A curated, non-chronological guide to the articles I would recommend first. If the blog archive is the full library, this page is the shelf of strongest starting points by topic.

LLM architectures From scratch Reasoning models PyTorch Beginner-friendly Glossary
Best LLM Architecture Articles

Start here if you want to compare modern open-weight models, attention variants, mixture-of-experts designs, and practical architecture tradeoffs.

HERO The Big LLM Architecture Comparison The main long-form comparison of recent LLM architecture choices. 51d52b9a E820 45d6 F94496ec1745 1600x900 A Visual Guide to Attention Variants in Modern LLMs A focused guide to MHA, GQA, MLA, sparse attention, and hybrid attention. Architecture Gallery HERO LLM Architecture Gallery A visual reference collection of architecture figures and compact fact sheets. 43071df5 B297 42f2 B875 36915f5e33a2 1826x1020 My Workflow for Understanding LLM Architectures How to inspect open-weight model configs and implementation details. HERO From GPT-2 to gpt-oss A bridge from classic GPT-style designs to newer open-weight releases. HERO From DeepSeek V3 to V3.2 A technical look at DeepSeek architecture, sparse attention, and RL updates.

See all LLM articles →

Best From-Scratch Implementation Articles

These are the most useful articles if your preferred way to understand a method is to implement it and inspect the moving parts.

HERO Understanding and Coding the KV Cache in LLMs from Scratch A practical explanation of one of the key mechanisms behind efficient inference. HERO Self-Attention from Scratch A step-by-step implementation of the core mechanism behind transformer LLMs. BPE Overview Byte Pair Encoding Tokenizer from Scratch A standalone implementation of the tokenization algorithm used by many LLMs. HERO Understanding and Implementing Qwen3 from Scratch A detailed implementation-focused look at a modern open-weight LLM family. HERO LoRA and DoRA from Scratch A code-oriented explanation of parameter-efficient adaptation methods. HERO The 4 Main Approaches to LLM Evaluation from Scratch Benchmarks, verifiers, leaderboards, and LLM judges with concrete examples.

See all LLM articles →

Best Reasoning Model Articles

Use this path for reinforcement learning for LLM training, inference scaling, distillation, and other advanced reasoning-model methods.

HERO Understanding Reasoning LLMs The high-level map of reasoning model strategies and where they fit. HERO Categories of Inference-Time Scaling A guide to using more test-time compute to improve LLM reasoning. HERO The State of Reinforcement Learning for LLM Reasoning GRPO, reinforcement learning, and post-training ideas for reasoning models. HERO Inference-Time Compute Scaling Methods A deeper look at inference-time compute methods for reasoning-optimized LLMs. HERO First Look at Reasoning From Scratch A bridge from the conceptual articles into the code-first reasoning book.
Best Practical PyTorch Articles

These articles and tutorials are useful when you want concrete PyTorch practice, training efficiency, and implementation details.

HERO PyTorch in One Hour A compact tutorial for getting productive with tensors, training loops, and GPUs. HERO Optimizing Memory Usage in PyTorch Practical memory-saving methods for LLMs and vision transformers. HERO Make Your PyTorch Models Train Faster A practical tour of performance improvements without changing model accuracy. HERO Mixed-Precision Techniques for LLMs A code-backed explanation of float16, bfloat16, and training speedups. HERO Finetuning Falcon LLMs with LoRA and Adapters A practical finetuning comparison for parameter-efficient methods. HERO Single-GPU LLM Finetuning with Gradient Accumulation A practical workaround for larger effective batch sizes when GPU memory is tight.

See all PyTorch articles →

Best Beginner-Friendly Articles

Start here if you want broader context before jumping into longer technical walkthroughs or code-heavy implementations.

HERO Understanding Large Language Models: A Reading List A guided set of papers and resources for building a first mental model of LLMs. HERO Developing an LLM: Building, Training, Finetuning A broad overview of the LLM development lifecycle. HERO New LLM Pre-training and Post-training Paradigms A readable tour of how modern LLM training has changed. HERO Keeping Up With AI Research and News A practical workflow for navigating the flood of AI papers and updates. HERO Getting the Most Out of a Technical Book Study advice for turning technical reading into working knowledge. 49b97718 57f4 99c8 8ad5c4d32af3 1548x862 Components of a Coding Agent A practical entry point into tools, memory, repo context, and coding agents.