Tokens Should Have Jobs — Katelyn Lesse & Angela Jiang, Anthropic
Source
youtube.com
Author
AI Engineer
Date
Context
An AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition →'s text budget can fund checking as well as doing. In this Anthropictalk, an executor can consult a separate adviser during a task, or submit its work to a grader that applies a predefined rubric and requests another attempt. These arrangements change how the model's text processing is allocated without necessarily increasing the total budget.
That changes what efficiency means. A financial-analysis answer with incorrect figures may require another run, so the cost of an acceptable answer includes unsuccessful attempts. The proposed comparison is between strategies that reach a usable result, rather than simply between the amounts of text generated in one attempt.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
token budget — A cap on how many tokens a task, session, or agent run may consume — the practical control on both cost and how long an agent will grind.
benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
videoHow to Write Your First AI Eval - SallyAnn DeLucia, Arize AI
videoBuild Your Own Team of Managed Agents - Jack Wotherspoon, Google DeepMind
videoHow to Build an AI Code Review Agent - Daksh Gupta, Greptile
videoHow to Build Voice Agents That Don't Fall Apart When You Interrupt - Jesse Hall, LiveKit
