What Is an LLM?
A practical mental model of large language models - next-token prediction, training vs inference, and what LLMs are and are not good at for operations work.
LLM Basics
Core mechanics of large language models for platform and infrastructure engineers who need to use them without being misled by them.
17 minutes read
Categories:
Tokens, Context Windows, and Limits
What tokens are, how context windows bound every LLM interaction, and what that means for how much infrastructure state you can feed an LLM at once.
Strengths and Failure Modes
Hallucination, confident wrongness, stale knowledge, and non-determinism - why verification and human-in-the-loop are non-negotiable for infrastructure automation.
Choosing a Model for Infrastructure Tasks
Capability vs latency vs cost, when a smaller model suffices, and how to match model to task - classification, generation, and reasoning.