LLM Guts

I am not bringing everything I have here partially because that is a lot to comb through and partially because a lot of it is focused on the inference infrastructure side of things as well inference user experience and strategies which is tangential to this but not clearly a part of it.

Collection of resources someone else made I approve of https://read.theaimerge.com/p/a-practical-roadmap-on-llm-systems

First some blogs I think reach a nice technical depth level. Some of these you gotta search for the in depth concepts underlying llm workings but they're there. https://hamzaelshafie.bearblog.dev/blog/https://huyenchip.com/blog/https://rajatpandit.com/ai-engineering/https://blog.ngxson.com/https://www.dbreunig.com/writing.htmlhttps://themlsurgeon.substack.com/archive?sort=newhttps://simonwillison.net/tags/llms/https://mbrenndoerfer.com/writing/categories/llm-genaihttps://newsletter.maartengrootendorst.com/archive?sort=new

More hands on for the reader https://xsxszab.github.io/posts/https://read.theaimerge.com/archivehttps://leimao.github.io/blog/

Here are some resources I am cheating to include that a bit more tangential which I think really still fit the convo for expanding how you see this stuff https://eugeneyan.com/writing/https://buttondown.com/ultradune/archive/https://huggingface.co/blog/mike-ravkine/can-ai-stop-thinkinghttps://bentoml.com/llm/https://github.com/p-e-w/heretichttps://www.micahlerner.com/https://eunomia.dev/blog/2025/02/18/os-level-challenges-in-llm-inference-and-optimizations/ also check out lavd with this from steam and meta

I don't read the papers often but if you are going to read any, this is a good collection https://huggingface.co/collections/nityan/mustread-papers

Seriously dense content but not paper level dense and very comprehensive https://aman.ai/

Stanford course material links https://www.reddit.com/r/LocalLLaMA/comments/1oakwgs/stanford_just_dropped_55hrs_worth_of_lectures_on/

The acronyms will never stop https://www.vectara.com/glossary-of-llm-terms

Individual articles where I'm not sure the full blog fits https://thinkingmachines.ai/blog/defeating-nondeterminism-in-llm-inference/https://bhavishyapandit9.substack.com/p/reward-models-in-llms

A few de-hype pieces here that I really think fits when talking about gen ai with people, easy to get sucked into the hype and awe as you dive because of how amazing the fact we solved an of these problems is. One on using them https://www.lesswrong.com/posts/nR3DkyivzF4ve97oM/how-go-players-disempower-themselves-to-ai One of how the current commercial and research incentive is more on advancement than bounding. https://www.lesswrong.com/posts/WewsByywWNhX9rtwi/current-ais-seem-pretty-misaligned-to-me

I am going to call it here for now.