How numbers learn to read, reason, paint and sing — told as a story, with every symbol pronounced out loud.
A 220-page tour of the mathematics that powers modern AI, written for everyone who ever stared at an equation and quietly decided it was not meant for them. From a single scalar to full transformers, diffusion models, audio codecs and AI agents — the whole modern map, in one coherent arc.
Pick your edition
Read it online in a protected browser reader with your email and phone — or hold it in your hands, printed and shipped.
Online edition
$3.99
One-time · 30 lifetime views
Printed edition
$14.89
+ delivery & 18% GST
I knew every English word around it. I understood, vaguely, that this line was the reason the machine on my phone could finish my sentences. And I could not read it. This book is the one I wanted that evening. — Vicky Yadav, from the preface
What’s inside
You will be able to open a modern AI paper when you finish — and hear it, rather than just see it.
Part I
Functions, vectors, matrices, probability, calculus, optimisation and information theory — rebuilt from the ground up, with every symbol pronounced in a plain-English table.
Chapters 1–9
Part II
From a single neuron to a full transformer: backpropagation, embeddings, attention, the block, the loop that turns it all into something that talks.
Chapters 10–15
Part III
How a model sees images, how diffusion paints from static, how sound becomes numbers — and how machines learn to speak, sing and listen.
Chapters 16–19
Part IV
Scaling laws, post-training, reasoning, agents, memory and retrieval, numerical tricks on real hardware, and why prediction becomes intelligence.
Chapters 20–25
Part V
Code you can run tonight — tool-calling by hand, Agno and PocketFlow agent frameworks, and a complete GPT built from scratch in ~130 lines of PyTorch (Karpathy-style).
Chapters 26–29
The back cover
29 chapters. 220 pages. One tea stall, one unreadable equation, and the ladder that was missing.
The paperback is the whole book on paper — made for a pencil, a cup of chai, and a slow afternoon. Every chapter ends with exercises, and the answers are at the back.
Read before you buy
Start where the book starts — at a tea stall outside the office, with a question about a chatbot that just wrote a monsoon poem.
Modern AI looks, from the outside, like a jungle of hundreds of unrelated techniques. From the inside it looks more like a tower, where each floor rests on the one below. Chapter 1 sketches that tower, writes down the one equation that rules them all, and introduces the two lives of every model: training and inference.
Open Chapter 1 →Meera found me at the tea stall outside the office, holding her phone out like evidence. On the screen a chatbot had just written her a perfectly decent poem about monsoon traffic.
“Explain this,” she said. “Not the marketing version. The real version. What is actually happening inside?”
I have been asked that question many times, and I have learned that the honest answer begins with a disappointment and ends with a wonder…
Why this book is different
01
Every equation is followed by a Reading the symbols table. ∇θ is pronounced “nabla theta” and explained in one plain sentence. No prior fluency needed.
02
With a smart, impatient student named Meera who keeps interrupting with the questions you are probably too polite to ask. The interruptions are where most of the real learning happens.
03
From one neuron to GPT-style transformers, diffusion, flow matching, audio codecs, agents and RLHF — plus a complete working GPT in ~130 lines of PyTorch.
The author
Eight years of building AI that millions of people use — at Google, at Ola’s Krutrim AI, and now in health research at Zyvelor AI Labs. This book is everything those years taught, written for the reader Vicky used to be.
Nov 2025 – present
Senior Data Scientist · health & medical AI
Built Kaira AI from scratch — a health assistant that logs meals from plain language and photos, and answers questions about body-composition (BCA) and continuous glucose monitoring (CGM) data. Designed the full LLM pipeline: query understanding, food-database retrieval, nutrition estimation and device-data context.
2025
Ads Solutions Architect (Product)
Search Ads 360 specialist. Architected backend integrations for campaign and ads management at scale for enterprise clients, and designed measurement and automation across Google’s ads product suite.
2024 – 2025
Senior Full Stack Lead
Led Krutrim AI Chat, the assistant built into the Ola app, and its ReAct agents (LangChain, LangGraph) for shopping, food, cab booking and support. Shipped agentic WhatsApp sales bots for 4,000+ Ola scooter centres, a Spark pipeline over 22 TiB that cut compute costs by a third, a 50–100 million image Indic text-image dataset, and live AI dubbing with ASR, TTS and face tracking for product launches.
2019 – 2024
From engineer to team lead
Enterprise AI assistants at Zluri; search infrastructure indexing 2M+ pages for Realtor.com, plus work for Microsoft and Dell at Brillio; a loan platform built with a team of twelve at RattanIndia; and the surveillance backbone now securing 50,000+ ATMs across India, alongside a medical dashboard for Uptake Medical, at QSS.
A few honest questions
The online edition is deliberately kept inside the browser, bound to the email and phone you used at purchase, with a lifetime cap of 30 views. That is what lets us price it at $3.99. If you want the book permanently on your shelf, the paperback is $14.89 plus delivery and GST.
The book is $14.89 (the same price converted to rupees for Indian buyers). Delivery is ₹70 to a Bengaluru address and ₹250 anywhere else, and 18% GST applies to the book and delivery together. The checkout shows the exact total, line by line, before you pay.
Prices on this page automatically switch to Indian Rupees when the request comes from an Indian IP, and Razorpay accepts UPI, cards, netbanking and wallets.
Go to books.threadlane.in/read, enter the email and phone you used at checkout, and type in the 6-digit code we email you. Each open counts as one of your 30 views.
No. The whole first part of the book rebuilds the mathematics it needs — numbers, vectors, probability, calculus — from scratch. If you can read this sentence, you can read this book.
Yours forever
Read it online for a few dollars, or hold the paperback in your hands. Either way, the whole mathematics of modern AI, one quiet step at a time.
The ebook — $3.99
🔒 Secure checkout via Razorpay. Cards, UPI, netbanking & wallets.