Code Safari
All subjects

The AI guides & blogs

AI

Build production features on top of LLMs — prompting, retrieval, agents — and evaluate, secure, and afford them.

56 guides & blogs·8 stages·≈ 10 hr·4 more being written

Start with Common Foundations — it counts on every subject.

  1. 01

    ML & neural network foundations

    ≈ 2 hr

    The ground floor: how models learn from data and what a neural network actually is.

    Machine Learning for Curious Minds (No Math Required)A plain-English guide to how machine learning actually works — learning from data, the three flavors of learning, overfitting, why data quality decides everything, how neural networks learn, and how a model goes from idea to deployment. No math required.6 chapters·≈ 1 hr·also under ML
    Neural Networks Basics (No Math Required)A plain-English guide to neural networks — what they actually are, the anatomy of a neuron (weights, biases, activation functions), training one end to end from random dials to reading digits, and the family tree of architectures: CNNs, RNNs, and the Transformer behind ChatGPT.4 chapters·42 min·also under ML
  2. 02

    How generative AI works

    ≈ 1 hr

    Next-token prediction, hallucinations, and the honest limits of generative models.

    How Generative AI Actually Works (No Math Required)A plain-English guide to how generative AI and large language models actually work — next-token prediction, training, hallucinations, context windows, and honest limits. No math required.6 chapters·≈ 1 hr
  3. 03

    Large language models

    ≈ 2 hr

    Tokens, embeddings, Transformers, attention, training, fine-tuning, and inference — the full anatomy.

    How Large Language Models WorkA clear, plain-English guide to how large language models actually work — from tokens and embeddings to Transformers, attention, the context window, training, fine-tuning, inference, and their real limits.10 chapters·≈ 2 hr
  4. 04

    Prompt engineering

    ≈ 2 hr

    From basic prompting to chain-of-thought, structured output, system prompts, and evaluation.

    Prompt Engineering from Beginner to AdvancedA practical, example-driven guide to prompt engineering — from basic prompting, zero-shot and few-shot, to chain-of-thought, structured output, prompt templates, system prompts, RAG prompts, and how to evaluate prompts properly.9 chapters·≈ 2 hr
  5. 05

    Retrieval-augmented generation

    ≈ 1 hr

    Embeddings, chunking, vector databases, retrieval, re-ranking, and evaluating the whole pipeline.

    RAG Explained: Retrieval-Augmented Generation, End to EndA plain-English guide to how Retrieval-Augmented Generation actually works — why RAG exists, embeddings and semantic search, chunking, vector databases, retrieval, re-ranking, and how to evaluate a RAG system. No hype.7 chapters·≈ 1 hr
  6. 06

    AI agents

    ≈ 1 hr

    Planning, memory, tool calling, multi-agent systems, MCP, and the frameworks that wire it together.

    AI Agents Explained: From LLM to Autonomous WorkerA plain-English guide to how AI agents actually work — the perceive-think-act loop, planning, memory, tool calling, multi-agent systems, the Model Context Protocol (MCP), and agent frameworks. No hype.7 chapters·≈ 1 hr
  7. 07

    The economics of AI

    ≈ 1 hr

    GPUs, training costs, per-token pricing, and power — what your AI feature actually costs to run.

    The Economics of AI: Why It Costs So Much (and Keeps Getting Cheaper)A plain-English guide to the money behind AI — why everything runs on GPUs, what it really costs to train a frontier model, why you pay per token, the data centres and power behind every reply, why prices keep collapsing anyway, and who actually makes money. No hype.7 chapters·≈ 1 hr
  8. 08

    Shipping AI features

    The gap between demo and product: evals, guardrails, fine-tuning decisions, and serving realities.

    Evaluating LLM outputsbeing written

    Will cover: golden datasets & regression evals · LLM-as-judge & its pitfalls · offline evals vs online metrics · evals as CI for prompts

    Fine-tuning in practicebeing written

    Will cover: when RAG beats fine-tuning (and vice versa) · LoRA & parameter-efficient methods · building a training set from logs · evaluating a fine-tune against the base model

    Guardrails & safetybeing written

    Will cover: prompt injection & jailbreaks · input/output filtering · PII & data-handling in prompts · human-in-the-loop patterns

    Serving LLMs in productionbeing written

    Will cover: choosing a model: quality vs cost vs latency · streaming, timeouts & retries · caching & batching strategies · observability for LLM calls

Shelves don't teach. Guides do.

Start: Machine Learning for Curious Minds