AI #

Most writing about AI is either breathless hype or dense research jargon. This section is neither. It is plain-language explanations of how the tools everyone is suddenly using actually work under the hood, written for someone technical enough to be curious who does not want to wade through a paper.

Nothing here is tied to one company or product. These explain the ideas, meaning how the models work, how agents work, and how to use them well, so they stay useful no matter which tool you pick up.

Foundations: how the models work #

• How Large Language Models Actually Work: the prediction engine behind all of it.
• What Is a Token, and Why It Shows Up on Your Bill: the unit AI runs on.
• Embeddings, Explained: how software turns meaning into geometry.
• Why LLMs Hallucinate, and How to Reduce It: why AI confidently makes things up.
• Context Windows, and Why AI Agents Forget: the invisible memory limit.

Agents: how AI takes actions #

• What Is an AI Agent?: the difference between a chatbot and an agent.
• The Agent Loop: How AI Agents Actually Work: the single mechanism behind every agent.
• How Coding Agents Work Under the Hood: tools, navigation, and self-verification.
• Tool Use and Function Calling, Explained: how a text-only model runs real code.
• What Is MCP (Model Context Protocol)?: the standard that plugs tools into any AI app.

Using AI well #

• Prompt Engineering: How to Actually Get Good Answers: the practical basics.
• What Is RAG (Retrieval-Augmented Generation)?: making AI answer from your own documents.
• Fine-Tuning vs RAG vs Prompting: which of the three you actually need, and why fine-tuning is usually the wrong answer.
• Reasoning Models: What “Thinking” and “Effort” Actually Mean: models that work it out first.

Running it yourself #

• On-Device AI: What It Means and Why It Is Suddenly Possible: a language model in your pocket, and the four constraints that shape it.
• Quantization, Explained: the compression trick that makes a huge model fit on a phone.
• Running AI Models Locally: free, private, offline AI on your own machine.
• Open Weight vs Closed Models: three categories, not two, and what the licenses actually permit.

Images, speech, and privacy #

• How AI Image Generation Actually Works: diffusion, and why the model was trained by destroying pictures rather than making them.
• How Speech Recognition Actually Works: from a pressure wave to a sentence, and why transcription fails the way it does.
• What Actually Happens to Your Data When You Use AI: where your prompts go, who can see them, and the settings worth changing.

These pair naturally with the apps I build, which lean heavily on running AI locally and cheaply, and with the cheap-AI frugality guide.