Skip to content
Clawdemy

References: Seeing it whole, and where next

Source material

Section titled “Source material”
Source curriculum (structural mirror, cited as further study):
• 3Blue1Brown, Neural Networks (series), by Grant Sanderson (text adaptation by Josh Pullen)
  Series index: https://www.3blue1brown.com/?topic=neural-networks
  License: copyright Grant Sanderson; videos published on his site and YouTube
This lesson is an original synthesis; it mirrors no single chapter. The track as
a whole follows the pedagogical arc of the 3Blue1Brown Neural Networks series.
Clawdemy's lessons are original prose; we do not reproduce or transcribe the
videos, and cite them as the recommended companion. All rights remain with the
creator.

Watch this next

Section titled “Watch this next”
  • The Neural Networks series (3Blue1Brown) by Grant Sanderson. You now have the foundation to watch the whole series end to end and follow every frame: Chapter 1 (what a network is), Chapter 2 (gradient descent), Chapter 4 (what backprop is doing), and Chapter 5 (the backprop calculus) map directly onto this track. If you watched chapters as you went, this is the moment to rewatch the series whole and feel how the pieces connect.

Going deeper

Section titled “Going deeper”

A short, durable list. Each link is a specific next step, not a generic pile.

  • Neural Networks and Deep Learning (full book) by Michael Nielsen. The free online book that has shadowed this whole track. Now that you have the intuition, reading it cover to cover fills in the equations and the practical details with the understanding already in place.

  • TensorFlow Playground. One last visit, now that every control means something to you: the layers, the neurons, the activation function, the learning rate, the loss falling. Nothing on that page is a mystery anymore.

Where to go next in Clawdemy

Section titled “Where to go next in Clawdemy”

The honest routing, by what you want to do.

  • Track 13 (Build Neural Networks from Scratch). If you want to build what you just understood, in Python from first principles. The gradient descent and backpropagation of this track, as code you run and watch learn.

  • Track 5 (AI Foundations). If you want to understand how modern language models work; it covers transformers and LLMs. A transformer is a particular kind of neural network, so this track’s forward pass, cost, gradient descent, and backprop carry straight over.

  • Track 20 (AI Agents and Tool Use). If you want to use AI to build things. Agents are wired up on top of trained networks; a different altitude, resting on the foundation you now have.

Adjacent topics

Section titled “Adjacent topics”

The pieces of this track, for quick reference.

  • The structure (lessons 1 to 4): a network as a function from 784 numbers to 10, built from layers of neurons, each computing weighted-sum-plus-bias-plus-squish, parameterized by about 13,000 knobs.

  • The learning (lessons 5 to 9): a cost function to minimize, a landscape to descend, gradient descent to take the steps, and backpropagation (the chain rule, run backward) to compute the downhill direction.