How Large Language Models Work
Open the black box of large language models: how they predict text, why they sometimes make things up, and the handful of settings that change how they answer.
System prompts and fine-tuning at a glance
Three ways to shape what a model does Once you understand prediction and temperature, the last piece is how you steer a model toward your particular needs. There are three common levers, from lightest to heaviest. 1. The system prompt A system prompt is a set of instructions placed above the conversation that sets the […]
What the model can and cannot see
Everything the model knows right now is in one buffer An LLM has no memory between requests. Everything it can consider — the system prompt, the conversation so far, anything you pasted, and its own reply as it forms — occupies a single fixed buffer called the context window. Nothing outside it exists for this […]
Cost, latency, and picking a model
Every provider offers a range, and the biggest is rarely the right default Models come in tiers — small and fast, mid-range, and large and capable — and the price between the ends of that range is often an order of magnitude. Reaching for the largest by reflex is the most common way to spend […]
Tools, functions, and agents
Giving a text predictor a way to act A model that only produces text cannot look anything up, do arithmetic reliably, or change anything in your systems. Tool use — also called function calling — closes that gap, and it is the single idea behind most of what gets marketed as an “AI agent”. How […]
Reading a model announcement without being sold to
New models ship constantly. Most announcements will not change your work Keeping up is exhausting and mostly unnecessary. What you need is a filter that turns a launch post into a decision in five minutes. The four things that actually matter to you Price per token, input and output. This is the number most likely […]