Skip to content
Hi, Bot

Hi, Bot · First Principles

48 lessons. One a day. No skipping.

It starts with writing instructions precise enough for a machine to follow, and it ends with you training a neural network you wrote yourself, in NumPy, with no framework imported anywhere. Every step in between is earned.

Each lesson ends with one question. Get it right and the next lesson opens on the spot — race through ten tonight if you want. Get it wrong and that same question comes back tomorrow. One attempt a day, which is exactly enough pressure to make you read carefully.

Start at lesson 1

By submitting, you agree to our Terms and Privacy Policy.

Free, forever, no account. Read lesson 1 first →

The climb

Six phases. Nothing appears before the thing it depends on, which is the entire reason this works when a playlist of videos does not.

Phase 1

Logic, functions, symbols

Lessons 1–8

Say a thing so precisely that a machine could do it.

  1. 1Instructions a machine could follow
  2. 2Loops and branches
  3. 3Variables and state
  4. 4Function as a black box
  5. 5Balancing equations
  6. 6The coordinate plane
  7. 7Python: syntax and types
  8. 8Python: files and automation
Phase 2

Structure, proof, discrete systems

Lessons 9–18

Learn what counts as knowing something for certain.

  1. 9Exponents and their inverse
  2. 10Logarithmic scale
  3. 11Polynomials and roots
  4. 12Complex numbers
  5. 13Deduction
  6. 14Proof technique
  7. 15Boolean algebra
  8. 16Counting
  9. 17Sets
  10. 18Arrays, lists, dicts, and Big O
Phase 3

Calculus and complexity

Lessons 19–27

Change, accumulation, and what an answer costs.

  1. 19Limits
  2. 20The derivative
  3. 21Rules of differentiation
  4. 22The chain rule
  5. 23Integration and the fundamental theorem
  6. 24Series and convergence
  7. 25Recursion
  8. 26Trees and heaps
  9. 27Sorting and searching under constraint
Phase 4

Many dimensions

Lessons 28–35

Do all of the above to space itself.

  1. 28Vectors
  2. 29Matrices as transformations
  3. 30Systems, determinants, rank
  4. 31Basis and change of basis
  5. 32Partial derivatives
  6. 33The gradient
  7. 34Jacobians and vector fields
  8. 35Graphs, DAGs, tensors
Phase 5

Spectra, probability, optimization

Lessons 36–43

Uncertainty, compression, and rolling downhill.

  1. 36Eigenvalues and eigenvectors
  2. 37SVD and dimensionality reduction
  3. 38Continuous distributions and PDFs
  4. 39Expectation, variance, covariance
  5. 40Bayes' theorem
  6. 41Convexity and gradient descent
  7. 42Lagrange multipliers
  8. 43Entropy, cross-entropy, KL divergence
Phase 6

Build it

Lessons 44–48

Write the model yourself, from nothing.

  1. 44Forward pass from scratch
  2. 45Backpropagation from scratch
  3. 46Attention from scratch
  4. 47Diffusion
  5. 48Reinforcement learning

At the top, a certificate

Clear all 48 and you get a permanent, shareable Hi, Bot certificate with your own verification code. It is worth something precisely because there was no way to skip a rung: every lesson on it was unlocked by an answer you got right.

Doing this with a friend roughly doubles the odds either of you finishes. Send them hibot.space/first-principles.