An employee of a fintech unicorn spent $81,000 on AI tokens to create a meme shooter. What went wrong?
A story that is both amusing and serves as a stark warning for the corporate world. The head of strategic verticals at fintech startup Slash, valued at $1.4 billion, "burned" over $81,000 on AI tokens in one week. The reason? He decided to create a meme game using the Claude language model.
Interestingly, the incident itself is not just negligence, but a direct consequence of corporate policy. A week earlier, the company urged employees to more actively use AI for writing code. Nicolas Brianté took this call literally. He spent an entire day developing a shooter game called Brainrot Shooter featuring characters like Skibidi Toilet and Tung Tung Tung Sahur.
An accident worth tens of thousands
According to Brianté himself, he simply "underestimated how quickly AI consumption accumulates." During active development, he repeatedly uploaded the context of the entire codebase. Each request to the model consumes tokens, and over one day the total grew to tens of thousands of dollars. He himself called what happened a "real accident."
The reaction from Slash was unexpectedly positive. Instead of a reprimand, the startup responded humorously on social media, suggesting employees "play the game" so it could be written off as marketing expenses. Moreover, after media coverage, the game unexpectedly found an audience: in the first 48 hours, 6,912 people played it, with total playtime reaching 8,986 hours. The finance department quickly reclassified the project from an "expense incident" to a strategic initiative.
A systemic problem, not an isolated case
Slash is far from the first company to face this issue. Earlier this year, Uber exhausted its annual AI budget in just four months, after which it introduced strict limits. A similar situation occurred with an unnamed company that received a $500 million bill for one month of using Claude from Anthropic — management simply hadn't set limits for employees.
Expert comment: This case clearly demonstrates a "blind spot" in corporate AI adoption strategy. Companies call for innovation but forget the fundamental rule: AI is not a free resource. Without clear limits and monitoring, one employee's enthusiasm can cost more than a whole department's annual budget. The market urgently needs built-in cost control mechanisms at the API level.