🤖 In a development that has caused Silicon Valley to pause its collective oat milk latte and stare briefly into the void, Andreessen Horowitz-backed AI startup Cluely became embroiled in scandal this week after CEO Roy Lee admitted to fabricating the company’s revenue numbers — claiming ARR had doubled to $7 million in a single week before walking the figure back to $5.2 million after “public backlash,” which is startup-speak for “everyone on Twitter read the math and it did not math.” According to a new report from the Institute for Startups That Maybe Should Have Checked the Numbers Before the Press Release, 2026 has now seen more AI company revenue fabrications than any year since the peak of the 2021 NFT boom, when several projects claimed to have “revenue” that was technically a JPEG of a monkey.
😂 The scandal came with a truly poetic twist: multiple sources confirmed that Cluely’s core product uses AI to help users respond more effectively in professional settings — meaning an AI company that sells AI-assisted communication used AI-assisted communication to communicate incorrect numbers about its AI-assisted communication business. 🌀 The irony was not lost on the tech community, where several engineers described the situation as “recursively fraudulent,” while at least one venture capitalist reportedly smiled and nodded in a way that suggested this was not their first rodeo. Cluely’s a16z backers have not commented, presumably because they are currently using AI to generate a statement about AI generating false statements.
🤯 In parallel chaos, a major AI academic conference this week announced it had rejected 497 submitted research papers — roughly 2% of all submissions — because their authors had used AI tools to write peer reviews of OTHER papers, which is a violation of the conference’s stated policies and also, scientists note, somewhat defeats the entire epistemological purpose of peer review. 📚 The Journal of Things Scientists Are Not Supposed to Do called it “a fascinating case study in the automation of the gatekeeping of the automation of science,” a sentence that took three AI tools and four humans to write. Separately, the Trump administration’s ban on Anthropic’s Claude AI in government systems was struck down by a federal judge on free-speech grounds — meaning the U.S. government must once again use the very AI it tried to ban, presumably while being extremely annoyed about it.
💬 When asked whether AI startups faking their revenue numbers undermined public trust in the sector, a prominent tech investor paused for a moment on a podcast, took a breath, and then delivered what is already being quoted in three newsletter issues: “Look, every industry has growing pains. The railroad barons faked numbers. The dot-coms faked numbers. The social media companies faked numbers. The Web3 guys faked numbers while also faking the entire concept of what a number IS. Cluely faked some numbers. This is, in the grand sweep of American capitalism, a perfectly normal Tuesday. I’ve already wired them a second check.”
Leave a Reply