Sebastian Raschka is a gift

I’m reading Sebastian Raschka’s new book, Build a Reasoning Model (From Scratch), with my AI reading group, and it is such a joy.

Like everyone in this space, I’m bombarded with hundreds of articles and newsletters related to AI and ML each week. I do a lot of reading, summarizing with Claude, scanning, and some experimenting, to stay on top of things. But it’s so gratifying to sit down and digest something deep.

Sebastian is my favorite technical writer on this topic. His first book, Build a Large Language Model (From Scratch), is a must-read to me. In particular, the chapter explaining (and coding) the attention mechanism that powers transformers is, and I rarely use this word, a masterpiece.

I first read the attention paper shortly after it came out, in a previous AI reading group, and I remember finding it so confusing. The explanation in the academic paper was just so dense. Then I read his chapter, and it was simple and crystal clear, like attention was the most natural mechanism in the world.

That’s how Sebastian writes: the most complex topics become clear and accessible. The new book is more of the same, and I’m savoring it.