Mission: Understand open-source LLMs

Mission: Understand open-source LLMs

Why

Build a reliable mental model of how open-source large language models turn a prompt into useful text. The goal is to be able to evaluate explanations and technical claims about LLMs without treating them as magic.

Success looks like

• Explain, in plain language, how tokenization, transformer attention, and next-token prediction fit together.
• Distinguish model training from inference and describe what a model's weights are.
• Trace the main stages by which an LLM is trained: data preparation, pretraining, evaluation, and post-training.
• Read a basic open-source model guide and recognize its core terminology.

Constraints

• Start from first principles; assume no prior machine-learning background.
• Use short lessons with retrieval practice and high-trust primary sources.

Out of scope

• Training or operating a production-scale model, deployment, and detailed linear algebra for now.