THE ECONOMICS OF AGENTS

TOKEN ECONOMICS First, we need to define a token. An LLM, defined in Chapter 2, is the kind of AI model that reads and writes text. An LLM does not read whole words the way you do. It breaks text into small pieces called tokens. A token might be a whole short word like "cat", part of a longer word, a punctuation mark, or a space plus a few letters. As a rough rule of thumb in English, one token is about three quarters of a word, so one thousand tokens is roughly seven hundred fifty words. This is only an approximation, and different models count tokens in slightly different ways. Token economics means understanding how token usage drives cost. Model providers, the companies that run the models and rent them out, usually charge by the token. They typically charge separately for two categories. Input tokens are the tokens you send to the model, such as your instructions, the conversation so far, tool descriptions, and any documents you include. Output tokens are the tokens the model writes back, such as its answer or its request to use a tool. Output tokens usually cost more per token than input tokens, because generating text is more work for the machine than reading it. Prices differ by provider and by model, and they change over time, so a good harness engineer always reads the provider's current price page rather than trusting a number remembered from a book. In this chapter every price is a made-up round number chosen to make the arithmetic easy. Please do not treat them as real prices.