• Cake is a Rust framework for distributed inference of large models like LLama3 based on Candle.
• YouTube / @TheOfficialACM / Understanding the LLM Development Cycle: Building, Training, and Finetuning
• middleware.io / What is LLM Observability? By Vivek Tilva
• Hackaday.com / An Animated Walkthrough Of How Large Language Models Work
• A ChatGPT clone, in 3000 bytes of C, backed by GPT-2 by Nicholas Carlini
  • Хабр / Клон ChatGPT в 3000 байтах на C, основанный на GPT-2
• Хабр / Google представили Titan: архитектуру нейросетей, которая может стать новой серебряной пулей LLM
  • Все современные LLM построены на архитектуре трансформера. GPT-4o от OpenAI, Gemini от Google, Claude Sonet от Anthropic, Grok от xAI...
• Ollama
  • ollama/ollama Get up and running with Llama 3.3, DeepSeek-R1, Phi-4, Gemma 3, Mistral Small 3.1 and other large language models.
• darrenburns/elia A snappy, keyboard-centric terminal user interface for interacting with large language models. Chat with ChatGPT, Claude, Llama 3, Phi 3, Mistral, Gemma and more.

MCP (Model Context Protocol)

• HuggingFace.com / Learn / MCP Course
• The GitHub Blog / What the heck is MCP and why is everyone talking about it? TL;DR: It’s an open standard for connecting LLMs to data and tools.
• GitHub / tadata-org/fastapi_mcp A zero-configuration tool for automatically exposing FastAPI endpoints as Model Context Protocol (MCP) tools.
  • Turn any FastAPI app into an MCP server!
• KDnuggets.com / Building A Simple MCP Server Give your LLMs the extra ability to fetch live stock prices, compare them, and provide historical analysis by implementation tools within the MCP Server.

From scratch, reinventing a wheel

• freeCodeCamp.org
  • Train Your Own LLM
  • Understand and Code DeepSeek V3 DeepSeek V3 is a popular Large-Language Model that's a good one to learn if you're interested in AI. And in this course, you'll learn how it works. It covers DeepSeek's Multi-Head Latent Attention mechanism, Mixture of Experts architecture, and lots more.
• YouTube / Build a Large Language Model (From Scratch) by Sebastian Raschka
• TowardsDataScience.com / Training LLM, from Scratch, in Rust
• Medium / Towards Data Science / How to Build Your Own AI: Creating an LLM from Scratch 🤯 by Leo Anello
• GitHub
  • naklecha/llama3-from-scratch llama3 implementation one matrix multiplication at a time
    • Someone implemented Llama 3 from scratch in Python. Explained step-by-step, one matrix multiplication at a time.
  • jaymody/picoGPT An unnecessarily tiny implementation of GPT-2 in NumPy.
    • This tutorial walks you through how to implement GPT from scratch in just 60 lines of numpy. You will then load the trained GPT model weights released by OpenAI into the implementation and generate some text.
    • GPT in 60 Lines of NumPy by Jay Mody

Evaluation, metrics, benchmarks

• SWE-bench Can Language Models Resolve Real-World GitHub Issues?
• SuperAnnotate.com / LLM Evaluation: Metrics, Frameworks, and Best Practices July 18, 2024
• ITC.ua / Математики разработали сложные задачи для проверки мышления Gemini, Claude и GPT-4o — они провалили почти все тесты 20.11.2024
• Confident-AI.com / Evaluating LLM Systems: Essential Metrics, Benchmarks, and Best Practices April 22, 2025
  • LLMARENA / Оценка систем LLM: основные метрики, бенчмарки и лучшие практики
• Хабр / Языковые модели ИИ тестируют на способность рассуждать в игре «Phoenix Wright: Ace Attorney» 27 апр 2025

Courses

• HuggingFace.com / Learn / LLM Course
• TowardsDataScience.com / The Large Language Model Course
• DeepLearning.AI / Getting Structured LLM Output
  • Andrew Ng