Quick Take
- Narration: Bryan Hughey delivers the conceptual material accessibly, though code-adjacent chapters will push most listeners toward the companion PDF.
- Themes: neural network fundamentals, gradient descent and backpropagation, Python as ML implementation language
- Mood: Introductory and aspirational, promising more hands-on depth than the runtime can fully deliver.
- Verdict: A brief orientation to neural network concepts that works as an entry-level survey but lacks the code depth and examples that practitioners need, approach with calibrated expectations.
I came to Building Neural Networks from Scratch with Python the way a lot of people approach introductory ML books: knowing enough about the field to have opinions, but wanting to understand what this particular title offered its intended audience. L.D. Knowings is positioning for the absolute beginner, and the runtime, three hours and fifty minutes, tells you something important about the depth of coverage. You don’t build much from scratch in under four hours.
The companion PDF is included with this Audible title. For any book making code-related claims, that PDF matters. The reviewers who rated this book well mention good educational content but wish for ‘more actual examples of code.’ The reviewer who rated it one star describes a book with ‘a lot of words but basically says nothing,’ written by someone who has been programming for a while and came hoping for implementation guidance. These two responses trace the exact divide that defines this kind of entry-level book.
What the Book Actually Covers in Its Runtime
The synopsis promises discovery of ‘the fundamentals of neural networks and their benefits,’ guidance on coding ‘without drowning in complex math equations,’ and coverage of ‘layers, gradients, and underfitting/overfitting.’ In a three-hour-fifty-minute audiobook, what these promises actually mean is: conceptual explanations of each topic, framed accessibly, without the mathematical rigor or code density that would make the concepts operational.
That’s not nothing. For a listener who has read AI news coverage but has no mental model for what a neural network actually is, what a layer does, why gradient descent is the optimization approach, what it means for a model to overfit, the book provides accessible explanations. Bryan Hughey’s narration is clear and keeps the pace moving without rushing through concepts that take a moment to absorb.
The Code Deficit That Divides Its Audience
The harsh one-star review describes what happens when a developer picks this up expecting implementation guidance. ‘This is the kind of book that has a lot of words but basically says nothing’ is a critique of the gap between the title’s promise (‘from scratch with Python’) and what the audio delivers. Building from scratch implies working through the mechanics of implementing a neural network, writing the forward pass, implementing backpropagation, handling the weight update, and a sub-four-hour survey that avoids complex math cannot deliver that.
The disconnect is partly a genre problem. ‘From scratch’ books in programming traditionally mean: here is how this thing actually works at the implementation level, with code you write yourself. This book uses ‘from scratch’ to mean: from no prior knowledge, building your conceptual understanding. Those are different things, and the title sets expectations that the runtime cannot fulfill for experienced developers.
The Comparison Point Listeners Should Have
If your goal is genuinely building neural networks from scratch with Python, Andrej Karpathy’s YouTube lecture series on micrograd is the free resource that actually delivers on that promise, you build a working neural network library from basic Python over the course of several hours, implementing every component manually. For an audiobook treatment with similar depth, Neural Networks and Deep Learning by Michael Nielsen (available freely online) does the mathematical groundwork. Neither is an audiobook, but they’re what a practitioner is actually looking for when they pick up this title.
What Knowings’s book offers that those resources don’t is a lower barrier to entry, shorter, more accessible, packaged for someone who isn’t sure they want to commit to a full technical deep dive. For that audience specifically, the book functions as intended.
Calibrated Expectations and the Right Reader
Use this book if: you’re a non-technical professional trying to understand what neural networks are before a meeting where someone will discuss them; you’re a complete beginner deciding whether machine learning is an area worth pursuing; you want a quick conceptual orientation before committing to a more demanding course. Don’t use it if: you know Python and want to implement something; you’ve already read any introductory ML content; you’re looking for the mathematical mechanics of backpropagation.
The title’s 3.7 rating reflects the exact split you’d expect: people who came for orientation found it adequate; people who came for implementation found it hollow. Be honest with yourself about which listener you are, and the book’s value proposition becomes clear.
Frequently Asked Questions
Does Building Neural Networks from Scratch with Python actually teach you to write code?
Not in a hands-on way. The book covers concepts and principles rather than walking through working code implementations. Reviewers who expected implementation guidance were disappointed, the ‘from scratch’ in the title refers to building conceptual understanding from zero, not implementing a network from the Python level up.
Is the companion PDF worth downloading for this title?
Yes. The PDF is included with your Audible purchase and contains supplementary material. For a book covering code-adjacent concepts, having visual reference available alongside the audio helps reinforce what the narration describes.
What book should I read after this if I want to actually build a neural network?
For a free, genuinely hands-on resource, Andrej Karpathy’s micrograd lecture series on YouTube takes you through building a working backpropagation library from scratch in Python. For a structured audiobook approach with more depth, explore books specifically designed as developer guides rather than introductory surveys.
How does this compare to other beginner ML books on Audible?
At under four hours, this is among the shortest ML introductions available. It’s appropriate for absolute beginners seeking orientation rather than developers seeking implementation guidance. Longer titles like those from O’Reilly’s catalog or Manning’s guides (many available on Audible with companion PDFs) provide substantially more depth.
