While the normies were resting, we mainlined tech discourse all weekend to bring you the latest trends, rumors, fights, and innovations from the most manic corners of the internet. This week: Oracle’s future looks murky, an Anthropic researcher uses Claude to solve a famous math problem, China’s Moonshot is paying off.
This is Manic Monday.
Who’s divining Oracle’s doom?
Oracle is a pinnacle company in tech, but the higher you climb, the longer the fall. The company is seeing soaring expenses, a credit-score disaster and, consequently, banks unwilling to give it more billions. The culprit looks to be Oracle’s $300 billion computing agreement with OpenAI, which has transmogrified from a crown to an albatross on founder Larry Ellison’s neck due to the challenges of building monolithic data centers across America.
“The people who got harmed the most are the ones who trusted Sam Altman,” says tech industry critic Ed Zitron, who pens the popular Where’s Your Ed At newsletter.
The weight has dragged Oracle’s credit score into the murky depths of “one level above junk status.” If this reeks of the 2008 subprime mortgage crash, well, you’re not alone in thinking that: Michael Burry, the hedge-fund genius who made $800 million shorting that market, feels Oracle is in freefall. To that point, Oracle’s stock has dropped more than 60 percent since September. Ellison himself has lost $213 billion in that time. Maybe that’s why the Oracle layoffs have killed 21,000 jobs in the last 12 months.
The problem isn’t just that people despise AI data centers enough to ban them and threaten AI execs with violence. The expectation of endless growth of AI is just running into material limitations: “Microsoft, Google, Amazon, Oracle and Meta are spending obscene amounts on AI infrastructure. They have the money. They have the chips. What they don't have is power,” Tiny co-founder Andrew Wilkinson wrote on X. (We’re talking literal, not metaphoric, power: Elon Musk recently dropped a cool billion to buy a gas turbine company to fuel data computing.)
Speaking of desperation, Ellison is also flailing on a second front: The blockbuster $110 billion merger of Paramount — led by his nepo baby CEO, David Ellison — and Warner Bros., a deal that’s currently being opposed by 12 different states. Daddy Larry promised to back the deal with $40 billion of his own cash, which, in hindsight, might’ve been a misstep! Maybe the Hawaiian gods are finally punishing Ellison for the crime of colonizing an entire island for the super-rich. —Eddie Kim
Claude (allegedly) disproves a famous algebraic theory
Levent Alpöge, a researcher at Anthropic, announced yesterday that he had disproven the 87-year-old algebraic problem Jacobian Conjecture with the help of his “close friend” Claude Fable 5.
Obviously, the discovery is not peer-reviewed nor are Fable 5’s math skills perfect: The model scored a modest 77 out of 100 for its math capabilities, according to math benchmark test ProofBench. That said, academics do appear to be taking Alpöge’s finding seriously: One mathematics professor at Queen Mary University of London told The New Scientist that “this is a pretty big deal,” calling it “probably…the biggest conjecture that AI has played a significant role in [proving or disproving] so far in mathematics.” Stanford mathematician Jared Duker Lichtman called it “remarkable.” We’ll just have to take their word for it; I am but a writer who never even took AP Calculus BC. To all the math-heads reading this, please tap in. —Cydney Hayes
The latest development in surveillance girl summer: MSG sues Wired
All summer so far, it seems like every big story has been about who’s watching, and who’s getting watched. First, our investigation into the ring of TSA-style face scanners at Castro bars rocked the neighborhood; a couple weeks later, Meta’s hidden-camera sunglasses sparked backlash from privacy advocates and people with love in their hearts. Now, Madison Square Garden is suing Wired for defamation, claiming its July 9 article titled “Madison Square Garden Kept a List of Gay Celebrities” created a “false narrative” that the venue was tracking and discriminating against LGBTQ+ patrons.
The July 9 article was not the first time Wired reporters had dug into MSG’s surveillance network: In April, the magazine published a shocking exposé that detailed how the venue and its owner, Jim Dolan, stalked one trans Knicks fan. Its dragnet systematically reported her whereabouts and behaviors in the arena, where the Knicks have played for 58 years, including when she would leave her seat, buy concessions, and use the restroom. Helmed by its unyielding editorial director Katie Drummond, Wired released a statement in response to the lawsuit on Thursday, saying they “look forward to continuing our coverage of MSG, and of billionaire James Dolan’s use of technology across his entertainment empire.” —C.H.
The Moonshot landing ain’t no fake
Once again, a Chinese AI company has emerged as a big baddie poised to take the lunch of the likes of OpenAI and Anthropic, and everyone’s paying attention. This time, it’s Moonshot, which released its newest model, Kimi K3, on Friday to raves. It’s an open-source model, unlike the ones used by ChatGPT and Claude, giving developers and users more creative freedom. That notion is so scary that Dean Ball, OpenAI’s head of strategy, panic-typed an entire essay about the threat of “full AI communism.” (Former White House AI czar David Sacks, on the other hand, mixed praise with existential anxiety: “We’re only making ourselves less competitive.”)
The Moonshot launch has resurfaced Donald Trump’s desire to ban Chinese open-source AI in the US, but many founders and AI nerds are marveling, not mad. Consider how, on Sunday, Moonshot paused new subscriptions (forgoing precious revenue) for Kimi to fix performance issues for existing users — a sharp contrast to Anthropic’s decision to throttle users in April when facing the same bugs. Commence more plaudits from the commentariat.
Pretty good for founder Yang Zhilin, a 33-year-old recluse, and his crack squad of programmers. Who needs a fancy office when you can just cram five dudes in front of two screens and let ‘em rip? As one dev put it, “Mfs are running a $20B startup and still live like this.” Moonshot is now poaching some of OpenAI and Anthropic’s top minds; meanwhile, as Y Combinator founder Anshu Chimala joked, the AI leaders in America are shilling “worse models” while distracted by “a blog post about the collapse of humanity.” Techno-optimism indeed. —E.K.
The war over VAR rages on
The World Cup ended on Sunday with Spain lifting the golden trophy despite dirty, violent, cheating antics by Argentina, which has the ignominious distinction of being the only team in World Cup history to not shoot a single target on goal for the entire match. Now, soccer fans’ fixation is on a different element of the game: The Video Assistant Referee, aka VAR.
The technology uses 42 cameras placed around a soccer field to capture fouls, offside players, and other infractions; the footage is reviewed by humans, and VAR is used to prove or overturn calls made on the field by human referees. On paper, it ostensibly reduces human bias and errors, making the game more fair. In reality, since its introduction in the 2018 World Cup, it’s viewed increasingly as an open Pandora’s Box that fans fear will slow the game down with constant reviews of minutiae. Many a goal has been overturned by VAR because it deemed a player was a sliver of a shoulder offside.
More controversially, however, VAR review is triggered by a human ref, meaning that a technology designed to be precise and bias-free is… completely dependent upon a human’s instinctive decision. Not to be conspiratorial, but do we really believe it’s fair that Argentina was never subject to VAR review in the entire Cup despite committing the most fouls? The whole discourse is a metaphor for tech: The perfect tool can’t be trusted if the wrong person is bending it to their will, to paraphrase writer Jay Kaspian Kang. —E.K.
