Anthropic’s Daniela – At a Bloomberg Tech conference in San Francisco, Anthropic co-founder Daniela Amodei was asked about tokenmaxxing—developers using as much AI as possible to rack up large bills with unclear business payoffs. She said models still have “a lot more distance to g
For the third morning of Silicon Valley debate, the question still came dressed like a dashboard: tokenmaxxing.
At Thursday’s Bloomberg Tech conference in San Francisco. Anthropic president and co-founder Daniela Amodei faced a room that has grown accustomed to ranking everything. The prompt was familiar—developers pushing as much AI usage as possible, driving huge bills with unclear business payoffs. Amodei answered with a mix of diplomacy and a sharper belief that the industry is asking the wrong question.
“I actually think there’s a lot more distance to go still for what the models will be able to do two to four, six to eight years in the future,” Amodei said.
Her point wasn’t that today’s tools are meaningless. It was that the current obsession—especially the culture of leaderboards—isn’t the end of the story. She said businesses will find new ways to use AI, and that teams will “learn together as a larger community.”
But she also acknowledged the human shift already underway: workers are reckoning with a new work paradigm where employers may direct employees to use AI and sometimes even track how much they do.
“There’s this feeling that’s like. ‘Oh. like AI. you know. the leaderboards. and it’s like I have to use it. and what am I going to use it for?’” Amodei said. “My hope is that over time it’ll be more incorporated into the day-to-day of how humans do our work. how we communicate together. and that there will actually be a lot more value realized in a way that feels really good to people.”.
Amodei’s stance on the leaderboard mentality was concrete. She said Anthropic does not have a token-usage leaderboard. The company does track general use of its Claude products as teams work, she said. Still, she drew a clear line between optional integration and enforced usage.
“But there’s not like, ‘You must use AI and you must use Claude,’” Amodei added.
That distinction matters because Claude Code—the company’s AI-coding product—is one of the main drivers of the tokenmaxxing discourse. The tool uses significantly more tokens—“the units of AI input and output”—than a classic chatbot interface. The gap grew more substantial with the advent of agents, which a user can direct to autonomously churn through tasks.
And as token consumption rose, so did revenue pressure. Anthropic and OpenAI partly charge on a per-token basis, and the boom in AI coding has fueled a surge in revenue.
In parallel, executives at major companies have been trying to control the spending and behavior that leaderboards can encourage. One recent example came from Amazon, which shut down an informal, employee-made leaderboard tracking the use of AI tokens.
“Please don’t use AI just for the sake of using AI,” Dave Treadwell, an Amazon senior vice president, told staff.
Similar dynamics played out elsewhere. A leaderboard at Meta was also taken down by the employee who made it, and it was titled “Claudeonomics,” as The Information reported.
Amodei’s remarks put Anthropic in a familiar place in the debate: not denying the economic logic of token-based pricing, but resisting the cultural logic of treating AI usage like a game to win.
Anthropic Daniela Amodei tokenmaxxing Claude Claude Code AI pricing per-token billing agentic AI Amazon leaderboard Meta Claudeonomics Bloomberg Tech conference