Curated by Priyanshu Singh
What this is: A carefully filtered collection of the best articles, papers, books, and videos for understanding how LLMs and transformers work from the inside — the architecture, the inference mechanics, the training pipelines, and how the field has evolved. Every entry was evaluated individually. Many were left out.
Who this is for: Engineers, researchers, and serious learners who want to go beyond using LLMs to actually understanding them.
What this is not: Prompt engineering, agent design, LLMOps, or application-layer tutorials. The focus is the internals.
⭐ Essential — foundational, high-signal, read regardless of experience level
✅ Excellent — strong depth, worth your time once you have the foundations
The core mechanism — start here
This is where every serious study of LLM internals begins. The transformer architecture, introduced in 2017, is the foundation of every model discussed in the full guide. These four resources approach it from different angles — visual, geometric, historical, and intuitive — and together they give a complete picture of what attention is and why it matters.
Jay Alammar · Article
⭐ Essential · Start here
This is the most widely used visual introduction to the transformer architecture in existence. Alammar walks through the encoder-decoder structure, multi-head attention, and positional encoding with custom illustrations that make the flow of information through the model genuinely clear. It has been translated into over a dozen languages and cited by learners across every level of the field. The reason it holds up years later is that it explains the architecture as a system — not just the math, but the shape of how data moves. If you read only one thing from this entire list first, make it this.