Primer · seven volumes
Field manuals you can poke.
Short labs, one idea each. Read the plain-English box, poke the diagram, keep one sentence. Algebra, learning, and information sit under the rest if a later volume feels like a foreign language.
In English
Each volume is a stack of short labs. Read the plain-English box, poke the diagram, keep one sentence. If later volumes feel like a foreign language, start with Algebra.
- Try this
- New to all of this? Algebra, then Information, then LLMs. Here for agents? Start with Tokens.
- Keep this
- One idea per lab. Click it. Then go on.
- Volume ILLMsHow a language model thinks — and how engineers wire it.Tokens, attention, the cache, the loop, the harness.
- Volume IIQuantumA state is a direction. Measurement is a question you only get to ask once.Superposition, interference, entanglement, Deutsch.
- Volume IIIComputingPhysics is the substrate. This is the computer: registers, oracles, Grover, the fridge.Registers, Grover, teleportation, the stack.
- Volume IVAlgebraThe shared language. Vectors, maps, the arrows that only stretch.Vector, matrix, eigen, projection.
- Volume VLearningThe model is a fitted function. Loss, a slope, a step, then the ways that fails.Loss, gradient, backprop, overfit.
- Volume VIInformationSurprise you can count. Entropy, KL, and the number an LLM actually minimizes.Bits, entropy, cross-entropy, compression.
- Volume VIICryptoOne-way functions, RSA, the discrete log — and the stories Shor makes false.Mod, RSA, hashes, the threat.