Practical. Deep. Built for developers.

Learn to build software from first principles.

Build real systems. Understand how tools work. Become a stronger engineer in the age of AI.

Build your own Neural Network

1 class Value:

2 def __init__(self, data, children=()):

3 self.data = data

4 self.grad = 0

5 self._prev = set(children)

7 def __mul__(self, other):

8 out = Value(self.data * other.data)

9 out._prev = (self, other)

10 def _backward():

11 self.grad += other.data * out.grad

12 other.grad += self.data * out.grad

Courses

Systems you can reason about.

Each course starts with primitives, then moves through real tests, performance tradeoffs, and implementation details.

Backpropagation, gradients, tensors, training loops, and a clean test suite.

Sockets, parsers, routing, concurrency, and the protocol details beneath frameworks.

Hashing, memory layout, collision handling, resizing, and cache-aware tradeoffs.

Tokenization, attention, transformers, training data, and inference from the metal up.