Marcel Zemp

🦔An NYU mathematician says OpenAI used his own progress against him to beat him to one of the biggest unsolved problems in mathematics. Tristan Buckmaster had been working toward a Millennium Prize proof using OpenAI's Codex when information about his progress reached OpenAI. Days later, OpenAI published a full proof of the same problem using the same uncommon approach, after burning $22.5 million in compute to get there. When Buckmaster confronted them, exec Sébastien Bubeck allegedly said "Why would you ruin your career?" and "If you don't want me to be nice, then I don't have to be nice." My Take OpenAI spent $22.5 million to solve a problem with a $1 million prize. They didn't do this for the bounty. They needed a headline that says "our AI solved one of the hardest problems in mathematics" and they needed it before someone else got credit. They started days after they heard about Buckmaster's progress and took the same uncommon approach he'd pursued for months. That's hard to explain as coincidence. Buckmaster did his work inside Codex. OpenAI reserves the right to train on Codex data. They admit they can't rule out that his usage helped improve their models. So a customer used their product, potentially handed them the roadmap, and then OpenAI outran him with $22.5 million in compute he could never match. I don't know if any of this was intentional. But if you're a researcher and you just watched this happen, I don't think you'd keep your best ideas inside someone else's product. Hedgie🤗 techcrunch.com/2026/09/08/op…

🦔AI can only fix security vulnerabilities 26% of the time. Researchers at 1Password ran over 6,000 AI-generated patches using Claude and ChatGPT against real vulnerabilities. Half the time the AI failed to fix the original bug. 4.5% of the time it created a brand new vulnerability that didn't exist before. And the researchers found that checking an AI-generated security patch takes more effort than just writing the fix yourself. Their conclusion was blunt. "The expected value of a fully LLM-generated, non-human-reviewed patch is a net-negative by a considerable margin." My Take OpenAI launched a program this summer called "Patch the Planet" where AI finds bugs and generates the fixes. These researchers ran 270 patches against one of those same bugs. Zero clean fixes. Not one. Every patch that fixed the original problem created a new vulnerability in the process. The partner that submitted a fix through OpenAI's program produced what the researchers classified as the worst possible outcome, it didn't fully fix the bug and it introduced a new exploit on top of it. Here's what this means if you don't write code for a living. Companies are using these AI tools to patch the software that runs your bank, your hospital, your phone. The pitch has been "AI finds and fixes security holes faster than humans." This study says the AI fix is four times more likely to be broken than working, and a third of the time it recreates the exact same mistakes human programmers already made. Reviewing the AI's work takes longer than doing it yourself. So the speed advantage disappears the moment you try to verify the output, which most companies won't do because the entire point was to move faster. I think we're going to see major breaches traced back to AI-generated patches that nobody checked, and the companies that shipped them are going to blame the tool instead of the decision to trust it. Hedgie🤗 Study: 1password.com/files/resource…

🦔Five ECB economists warned that an AI market correction is likely whether current valuations are justified or not. They studied every major tech revolution and found the same boom-then-correction pattern each time, even when the technology was transformative. European households hold about €440 billion in Mag 7 stocks through index funds and pensions. The ECB said policymakers have less room to cushion a correction now than during dot-com. Their line: a US AI correction would not remain a US problem. My Take The ECB isn't saying AI is fake. They're saying the technology can work and the stocks can still crash, because that's what happened with railroads, electricity, and the internet. The tech survived every time. The investors who bought at the top didn't. Lucent made the equipment that built the internet, dominated the market, and lost 98% of its value when the financing behind the buildout collapsed. Nvidia sells the chips that build AI and is now guaranteeing the leases and financing the purchases of its own customers. The technology was legitimate both times. The financial structure around it was not. €440 billion in European retirement savings is exposed to Mag 7 stocks through index funds, and the same concentration exists in US 401k target-date funds. If you hold an S&P 500 index fund, roughly 30% of it is in a handful of AI companies. The ECB flagged that the tools to respond to a correction are weaker now than in 2000. Rates are high, deficits are blown out, and the fiscal room to cushion anything has shrunk. That means if the correction comes, the people holding these stocks through their retirement accounts absorb more of the loss than they did last time, with less help on the other side. Hedgie🤗

