‘Tokenmaxxing’ Is the New Silicon Valley AI Debate

Overlook traces of code written, engineers have a brand new method to compete amongst one another.

Welcome to the period of “tokenmaxxing.”

Armed with shiny new AI coding instruments, software program builders throughout the tech trade have a pockets stuffed with tokens to spend. Tokens, a measure of computing that determines how AI work is priced, have been floated as a type of compensation for engineers, even cropping up in job descriptions for AI fellowships at OpenAI and Anthropic.

However is token spending an excellent measure of developer productiveness?

The query has lit up on social media this week as techies debate the idea of tokenmaxxing after The Information reported that some Meta engineers are racing to spend tokens to rank on an employee-made “Claudeonomics” dashboard that tracks utilization and lets workers compete for titles like “Token Legend.” The corporate did not reply to a request for remark from Enterprise Insider.

Some say it is a useful marker of workers embracing new instruments; others say it might incentivize inefficient use of AI inside corporations — resulting in performative gaming of the metric.

“Rating engineers by token spend is like me rating my advertising workforce by who spent essentially the most cash,” Linear COO Cristina Cordova wrote on X. “Do not mistake a excessive burn price for a excessive success price.”

What’s tokenmaxxing?

To know tokenmaxxing, you first must know what a token is. Giant language fashions break phrases into numerical inputs, treating every token as roughly ¾ of a phrase. AI fashions cost based mostly on the variety of tokens used.

Tokenmaxxing, then, is the drive to spend as many tokens as doable. Meta and OpenAI are simply a number of the tech corporations with token leaderboards, The New York Occasions beforehand reported.

Whereas it is troublesome to measure how widespread tokenmaxxing has change into, corporations’ AI spending is clearly on the rise. The fintech company Ramp known as it a “$1 trillion blind spot” on X, citing Gartner information exhibiting that month-to-month AI spending amongst companies has quadrupled during the last yr.

It is also a flex. Founders and future-forward engineers put up their token spending on X to sign how all-in they’re on AI. One xAI worker wrote that tech was turning each good concept “into theater.”

Y Combinator CEO Garry Tan seems to be a fan. Quoting a prior post that chided corporations which might be “stingy” with tokens, Tan wrote: “We have been tokenmaxxing longer than most individuals.”

Is tokenmaxxing an excellent incentive?

Some throughout the tech world argue tokenmaxxing is an efficient metric; others name it reckless spending.

Khosla Ventures accomplice Jon Chu known as token spending measurement an “completely silly coverage” on X.

“Loads of my Meta buddies advised me people have been constructing bots that simply run in a loop burning tokens as quick as they will resulting from this coverage,” he wrote.

Cursor worker Edwin Wee Arbus supplied a nuanced take on the metric, calling it a “helpful, quick proxy, however barely flawed.” He in contrast it to physique mass index, or BMI, which might present well being insights however doesn’t seize muscle or bone mass.

Whereas Nvidia CEO Jensen Huang hasn’t instantly weighed in on “tokenmaxxing,” he has harassed the significance of engineers using a lot of tokens, saying that if a $500,000 engineer did not devour at the least $250,000 value of tokens, he can be “deeply alarmed.”

“The Pragmatic Engineer” publication creator Gergely Orosz known as the observe wasteful. “Devs sport every little thing and something seen as a goal for extra bonus or promos,” he wrote. “This was no totally different.”

BEP Analysis founder Ben Pouladian pulled a unique takeaway from the pattern, calling compute the bottleneck for innovation. “Within the AI period, each worker turns into a compute shopper,” he wrote on X.

“Token spend is all the time an output not an enter,” wrote Persona software program engineer Arush Shankar, who beforehand labored at Sq. and Microsoft, in response to his LinkedIn. “Price taking a look at, however by no means in isolation. It is a sign however not THE sign.”

Does your organization observe token spending or token use? Contact the reporter from a non-work e mail and system at hchandonnet@businessinsider.com, or on Sign at henrychand.30