You’re sitting in a lecture hall. The professor is scribbling massive grids of numbers on a chalkboard, muttering about row reduction and pivots. It feels like accounting, but worse. If you’ve ever felt like linear algebra was just a series of tedious, soul-crushing calculations, you aren't alone. Most people learn it backward. They start with the "how" (crunching numbers) instead of the "why" (the actual geometry and logic). This is exactly why Sergei Treil’s Linear Algebra Done Wrong became a cult classic in the math world.
The title is a tongue-in-cheek jab at another famous book, Sheldon Axler’s Linear Algebra Done Right. While Axler’s book is great, Treil wrote his version for the honors linear algebra course at Brown University because he felt students needed a more honest, rigorous approach that didn't shy away from the "wrong" things—like determinants—early on. It’s a masterpiece of mathematical pedagogical rebellion.
The Determinant Controversy
Standard curriculum treats the determinant like a magic number you find by doing a bunch of cross-multiplication. It's boring. Honestly, it’s frustrating. Treil argues that avoiding determinants (the "Done Right" approach) actually makes things harder to visualize in the long run. In Linear Algebra Done Wrong, the determinant isn't just a formula; it’s a tool for understanding how much a transformation scales an area or volume.
Think about it this way. If you have a square with an area of 1 and you apply a linear transformation to it, the determinant tells you the area of the new shape. If the determinant is 0, your square just got squashed into a line or a point. It vanished. That’s why a zero determinant means a matrix isn't invertible. You can't "un-squash" a point back into a square. There’s no information left to work with.
Treil’s book gets into this early. He doesn't wait until the end of the semester to tell you why you’re doing the math. He treats you like an adult. This approach is what makes it feel so different from the dry, corporate-style textbooks pushed by major publishers. It’s gritty. It’s dense. It’s brilliant.
Why Eigenvalues Are the Secret to Everything
If you’re into machine learning, data science, or even just physics, you’ve heard of eigenvalues and eigenvectors. Most textbooks make them feel like an abstract chore. You solve $det(A - \lambda I) = 0$, you get some numbers, you move on. But Treil frames these concepts as the "characteristic" traits of a transformation.
Basically, an eigenvector is a direction that doesn't change when you apply a transformation. It might get stretched or shrunk (that’s the eigenvalue), but it stays on its original line.
🔗 Read more: Logitech Wireless Computer Mouse: Why Your Choice Actually Matters
- In facial recognition, these are "Eigenfaces."
- In Google’s original PageRank algorithm, the "most important" web page is essentially an eigenvector.
- In bridge construction, eigenvalues help identify resonance frequencies so the wind doesn't knock the whole thing down.
The book focuses heavily on the Spectral Theorem. It’s a big deal. Treil shows that for certain types of matrices (self-adjoint), you can find a set of eigenvectors that are all perpendicular to each other. This is essentially like finding the "natural" coordinate system for a problem. Instead of using the $x$ and $y$ axes we’re used to, we use the axes the data wants us to use.
The Problem With "Cookbook" Math
Most undergraduate courses are "cookbook" courses. Here is a recipe for a 3x3 matrix. Here is a recipe for a system of linear equations. Follow the steps, get the grade, forget everything in two weeks. Treil hates this. Linear Algebra Done Wrong is built on the idea that if you understand the underlying structure—vector spaces, linear maps, and inner products—the calculations become trivial.
It’s hard. I won't lie to you. The book moves fast. It expects you to handle proofs from page one. But that’s the point. Linear algebra is the language of modern technology. Every time you filter a photo on Instagram, a matrix is doing the work. Every time an AI generates a sentence, it’s navigating a high-dimensional vector space. If you only learn the "cookbook" version, you’re just a user. If you learn the Treil version, you’re an architect.
Treil, who is a professor at Brown, actually made the PDF available for free online. This contributed to its "underground" success. It wasn't pushed by a sales rep from a multi-billion dollar publishing house; it was shared by students and researchers who finally found a text that didn't talk down to them.
Real-World Applications You Actually Care About
We often talk about math in the abstract, but let's look at the Singular Value Decomposition (SVD). Treil covers this toward the end of the book. SVD is the backbone of image compression. If you have a high-resolution photo, it’s a massive matrix of pixel values. Most of those values are redundant. SVD allows you to strip away the "noise" and keep only the most important parts of the image.
$$A = U \Sigma V^T$$
This formula looks scary, but it’s basically just saying that any transformation can be broken down into a rotation, a stretch, and another rotation. That’s it. By focusing on the "stretch" (the singular values), we can decide how much information to throw away. This is how a 10MB RAW photo becomes a 200KB JPEG that still looks great.
Treil’s rigorous treatment of inner product spaces makes these concepts click. When you understand that the "dot product" is just a way to measure how much two things have in common, everything from recommendation engines to quantum mechanics starts to make sense.
Is This Book Right For You?
Honestly, it depends. If you just want to pass a basic intro course and never look at a matrix again, stick to the standard stuff. It’s easier. But if you want to actually understand how the world is built, you need to read Linear Algebra Done Wrong.
📖 Related: Why You Should Make a Picture Video With Music This Way
It’s not just for math majors. Computer scientists need this for graphics and optimization. Engineers need it for stability analysis. Economists need it for modeling complex systems. The beauty of Treil’s writing is that he doesn't separate the "pure" math from the "applied" math. He recognizes that they are the same thing.
You’ve got to be willing to sit with a single page for an hour sometimes. You’ve got to be willing to fail at the exercises. That’s where the learning happens. The "wrong" way is often the only way to get it right.
How to Start Learning Linear Algebra the "Wrong" Way
Don't just dive in and try to read it like a novel. You'll burn out by chapter two.
- Download the PDF. It’s legally free on Sergei Treil's Brown University faculty page.
- Get a notebook. You cannot do this in your head. You need to write out the proofs.
- Focus on Chapter 1 and 2 first. If you don't grasp vector spaces and linear maps, the rest is gibberish.
- Use 3Blue1Brown as a supplement. Grant Sanderson’s "Essence of Linear Algebra" YouTube series provides the visual intuition that perfectly complements Treil’s rigorous proofs.
- Look at the SVD chapter early. Even if you don't understand the math yet, look at what it does. It will give you the motivation to slog through the harder proofs.
Linear algebra is arguably the most important branch of mathematics in the 21st century. It is the engine under the hood of the AI revolution. Whether you use Treil’s book or another rigorous text, the goal is the same: stop memorizing formulas and start seeing the transformations. Stop doing it "right" and start doing it with depth.