Tokenmaxxing is more than just a trend. This is honest advice from Silicon Valley’s leading startup accelerator.
In a new episode of Y Combinator’s “Startup School,” partner Diana Hu teaches founders how to build AI-native companies. Hu himself is a YC-backed founder who founded the augmented reality company Escher Reality, which was acquired by Niantic.
“The key change will be to maximize token usage, not headcount,” Hu said. “The best companies will be the ones doing token maxing.”
The token measures the cost of AI computing. The more tokens spent, the more an individual employee or developer used an AI tool. (Importantly, more tokens does not necessarily mean greater effect.) Some companies build token leaderboards or encourage token maxing, the process of consuming as many tokens as possible.
Business Insider asked startup leaders about the trends. Some say tokenmaxxing is easy. Some said their size was meaningless.
Hu, like his boss Gary Tan, is an unabashed advocate. She described a “trade-off” between labor and token spending.
“One person who can use an AI tool could be the equivalent of a large engineering team in a pre-AI company,” Hu said. “This means our engineering, design, human resources and management teams will be dramatically leaner.”
Startup founders “have to run obnoxiously high API bills because they’re replacing costs that would have been much more expensive and headcount inflated,” Hu said.
While Y Combinator’s advice for startups may not apply to large companies (though there are plenty of voices in venture capital and the tech industry to argue that it should), this explainer video provides an interesting look at the business values being instilled in the next generation of up-and-coming CEOs.
President Hu also expressed support for a three-pillar employee base. There are individual contributors (who build things), individuals with direct responsibility (who focus on strategy), and AI founders (who lead while they build).
This is a similar structure to Jack Dorsey’s redesign of his payment processor, Block. After laying off around 40% of his staff, Mr Dorsey unveiled a new three-pronged structure to transform the block into a “mini-AGI”.
Founders should also try out the tools themselves. Hu said leaders should not “outsource” their belief in these AI tools.
“You have to use them and develop it yourself until you actually sit down with a coding agent and start breaking out your own prior knowledge of what you can build,” Hu said.