How Large Language Models Work Inside
For readers who want to understand the machinery: attention, training, fine-tuning and what happens inside a model. These sources go deep but explain clearly, many with diagrams or code. A little maths and Python helps. If the maths gets too heavy, the research-explained-simply collection covers many of the same ideas in lighter form. Most of these writers also link to free code you can run.
- Blog
Sebastian Raschka
AI and LLM articles, books and research notes by Sebastian Raschka.
Pick 1
Sebastian Raschka
Why we picked it: Articles and books on building LLMs from scratch. Best for readers who want to understand LLMs deeply; expect thorough articles with diagrams and code.
- Newsletter
Language Models & Co.
Jay Alammar on large language models, their internals and applications.
Pick 2
Language Models & Co.
Why we picked it: Illustrated guides to transformer LLMs. Suits visual learners and developers; expect illustrated explainers on how LLMs work.
- Newsletter
Deep (Learning) Focus
Cameron Wolfe explains important topics in AI research.
Pick 3
Deep (Learning) Focus
Why we picked it: Careful explainers of important AI research topics. Best for technical readers; expect long, well-referenced overviews of research topics.
- Blog
Transformer Circuits Thread
Anthropic's interpretability research on reverse-engineering language models.
Pick 4
Transformer Circuits Thread
Why we picked it: Research on reverse-engineering what models compute. For readers ready for technical depth; expect long interactive research articles on model internals.
- Blog
Andrej Karpathy (bear blog)
Andrej Karpathy's current blog, started March 2025.
Pick 5
Andrej Karpathy (bear blog)
Why we picked it: Andrej Karpathy's current writing on LLMs. Suits anyone curious about LLMs; expect occasional, thoughtful essays from a leading AI researcher.
- Newsletter
Exploring Language Models
Maarten Grootendorst's visual guides to how language models work.
Pick 6
Exploring Language Models
Why we picked it: Visual guides to model architectures. Best for visual learners and developers; expect long illustrated guides that build up an idea one figure at a time.
- Newsletter
Interconnects
Nathan Lambert on frontier AI and open models, minus the hype.
Pick 7
Interconnects
Why we picked it: How frontier models are trained, from a researcher who trains them. Best for readers who want insider detail on model training; expect long analyses of open and frontier models.
More Collections
- Start here if you're new to AI8 sources · updated 2026-11-21
- Research explained simply7 sources · updated 2026-11-21
- AI podcasts worth a commute7 sources · updated 2026-11-21
- Best AI newsletters for developers8 sources · updated 2026-11-21