← BACK TO CATALOGUE A KNOWLE · FIRST EDITION · FOLIO SERIES
How AI Really Works by Leo Odåsnac
FORMATS Digital (WAB) · Paperback · Kindle
DIMENSIONS 100 pages · 18,500 words · 90 min read
MONOGRAPH · VOLUME 02 — TECHNOLOGY

How AI Really Works

From Prediction to Reasoning: What Happens Inside an LLM When It Learns, Answers, and Thinks
LO

AUTHOR

Leo Odåsnac

CURATED KNOWLEDGE FOLIO
Every day, millions of people ask AI questions and receive astonishingly human-like answers. But what is actually happening inside the machine? Is ChatGPT really thinking? Does it understand what we say? Where does its knowledge come from? And why can something so intelligent sometimes make such ridiculous mistakes? H…
menu_book READ FREE ON WAB arrow_forward BUY ON AMAZON open_in_new

verified Always 100% free to read online. Physical editions printed on demand on sustainably sourced unbleached paper.

MONOGRAPH PRÉCIS

The Core Argument

Follow a prompt from words to numerical patterns and back again, and discover how training, attention, and token-by-token generation let an LLM do far more than simple autocomplete. You’ll see how prediction can support reasoning—and why the same machinery can produce both surprising insights and convincing mistakes.

SEQUENCE · COMPLETE TEXT

The Reading Path

Crafted for a single contemplative evening or two morning sittings. Each chapter functions as a self-contained mental model and immediate intervention.

FORMAT Printed Monograph & Web
LENGTH 90 Min Read
ACCESS Complete & Unrestricted
READ FIRST CHAPTER →
Chapter 1

The Prediction Puzzle

  • A Prompt Is Not Yet Meaning
  • The Next-Piece Game
  • The sentence is only the beginning
  • The Question Behind the Question
Chapter 2

Words Become Numbers

  • Tokens: The Pieces the Model Receives
  • An Invisible Map of Relationships
  • Billions of Adjustable Settings
Chapter 3

How Prediction Builds a Model

  • Training: The Enormous Prediction Game
  • Layers Transform the Clues
  • Inference: Using What Training Built
Chapter 4

Context Changes What Matters

  • Attention: Selective Listening in the Sentence
  • The Context Window and Its Limits
  • One Prompt, Several Relationships
Chapter 5

From Likelihoods to an Answer

  • Choosing the Next Token
  • A Prompt’s Journey Through the Machine
  • A Branching Set of Possible Answers
  • Why Fluent Does Not Mean Certain
Chapter 6

When Prediction Starts to Look Like Reasoning

  • Working Through a Harder Problem
  • Tools, Checks, and Course Corrections
  • A Powerful Pattern Learner, Not an Oracle
FOUNDATIONS

Scholarly References

Peer-reviewed sources, field research, and primary literature grounding this monograph.

  1. Reflective Reasoning System: Inference-Time Self-Diagnosis and Self-Correction for Large Reasoning Models (2025)
CURATED FOLIOS

Explore More Knowles

VIEW ENTIRE LIBRARY →
AI Agents: When AI Starts Doing the Work
A KNOWLE · 90 MIN

AI Agents: When AI Starts Doing the Work

AI Agents: When AI Starts Doing the Work provides an authoritative and accessible exploration into its subject matter. Designed specificall…

INSPECT →
Introvert, so what?
A KNOWLE · 90 MIN

Introvert, so what?

Introverted, So What? provides an authoritative and accessible exploration into its subject matter. Designed specifically for readers seeki…

INSPECT →
OPEN FOLIO ACCESS

Begin reading "How AI Really Works" right now.

Free unabridged edition with instant cloud bookmarking.
auto_stories OPEN WEB READER