.
Recommended path
Understand first. Calculate next. Then challenge yourself.
Five laws to navigate representation, uncertainty, learning, optimisation and limits.
02Test the mechanismsChange parameters and observe the consequences without installing anything.
03Train properlyEighty exercises with immediate solutions and progress saved on your device. In Italian.
04Prove what you knowSixteen questions, four levels and nowhere to hide.
Moment of truth
Think you know AI? Prove it.
It starts simple. It does not stay simple. Concepts, systems, calculations and mathematics: this quiz rewards understanding, not acronym collecting.
01 · Foundations
The five laws, without the superficial version
These long-form guides and exercises are currently available in Italian.
Embedding, feature e spazi vettoriali
Come feature, embedding e spazi vettoriali trasformano dati e significato in una forma che i modelli possono elaborare.
Read the guide →02 · Law IIProbabilità, softmax e modelli generativi
Come probabilità, campionamento e modelli generativi rappresentano e trasformano l’incertezza nelle risposte dell’AI.
Read the guide →03 · Law IIICome una rete impara dai dati
Come dataset, obiettivi, architetture e addestramento trasformano esempi in parametri e comportamenti appresi.
Read the guide →04 · Law IVLoss, gradiente e ottimizzatori
Come loss, gradiente, learning rate e ottimizzatori guidano il training e l’efficienza dei modelli.
Read the guide →05 · Law VOverfitting, bias e collasso dei modelli
Come overfitting, bias, valutazione, sicurezza e governance rendono visibili e gestibili i limiti dei sistemi AI.
Read the guide →02 · Tools
Do not just read the formula. Move it.
Conv2d output size calculator for CNN layers
Calculate output shape and parameters for Conv1d, Conv2d, ConvTranspose2d and pooling, with stacked layers and PyTorch export.
Open the lab ↗02Confusion matrix calculator: precision, recall, F1, MCC
Calculate binary and multiclass metrics with an editable matrix, macro, micro and weighted averages, F-beta and MCC.
Open the lab ↗03Transformer parameter count and memory calculator
Calculate Transformer parameters by component and memory, with BERT-base and GPT-2 small presets and an Adam estimate.
Open the lab ↗04Cosine similarity and vector distance calculator
Calculate cosine similarity, dot product, distances and norms for two or more vectors, with a shareable similarity matrix.
Open the lab ↗05Learning rate scheduler visualizer with PyTorch code
Visualise and customise learning rate schedules with warmup, cosine, one-cycle, step and copyable PyTorch code.
Open the lab ↗06Token counter for GPT-style tokenizers
Count tokens, characters and words locally with cl100k_base and o200k_base, including highlighting and language comparison.
Open the lab ↗07Softmax and temperature explorer
Adjust logits and temperature, inspect softmax probabilities and measure the entropy and confidence of the resulting distribution.
Open the lab ↗08Activation function and derivative visualizer
Compare sigmoid, tanh, ReLU, GELU and SiLU with derivatives, saturation and interactive curves on a shared chart.
Open the lab ↗09MSE, RMSE, MAE, Huber and R² calculator
Calculate regression losses, residuals and R² from paired actual and predicted values, with an adjustable Huber parameter.
Open the lab ↗10Gradient descent, momentum and Adam simulator
Simulate SGD, momentum and Adam on convex and non-convex functions with adjustable trajectory, gradient and learning rate.
Open the lab ↗03 · Exercises
This is where theory stops protecting you
Embedding, feature e spazi vettoriali
16 solved exercises, step-by-step solutions and verification tools. In Italian.
Start →Law IIProbabilità, softmax e modelli generativi
16 solved exercises, step-by-step solutions and verification tools. In Italian.
Start →Law IIICome una rete impara dai dati
16 solved exercises, step-by-step solutions and verification tools. In Italian.
Start →Law IVLoss, gradiente e ottimizzatori
16 solved exercises, step-by-step solutions and verification tools. In Italian.
Start →Law VOverfitting, bias e collasso dei modelli
16 solved exercises, step-by-step solutions and verification tools. In Italian.
Start →Exam50 questions with answers
A final review to discover what you can really explain. In Italian.
Review →