# Mission: Understanding Large Language Models (LLMs)

## Why
I am an AI application developer with programming skills (Python, APIs) but no formal ML background. I want to understand how LLMs work internally so I can build better AI applications — making smarter architectural decisions, writing more effective prompts, debugging unexpected model behavior, and knowing when an LLM is the right tool for a job versus when it isn't.

## Success looks like
- Can explain the core mechanism of how an LLM processes input and generates output, at a level sufficient to reason about application behavior
- Understands the practical implications of tokenization, context windows, attention, and temperature for application design
- Can read and evaluate AI research announcements and technical blog posts without getting lost
- Can make informed decisions about model selection, fine-tuning vs. prompting, and RAG architectures
- Can anticipate failure modes of LLMs (hallucination, reasoning errors) from architectural understanding

## Constraints
- No formal ML/math background — need concepts built up from programming intuitions, not equations
- Learning in parallel with building AI applications — lessons should connect to real development scenarios
- Chinese is preferred for explanations, with English technical terms alongside

## Out of scope
- Implementing a transformer from scratch (not needed for application development)
- Deep mathematical proofs of convergence or optimization theory
- Training infrastructure, distributed computing, or MLOps
- Computer vision or multimodal models (focus on text LLMs first)