AI Foundations Guide

From biological intelligence to the frontier — 27 chapters.

Part I — Intelligence in Nature

  1. 01 From Reflex to Reason Single-cell organisms, nervous systems, the evolutionary ladder from stimulus-response to planning.
  2. 02 The Brain Neurons, synapses, brain structures, and how biology implements learning.
  3. 03 From Neuroscience to AI McCulloch-Pitts, Hebb, the perceptron, XOR, the AI winter. Connectionism vs symbolicism.

Part II — The Math Behind It

  1. 04 Mathematical Foundations Linear algebra, calculus, probability, and information theory — each tied to where it appears in AI.

Part III — Building Artificial Neural Networks

  1. 05 Neurons to Networks Weights, activations, the math of a single neuron, layers, and deep networks.
  2. 06 The Training Loop Forward pass, loss functions, backpropagation, gradient descent, overfitting, regularization.
  3. 07 GPUs and the Hardware Revolution Why parallel matrix math on graphics cards was the unlock. The CUDA moment.
  4. 08 Architectures CNNs, RNNs, LSTMs, autoencoders, GANs — what each solved and what it couldn't.

Part IV — How LLMs Work

  1. 09 The Transformer Attention, self-attention, multi-head attention, positional encoding. Why this architecture won.
  2. 10 Tokenization Text to numbers. BPE, vocabulary, subword units. What the model actually sees.
  3. 11 Pre-training Next-token prediction, the data, the scale. What emerges and why nobody expected it.
  4. 12 What the Model "Knows" Embeddings, latent space, superposition. Why it's compression, not growth.

Part V — Shaping and Using a Model

  1. 13 Fine-tuning SFT, RLHF, instruction tuning, alignment. What changes in the weights.
  2. 14 LoRA and Efficient Methods Low-rank adaptation, QLoRA, adapters. Why you don't need to retrain everything.
  3. 15 Inference Autoregressive decoding, temperature, sampling, KV cache, context windows.

Part VI — The Field Today

  1. 16 The Communities cs.AI vs cs.ML vs cs.IR vs cs.CL. Conferences, journals, how to read the landscape.
  2. 17 The Labs OpenAI, Anthropic, DeepMind, Meta AI, Mistral. The open vs closed debate.
  3. 18 The Hardware Stack Nvidia, AMD, TSMC, Google TPUs, HBM. What each layer does and why Nvidia dominates.
  4. 19 Hyperscalers and Enterprise AI Azure, AWS, GCP. How AI gets deployed at scale.
  5. 20 The Applied Landscape RAG, tool use, MCP, agentic engineering, multi-agent systems.

Part VII — The Frontier

  1. 21 Beyond LLMs Reinforcement learning, world models, JEPA, neuroevolution.
  2. 22 Continual Learning Catastrophic forgetting and why models can't just keep learning.
  3. 23 Intrinsic Motivation and Curiosity Prediction error as drive, open-ended learning. The constraint-as-feature insight.
  4. 24 Quantum Computing and AI What quantum gives you, what it doesn't. Honest state of the field.
  5. 25 Neural Interfaces BCIs, Neuralink, convergence of biological and artificial intelligence.
  6. 26 The Gap What exists vs what's missing. Growing architecture, survival drive, persistent learning.

Part VIII — Your System

  1. 27 Connecting It Back What this means for what you're building. Hardware, research opportunity, next steps.
kusp — 2026