🦔AI companies have borrowed so much money this year that they're pushing up interest rates for the entire economy. Nomura estimates tech borrowing alone now equals 25% of what the US Treasury issues in bonds, five times more than last year. Bank of America says the surge has added about 0.3 percentage points to the 10-year yield. Bond managers are selling Treasuries to buy AI corporate debt instead because it pays more. My Take AI companies are now competing with the US government for the same pool of lenders, and the lenders are picking the corporate bonds. Alphabet's 30-year pays 6.4%. A Meta data center bond pays over 7.5%. At those rates, a 5.2% Treasury loses the fight for capital every time. That's one of the reasons long-term rates have stayed so stubborn even as the Fed tries to bring them down. JPMorgan expects $5.5 trillion in AI infrastructure spending through 2030, and most of it will be borrowed. That borrowing raises the cost of money for everyone, the government, your mortgage, small businesses trying to get a loan. The AI buildout has reached the scale where it moves rates for the whole economy, and most people paying higher borrowing costs have no idea that a data center arms race is one of the reasons why. Hedgie🤗

🦔Leopold Aschenbrenner's hedge fund Situational Awareness just got forced out of all its public holdings after margin calls from Bank of America, Goldman, and JPMorgan. The fund hit $45 billion at the start of July. His biggest bets were SK Hynix, Nebius, SanDisk, Micron, and CoreWeave, all down more than 35% this month. His shorts on software names like Adobe went against him too. Citadel bought the bulk of what was left. He's 25 years old. My Take Aschenbrenner wrote the essay series that became the intellectual blueprint for the AI infrastructure trade. Massive expansion of chips, memory, data centers, power. He was probably right about the demand. He just needed it to arrive before his margin calls did, and it didn't. When both sides of your book move against you at once, longs down and shorts up, there's no way out. His prime brokers sold for him. He named the fund after the ability to see what's coming, and got margin-called because he didn't see the turn. $200 million into $45 billion in two years, given back in a month. Korea liquidated 75% of its leveraged positions this month. Now Wall Street funds are getting the same calls. The AI infrastructure thesis might be right over the long run, but a leveraged fund has to survive until then, and his didn't. Hedgie🤗

🦔For somewhere between $1 and $4, an AI can now figure out who you are from your anonymous internet posts. Researchers at ETH Zurich tested this on Hacker News users and matched 67% of them to their LinkedIn profiles at 90% accuracy. The AI read their posts, pulled out small details they'd dropped over months and years, and searched for the person who fit. Classical pre-AI methods scored near 0% on the same task. One of the paper's co-authors works at Anthropic. The full paper is free to read and I'd encourage everyone to look at it. My Take Most people assume that if they don't put their name on something, nobody can trace it back to them. That assumption is done. The AI doesn't need you to slip up in one big way. It collects the small stuff you've scattered across years of posts, your city, your job field, a pet's name, a repeated opinion, a favorite subreddit, and assembles a profile that narrows you down from millions of candidates to one. The researchers estimated it still works against a pool of a million people, at about 45% accuracy. Anyone with API access and a few dollars can run it. A government, a corporation, an ex, a stranger. The barrier is gone. Every anonymous account you've ever posted from, Reddit throwaways, forum accounts, anonymous reviews, is now potentially traceable to your name and employer by someone with a credit card and a few hours. If you've talked about your health, your workplace, your politics, or your personal life under a pseudonym, the protection you thought you had is much thinner than you assumed. The paper's authors put it simply: the practical obscurity that protected anonymous users online no longer holds. We should all take that seriously. Hedgie🤗 Link to research: arxiv.org/pdf/2602.16800