“Is this like a weird porn thing?” asked one concerned driver.
Photo by Amit Roth
The high-tech industry, quartered in California’s Silicon Valley, continues to apply AI solutions to businesses’ money problems. Entrepreneurs have inserted large language models (LLMs) in search engines, customer service, and dating simulators, and are “always on the hunt” for more features to replace. Brill.ai, a Sunnyvale-based startup, disrupted the advertising industry with “Brill-iance”-capable billboards. Business-to-business advertisers were able to display commercial messages for AI tools that react to on-road events such as traffic jams and honking drivers. However, repeated and unintended instances of rogue graphics led to the decommissioning of “Brill-boards” across the Bay.
Commuters on the 101 Highway reported “unfathomable slop” projected onto the Brillboards, ranging from rapid flashing GIFs, to fingers added in real time, to Reddit stories. This distracted South Bay-ers and caused average rush hour speeds to halve to ten miles per hour. San Jose resident Sam Carlos “ratioed” a Xeet by Brill.ai, tweeting: “You;d [sic] think that as a b2b saas company Brill would treat
their seed rounders better than to slop it out on main. If you have the venture capital to platform FAANG and even incubate smaller AI tools, don’t worsen the AI sector’s reputation.”
After Labor Day, Ian Grokholsky, CEO, founder, chief programmer, and senior advisor of Brill.ai publicly announced the final recall of the company’s sole product. “Our fundamental concept, while revolutionary, came too early. With a few more lines of code, our for-businesses ad platform could really take off. We still have the greatest minds in Silicon Valley working on it — I’ve assembled a top notch team here at Brill.ai — and we’re going to find the AI solution for this one too. Inves- tors willing, we’ll be back out there on your commute.”
An investigation into Brill.ai’s eight-person team revealed that most of the real-time “Brill-ance” advertising used manual reviewers based abroad. Reviewers monitor a live feed and send prompts to the “proprietary” Brill LLM, which runs on GPT-5. “It was a fault in what Brill could imagine, really,” said an anonymous Brill.ai whistleblower. “People are just plain funnier. They can improv much better content ideas than a visualizing AI could generate based off prompts. I mean, while Google thinks you need Gemini to handhold you through babysitting a goddamned kid, us manual reviewers needed to pass the time somehow. And I guess ‘make the babysitter a redditor and soy react tenor gif to gemini’ was hard to interpret.”
Grokholsky denies that manual review disqualifies Brillboards from being marketed as an “automated” product, citing “if Just Walk Out can do it,” but most of their previous “ride or dies” have moved on to other market-disrupting cash cows.
“Actually, god bless regulation in the AI sector,” tweeted Carlos in an Xeet that would go on to be ratioed by antiaccelerationists. “We need these kinds of pro- active and preemptive investigations after we find out an AI company is shitting itself this badly.”
Amit is a cog in this machine. But doesn't everything run on optic cables or something?
