Token Maxxing
In Jiaona Zhang’s usage on Chain of Thought, token maxxing means maximizing visible artificial intelligence (AI) use without showing that it improves the outcome. It names an adoption anti-pattern, rather than a technical method for making a model perform better.
Also known as: token maxing
Enterprise AIAI Evaluation & Reliability
In episode 64, Jiaona Zhang describes people reaching for AI because they have been told to use it, including when adoption is tied to performance reviews. Her example is using AI to redo a deck’s font when a button click would have handled the change. The criticism concerns inefficient use, not the mere presence of tokens in a workflow.
This is informal language, and usage can differ elsewhere. Here it is grounded in Zhang’s argument about measuring AI adoption. A long run may be justified for a difficult task; a short run can still waste effort. Token volume by itself does not decide either case.
The practical question is what the work returned. Compare total time, cost, correction effort and output quality with the previous workflow. Zhang recommends naming the intended outcome before measuring spend. That makes it possible to distinguish learning through useful experiments from activity performed mainly to satisfy a usage target.
Hear it from the guest
“And the measure of success often is are you just doing it versus are you using it efficiently?”
Quotes lightly edited to remove filler words.
Sources
- Chain of Thought, episode 64: Jiaona Zhang on token maxing — Zhang contrasts visible adoption with efficient use; her font-change example appears in the opening guest turn.
Go deeper
From the conversation
-
Stop Token Maxxing: Find Where AI Actually Pays Off | Jiaona Zhang, Laurel -
Your Best AI Engineer Might Have the Worst Metrics | Sonar CTO Andrea Malagodi -
Parenting Your AI Agents for Best Results | Cisco's Jeetu Patel -
250,000 Lines of Code/Week: Inside an AMD VP's Agent-First Workflow | Anush Elangovan