Learning paths
Start from a goal. Work backward into what you need, how deep to go, and who is already on the way. Leave traces so the next person does not start from nowhere.
Community path
Learn Spanish
learn spanish
Community path
Implement a transformer
I want to understand how transformers work well enough to implement one.
Community path
Write a rom-com novel
I want to write a rom com novel
Community path
Build a tree house
I want to build a tree house
Community path
Host a dinner
I want to cook a dinner for friends
Community path
Play a song on guitar
I want to play a song on guitar
A topic is an atomic unit a learning path can contain — one idea, only as deep as the goal requires. Attach resources and your notes to a topic, detail which part of the resource helped the topic click.
Trending · 1
Attention
How tokens look at each other — the mechanism that makes a transformer a transformer.
Trending · 2
Neural networks
Layers, nonlinearities, and the idea that a network is a composition you can train.
Trending · 3
Backprop
Why autograd lets you write the forward pass and still debug when the loss does not move.
Trending · 4
Character desire
What each lead wants, what they are afraid of, and the private lie the story has to undo.
Trending · 5
Choose the tree
Species, health, and whether a living trunk can take a load without being wounded into decline.
Trending · 6
Linear algebra
Dot products as similarity, and the shapes you keep in your head while coding attention.
Have a goal of your own? Map the topics you need, how deep to go, and leave traces for whoever comes next.