The first two posts (A Real-World Case Study and How Prompt Caching Actually Works) in this series were descriptive: how tokens are counted, where they actually go in a real $172.58 Claude Code session, what the prompt cache is doing on the … Continue reading →
The first post in this series, Part 1: A Real-World Case Study, ended with a single number: a 31-hour Claude Code session that cost $172.58, of which $114.98 — about 66% of the bill — was cache reads. Caching wasn’t a side effect … Continue reading →
Most developers who use a large language model — through an API, through an agent framework, or through a product like Claude Code, ChatGPT, or Gemini — have only a fuzzy sense of what they’re actually paying for. The dashboard … Continue reading →
For most people, artificial intelligence means sending your questions to a distant server — paying for a subscription, agreeing to terms of service, and hoping your data stays private. That’s the world ChatGPT, Gemini, and Claude live in. But just … Continue reading →
Introduction AI assistants have become a staple in most developers’ workflows — but switching between your terminal and a browser chat window breaks focus and slows you down. What if your AI assistant just lived in the terminal, where you … Continue reading →