Index Ai Academy
Creator of Index Ai Academy

Luke McLaughlin

Luke built Index Ai Academy, a free course that takes you from your first line of Python to building AI systems.

Luke works in digital marketing and trained as a molecular biologist. His interest in AI started with one idea: a large language model isn't written line by line like ordinary software. It is grown. Nobody types in what it knows. It develops through training, much as a brain develops through experience, only mathematically instead of organically.

For someone with a long interest in neuroscience and in how a brain develops, that was hard to leave alone. This page says why, and what he has made since.

Luke McLaughlin
Why AI

Grown, not written.

A newborn already has nearly all the neurons it will ever have, about 86 billion. What it doesn't have yet is the wiring. Over the first years of life the connections between neurons multiply at a huge rate, and then experience edits them: the ones that get used grow stronger, and the ones that don't are pruned away, a process that runs on through adolescence. No gene lists those connections one by one. There are trillions of them and only about 20,000 genes. The genes set the rules, and experience does the rest.

A language model is trained in a way that echoes this. It begins as a network of artificial neurons joined by billions of connections, each with a strength set at random. Training shows it text, has it predict the next word, and nudges the strength of each connection a little toward a better guess. After billions of words, the network can translate, write code and explain a joke. Nobody wrote those abilities in. They are held in the connections, put there by experience, which is where a brain is thought to keep what it has learned too.

This is more than a figure of speech. The first artificial neural networks were modeled on neurons, and people who build today's models describe them as "grown more than they are built". The researchers who study them borrow from neuroscience as well. They look inside a trained model for the units that respond to a particular idea, and trace how those units are wired into circuits, as a neuroscientist does with a brain. They have one advantage a neuroscientist would envy: in a model, every neuron can be read at once.

The likeness only goes so far, and Luke finds the differences as interesting as the likeness. A child hears a few million words a year, and a model reads billions. A brain keeps learning for life, while a model's connections are fixed once training ends. A brain adds and prunes its wiring, while a model only changes the strength of connections it had from the start. And the method that trains a model works out every adjustment with calculus, which, as far as anyone has shown, is not how a brain does it. That is the "mathematically instead of organically" part. Learning happens in both, by different roads.

The question he keeps coming back to is how a network of simple units, each doing very little, comes to behave as if it understands, whether it is made of cells or of numbers. It is why Index Ai Academy doesn't stop at how to use a model. It goes on to how one is trained, and to the math underneath.

What he works on now

Models you can run yourself.

Luke is especially interested in local LLM development: working with open-weight language models that you download and run on your own computer, with nobody else's server in between. A model on your own machine can be taken apart. You can read its weights, measure it, shrink it, change it and watch what happens, and your data stays where it is.

Not long ago a capable model needed a rack of expensive hardware. Now one with billions of weights can be stored in fewer bits per weight until it fits in the memory of an ordinary computer. He runs open models such as Qwen, Kimi and DeepSeek this way with Ollama, compares them with the large hosted models, and builds small tools on top of them.

The academy follows the same interest. It has a lesson on running open models locally, lessons on quantization and on working out whether a model will fit in memory, and a course on editing the weights of an open model directly.

Publications

Six books, all free.

Each one is a complete book, not a sample. They are free to download at aisearchhandbook.com, with no payment and no form to fill in.

To understand how a model is grown, Luke went back to the start: the math first, then programming, then how the models themselves are built and trained. As his knowledge grew he wrote it down, and that is where the books came from.

Math & Logic Essentials and Math & Logic Essentials for Programming are the mathematics everything else stands on. From Marketing to Technical Marketing is the road he took himself, from marketer to someone who writes Python and SQL. AI Engineering Essentials goes from machine learning to neural networks and transformers. And because his day job is search, Technical SEO for Digital Marketing and The AI Search Handbook cover where the two meet: how search engines, and now AI models, find and cite what they read.

Index Ai Academy is the same learning in another form, set out one step at a time. The books and the course are free.

Free PDF 391 pages

The AI Search Handbook

A complete guide to generative engine optimization: how AI search retrieves, weighs and cites information, and how to become a source that an AI answer names.

Free PDF 854 pages

AI Engineering Essentials

From NumPy and data preparation through machine learning, neural networks and PyTorch to transformers, RAG and MLOps. Every project has code you can run.

Free PDF 2,100 pages

Technical SEO for Digital Marketing

A hands-on course in fourteen chapters. Each project is a small Python tool that crawls and checks a site the way a search engine does.

Free PDF 874 pages

From Marketing to Technical Marketing

An illustrated path for marketers into Python, SQL and automation, with exercises and their solutions.

Free PDF 578 pages

Math & Logic Essentials

Mathematics and logic from the ground up: arithmetic, algebra, Boolean logic, trigonometry, geometry and calculus, with worked examples and a quiz in every chapter.

Free PDF 698 pages

Math & Logic Essentials for Programming

The same foundation with each idea also written as Python, so it runs as real code as soon as you understand it.

Elsewhere

Where to find Luke.