A startup engineer’s side project has become a cautionary tale about uncontrolled AI spending. What began as a simple game-building experiment reportedly turned into an AI bill of more than $80,000 in just one week.
The case came from Slash, a San Francisco fintech startup, which said one of its employees used far more AI credits than expected while making an online game called Brainrot Shooter. The company even joked on X that the game should be played enough to count the expense as marketing.
How the costs escalated
The project was part of what the company described as “vibe coding,” a rapid software-building approach that relies heavily on AI tools. Slash has said it encourages employees to experiment with that method, but this case showed how quickly the costs can escalate when usage is not tightly controlled.
According to the employee, Nicolas Brilliante, the AI spending reached $81,267, based on a dashboard screenshot he shared on X. He described the incident as a real accident and said he had underestimated his own capabilities.
The game itself was simple in concept. It used a block-style world similar to Minecraft and asked players to shoot meme-inspired characters, including “skibidi toilet” and “tung tung tung sahur.”
| Item | Detail | Source |
|---|---|---|
| Company | Slash, a San Francisco fintech startup | Company statement on X |
| Game title | Brainrot Shooter | Company and employee posts |
| AI spending | $81,267 | Shared dashboard screenshot |
| Timeframe | Seven days | Company statement |
Why companies are paying closer attention
Slash’s case is landing at a moment when many businesses are rechecking their AI budgets. Coding assistants powered by AI can speed up development, but they can also become expensive quickly when high-powered models are used repeatedly.
The issue is no longer just whether AI tools are effective. Companies are increasingly asking whether the spending they generate is matched by real productivity gains and business value.
Polymarket has described the Slash incident as an example of how AI coding costs can surge when use is left without strict oversight. That concern is not limited to one startup, either.
One recent report said an unnamed company may have spent around $500 million in a single month on Anthropic’s Claude AI after failing to set usage limits for employees. The pattern is similar: enthusiasm for AI adoption moving faster than financial controls.
Big tech firms are adding guardrails
Some large companies are now tightening their rules around AI use at work. Uber, Coinbase, and Walmart are among the names reported to have introduced restrictions on employee AI access.
Bloomberg reported that Uber recently imposed monthly limits on AI coding tools after the company was said to have used up its annual AI budget within the first few months of the year. Walmart has also reportedly tightened controls to reduce unnecessary coding experiments and repetitive tasks that do not create meaningful business value.
The shift shows that AI is being treated less like a novelty and more like a major infrastructure expense. As adoption grows, the key question is becoming who can use which model, for what purpose, and up to what limit.
Slash’s experience captures the risk clearly. Even a small meme-themed game can become a five-figure AI expense in days when spending is left open-ended.