Anthropic and the Tokenmaxxing Debate: Optimism and Caution

At the Bloomberg Tech conference held in San Francisco, Daniela Amodei, president of Anthropic, spoke on the controversial topic of tokenmaxxing. This phenomenon concerns companies that are heavily investing in artificial intelligence, accumulating high bills without always achieving clearly defined returns on investment. Amodei responded to a question on this subject, taking a diplomatic approach while displaying a certain optimism about the future of AI models.

She emphasized that, although AI models have seen significant improvements over the past two years, they have not yet reached their full potential. "I think there is still a long way to go for what models will be capable of doing in two to four, six to eight years," she stated, expressing her belief that the future capabilities of AI models will be much more advanced.

Amodei also discussed the evolution of companies in their use of AI tools, predicting that they will discover new ways to integrate them into their operations. She acknowledged that workers must adapt to a new paradigm where the use of AI is encouraged, and sometimes even monitored by their employers. "Today, there is this feeling that is like, 'Oh, AI, you know, the rankings, and it's like I have to use it, and what am I going to use it for?'" she explained. She hopes that, over time, AI will integrate more naturally into professional daily life, bringing tangible added value.

Regarding Anthropic, Amodei clarified that the company does not have a ranking imposing the use of tokens, although it monitors the general usage of its products like Claude. "But there is not like, 'You must use AI and you must use Claude,'" she added. Claude Code, Anthropic's flagship product for coding, is a key player in the tokenmaxxing debate. This product uses significantly more tokens than traditional chatbot interfaces, especially with the advent of autonomous agents. Anthropic and OpenAI charge partly on a per-token basis, which means that the rise of AI coding has contributed to an increase in revenue.

Companies are navigating an environment where managing AI spending has become crucial. Some large companies have encouraged uninhibited use of AI, but initiatives have been taken to moderate this trend. Amazon, for example, recently shut down an informal ranking created by employees to track AI token usage. Dave Treadwell, senior vice president of Amazon, advised staff not to use AI without a valid reason. Similarly, at Meta, a similar ranking titled "Claudeonomics" was removed by the employee who created it, according to reports from The Information.