Why Companies Keep Tokenmaxxing - by Amber Case

Cayce’s Substack

SubscribeSign in

Why Companies Keep Tokenmaxxing<br>The Hidden Cost of Losing Touch

Amber Case<br>May 29, 2026

Share

Uber burned through its entire 2026 AI budget in four months.<br>A recent headline had me thinking even more about what it means to work, do work that looks like working, and to rush to keep up with a treadmill of interaction that feels like something real but doesn’t produce what it promises.<br>Cayce’s Substack is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Subscribe

With many companies falling victim to “tokenmaxxing”, blowing out their programmers’ allotted budget, it’s past time to consider why this phenomenon keeps happening.<br>The typical response is to blame the declining quality of LLM models that power platforms like Claude and ChatGPT. But I believe there’s a far less appreciated part of the problem: The declining quality of human input.<br>For instance, consider the cognitive richness that is lost when we enter prompts through our computing devices:<br>When you’re typing something or you’re touching a glass screen with your thumbs, all of that flatness isn’t coming back to you with anything unique. By contrast, when you write in longhand, every single letter you write has its own shape. And that allows your brain to record those memories and stimulate those memory centers of the brain better because each is unique.<br>The problem is worsened by LLMs being designed to keep us in our seats, typing out the next prompt. As we get deeper and deeper into a chat with an LLM, it starts over-indexing on what you first input, and doesn’t go broader than that. We can get down into a thread and wind up working on the same thing we started with, again and again. We’re not being productive, but getting further saturated in our original task.<br>That’s often what is really happening when people say they are tokenmaxxing.<br>For us to think and thus prompt better, we need to get out of our blind spots and local minima.<br>Whenever possible, we need to physically leave our seats and engage with the fractals of nature. Studies have consistently shown how valuable this is for our brain activity.<br>People who want to work better with digital thinking tools and consider taking a break outside, preferably amid the fractals of nature, and start thinking about the actual nature of the problem, not just typing something in and getting maddened by the output.<br>While that may seem old school in Silicon Valley, it’s been proven that writing on paper stimulates the part of your brain that helps you store memories, and begin to create a base of knowledge from which to write down fully contextualized but highly succinct prompts.<br>“The quality of your life,” as a colleague once told me, “is based on the quality of the questions that you ask.” Sitting at a screen isn’t always the way to find the right questions.<br>Cayce’s Substack is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Subscribe

Share

Discussion about this post<br>CommentsRestacks

TopLatest

No posts

Ready for more?

Subscribe

© 2026 Cayce · Privacy ∙ Terms ∙ Collection notice<br>Start your SubstackGet the app

Substack is the home for great culture

This site requires JavaScript to run correctly. Please turn on JavaScript or unblock scripts