@LinusEkenstam: ROT — Return on Tokens. We all knew we would end up here at some point. Tokenmaxxing was a dumb idea from the start. It…
Summary
The author critiques the trend of token maximization in LLM usage and argues for a shift toward Return on Tokens (ROT) through optimization and routing for sustainable AI deployment.
View Cached Full Text
Cached at: 06/29/26, 08:32 AM
ROT — Return on Tokens.
We all knew we would end up here at some point.
Tokenmaxxing was a dumb idea from the start. It’s just that when everyone is drinking from the coolaid it’s hard to resist.
But for some of us it’s been clear as day for years that first we will see unhinged growth no matter the cost.
Because nobody will want to be left behind. So we throw everything/everyone/every dollar at the table.
Leaderboard with top token spenders as if it was a direct correlation to output and value. When in reality it’s a spend leaderboard and nothing else.
in 2022, a few of us openly debated around how long it would take before we would start to see token optimization become a real thing, because in the early days you had to, simply because context was short, so you optimized everything everywhere. Then all of a sudden, price kept coming down, and context kept going up.
We stopped optimizing and started Maximizing. Because why hold back.
Well, this is where @markiewagner has been able to do a fantastic job and her dept on the subject is zero to none. Her essay is linked below.
But I want to add, ROT, return on token, is not new, it’s simply becoming the right thing to do, as It should have been from the very start.
But we humans are simple, we like casinos, we like to get delusional together and feed the beast.
You don’t employmaxx and wait years to figure out that ohh, hiring more people infinitely is not equal to success, you need to have return on your hires too.
In this guilded age, it seems only natural that we then decide to move towards a more sustainable future outlook on how to operate token usage.
when we back in 2022 described the future systems, we all had strong consensus around routing.
Being able to route well is a key lever to good RUT. While dropping usage is the simplest, routing might be the hardest. Because of its near infinite possible outcomes.
I’m happy we’ve come full circle on this. Excited about the future of actually building out behavior and systems that focuses less on maximizing spend, as if we just won the lottery, and instead maximize return as if we just reinvented capitalism
Similar Articles
Tokenmaxxing is dead, long live Tokenmaxxing
An analysis of the 'tokenmaxxing' phenomenon at companies like Meta, arguing that executives intentionally encouraged wasteful AI token usage to drive tool adoption, contrary to the perception of accidental mismanagement.
@sdianahu: tokenmaxxing isn't "spend more on tokens" it's the opposite tokenmaxxing = picking the right stat to max, then making e…
A tweet explains that 'tokenmaxxing' is about optimizing for the right metric while minimizing costs, leveraging the declining cost of intelligence, and suggests taste is the scarce input.
Tokenmaxing is out - Frugal AI is the new trend
The era of tokenmaxing (unlimited AI token usage) is ending as companies face high costs and ecological damage, giving way to tokenminimizing—a focus on efficiency and choosing the right AI model for tasks.
@HaydnBelfield: The main benefit of hearing about tokenmaxxing experiments and token leaderboards seems to be an inspirational/explorat…
Haydn Belfield discusses how tokenmaxxing experiments and token leaderboards serve an inspirational and exploratory purpose by testing AI model limits and discovering new workflows.
@dabit3: Tokenmaxxing is dead. Everyone's realized that token usage is a horrible way to measure productivity. So where do we go…
Discusses the shift from token-based productivity metrics to output, impact, and value measurement in AI adoption, highlighting Cognition's solutions: adaptive routing, spend attribution, automations, and a productivity guarantee.