Employees burned $80k worth of tokens to create a "meme shooter game" using AI. After seeing the bill, the boss begged the whole internet to play it.

San Francisco fintech company Slash encouraged employees to try "vibe coding," resulting in a manager burning through $81,267 in AI tokens in one week to develop a shooting game full of internet memes.
(Background: From forcing employees to use AI to fearing burning too many tokens: more companies are tightening internal AI usage quotas)
(Background supplement: CFTC pushes back against Kentucky! Who says prediction markets should add a 14% tax?)

San Francisco fintech company Slash recently launched an internal initiative encouraging employees to try vibe coding, but the project's outcome far exceeded the company's expectations and unexpectedly sparked discussion online. Nicolas Brillante, Slash's Head of Strategic Vertical (X account @nickbruhman), spent $81,267 in AI token fees through the company credit card in just one week.

What he developed was a shooting game called Brainrot Shooter, set in a Minecraft-style environment where players shoot characters named after internet memes, including "skibbidi toilet," "tung tung tung sahur," and other meme素材 that have recently gone viral among younger demographics.

Afterward, Slash did not try to suppress the incident but instead proactively brought it into the open. Slash publicly announced on X with a half-joking tone, explaining that the company had been encouraging employees to try vibe coding, but one employee ended up spending $80k on the company credit card to develop a "meme shooter game."

Slash then immediately asked the public to try out the game, saying that if traffic was high enough, the company could classify the expense as marketing costs to make the books look better. This move itself also became another form of PR material. It is rumored that they have already received inquiries about advertising partnerships.

If you're a VC-backed startup & don't use Slash, shoot me a message.

Not only will I get you the best cashback, APY, and sign-on bonus, but I'll also throw in free billboard + menu branding placements in my game. https://t.co/Ke69H9ZYlS

— Nicolas Brillante (@nickbruhman) June 24, 2026

AI bills in the tech industry are spiraling out of control

This incident is not an isolated case—it is a microcosm of a problem spreading across the entire tech industry.

Over the past year, with the prevalence of AI coding tools like GitHub Copilot, Cursor, and Claude Code, "vibe coding" has rapidly spread from a niche practice among Silicon Valley engineers to a productivity strategy pushed by various companies. The problem is that the billing models for these tools mostly depend on token usage, and token consumption rates often far exceed the intuitive expectations of users without a technical background.

This incident also reflects a broader industry trend. Walmart was the first to explicitly stipulate that employee spending on AI tools must be subject to caps, with the core consideration being to prevent resources from being used on personal AI experiments unrelated to business. Uber and Coinbase have also followed suit, setting monthly or quarterly spending limits. This collective shift by large enterprises signals that the honeymoon period of "letting employees freely explore AI" is ending, and companies are starting to use more pragmatic financial discipline to rein in this wave of technology.

This marks a turning point: the "good enough" experimental phase of AI is ending, and companies are beginning to seriously calculate how much of their AI bills are genuine productivity and how much is just an illusion of "feeling productive."