SAN FRANCISCO — One employee at a fintech startup burned through $81,267 in artificial-intelligence computing costs in a single week. He was building a video game based on internet memes. The tab landed on his employer.
Nicolas Brillante, head of strategic verticals at the San Francisco-based firm Slash, used Anthropic’s coding assistant Claude to build what he called “Brainrot Shooter.” The game drew on viral memes including “Tung Tung Tung Sahur” and “Skibidi Toilet.” The project was part of a company push to get staff using AI coding tools to speed up software development. Each time Brillante asked Claude to analyze and then update large blocks of code, the AI token count climbed.
By the end of the week, the company’s AI bill hit $81,267. Brillante posted a screenshot of his usage dashboard on X. He called it an honest mistake.
“This was a genuine accident, I underestimated my own ability,” he wrote. In a subsequent post, replying to a Polymarket post on X, he added: “This is actually insane, am I going to become a case study for how AI spend can get out of control.” The incident has ignited an online discussion about the hidden costs of AI-powered development.
What began as a harmless experiment in productivity turned into a costly lesson for the firm. Sources familiar with the matter note that the episode underscores the need for businesses to enforce stronger spending controls and limits on employee use of advanced coding tools. Here is the problem.
AI coding assistants like Claude charge by the token — basically, by the chunk of text they process. A developer asking the tool to analyze an entire codebase, then rewrite it, racks up tokens fast. The cost is invisible until the bill arrives.
Slash is a fintech company. That means it handles money, or software for handling money.
The company pushed employees to use AI tools to accelerate development. No one set spending limits. No one flagged that a single employee could spend $81,000 building a meme game.
The stakes here are not small. Companies across the tech industry are racing to integrate AI assistants into their workflows.
They tout productivity gains. They talk about faster coding cycles. They rarely talk about the cost structure, which is opaque and easy to ignore until someone runs up a bill equal to a year’s salary for a junior developer.
For Slash, $81,267 is a real loss. The company paid for computing time that produced a meme game. The game may have no commercial value.
The money is gone. For the broader industry, the episode raises a practical question.
If a fintech company cannot control AI spending by a single employee on a lark, how many other companies are bleeding cash the same way? How many finance departments will get a surprise bill for five or six figures and have no policy in place to stop it? Brillante’s own words suggest he knows he is now a cautionary tale.
“Am I going to become a case study for how AI spend can get out of control,” he asked. The answer appears to be yes.
Artificial intelligence is transforming how developers write software. That is not in dispute. But one fintech employee learned that faster coding and greater productivity come with a hefty price tag after unintentionally spending more than $81,000 on a single project.
The company paid for that lesson.
