How AI Actually Works

A Living Companion to the Compression Series

Free · Version 1.0 · Updated as the series grows

AI sounds confident whether it's right or wrong. That's not a bug — it's the mechanism. This free companion explains why, in plain language, and ties it directly back to every book in the Compression Series.

Works whether you read it before or after The AI Compression Playbook. It's short, it's free, and unlike the rest of the series, it's a living document — versioned and updated as both the underlying tools and the series itself change.

What's inside

• What an LLM actually does — next-token prediction, not knowledge or belief, explained in one page
• Why confidence isn't correctness — the mechanism behind every failure mode named across the series: delusion, abdication, misrepresentation, inflation
• What training gives it (and what it doesn't) — context windows, and why AI doesn't know your specific business unless you tell it
• Why this explains every book in the series — the one mechanism, named four different ways
• 10 prompts for checking AI's own confidence — not prompts that make it more accurate, prompts that make its uncertainty visible to you
• A short glossary — token, context window, hallucination, training data, inference, in plain language
• A version history — because this document is built to be updated, not republished as a new product every time something changes

Who this is for

Anyone using AI for real work who wants the actual mechanism behind why it sometimes sounds completely sure about something completely wrong. No prior reading required — though it connects directly to the rest of the Compression Series if you have read it.

Why it's free

This isn't a lead-in to a paywall — it's genuinely free, and it's meant to stay useful without asking you to buy an update. New versions replace this one at the same link as the tools change and the series grows.

What you get

Instant download in PDF, EPUB. 11 pages.

How AI Actually Works — by Nathaniel Grey, published by Second Brain Publishing.