‘Tokenmaxxing’ is Silicon Valley’s new AI controversy

Forget about written lines of code, engineers have new ways to compete with each other.

Welcome to the era of “tokenmaxxing”.

Software developers across the tech industry are armed with shiny new AI coding tools and wallets full of tokens to spend. Tokens, a measure of computing that determines the price of AI work, have emerged as a form of compensation for engineers, appearing in job descriptions for AI fellowships at OpenAI and Anthropic.

But is token spend an appropriate measure of developer productivity?

The question surfaced on social media this week as technologists discussed the concept of token maxing after The Information reported that some meta-engineers were competing to spend tokens to track usage and rank on employee-created “Claudeonomics” dashboards where employees could compete for titles such as “Token Legend.” The company did not respond to Business Insider’s request for comment.

Some say this is an indicator of whether employees are adopting new tools. Some say it could encourage inefficient use of AI within companies and lead to performative gaming of metrics.

“Ranking engineers by token spending is like ranking marketing teams by who spends the most money,” Linear Chief Operating Officer Cristina Cordova wrote in X. “Don’t mistake a high burn rate for a high success rate.”

What is Token Max Thing?

To understand tokenmaxxing, you first need to know what a token is. Large language models split words into numeric inputs, treating each token as approximately 3/4 of a word. The AI ​​model charges fees based on the number of tokens used.

Tokenmaxxing is therefore the willingness to spend as many tokens as possible. Meta and OpenAI are just a few of the tech companies with token leaderboards, The New York Times previously reported.

While it’s difficult to measure how popular tokenmaxxing has become, enterprise AI spending is clearly on the rise. Fintech company Lamp Inc. cited Gartner data showing that monthly AI spending among companies has quadrupled in the last year, calling it X’s “$1 trillion blind spot.”

It’s also flex. Founders and forward-looking engineers post their token spending on X to show how committed they are to AI. One xAI employee wrote that technology is turning every good idea into a “theater.”

Y Combinator CEO Garry Tan also seems to be a fan. “We have been token maxing longer than most people,” Tan wrote, citing a previous post in which he scolded companies for being “stingy” with their tokens.

Is tokenmaxxing a good incentive?

Some in the tech industry claim that tokenmaxxing is an effective metric. Some call it reckless spending.

Jon Chu, a partner at Khosla Ventures, said measuring token spending is “an absolutely stupid policy” for X.

“Many meta friends of mine have told me that because of this policy, they are building bots that just burn tokens as quickly as possible in the loop,” he wrote.

Edwin Wee Arbus, an employee at Cursor, offered a more nuanced view of the metric, saying, “It’s a useful and fast proxy, but it’s a little flawed.” He compared this to body mass index (BMI). BMI provides insight into your health, but it doesn’t provide insight into your muscle or bone mass.

While Nvidia CEO Jensen Huang did not directly address “token maxing,” he emphasized the importance of engineers using large amounts of tokens, saying he would be “very concerned” if a $500,000 engineer did not spend at least $250,000 worth of tokens.

Gergely Orosz, author of “The Pragmatic Engineer” newsletter, said this practice is wasteful. “Developers gamify everything to get more bonuses and promotions,” he writes. “This was no different.”

Ben Pouladian, founder of BEP Research, disagreed with the trend by saying that computing is the bottleneck to innovation. “In the AI ​​era, every employee will become a consumer of computing,” he wrote in X.

“Token spending is always an output, not an input,” writes Persona software engineer Arush Shankar, who previously worked at Square and Microsoft, according to LinkedIn. “It’s worth noting, but by no means alone. It’s a signal, but it’s not a complete signal.”

Does your company track token spending or token usage? Contact reporters from non-work emails and devices. hchandonnet@businessinsider.comor Signal by henrychand.30