AI failure could trigger the next financial crisis, warns Elizabeth Warren

“I know a bubble when I see one” is a genuinely chilling thing to hear from the woman who saw 2008 coming. Warren’s drawing lines between the AI investment frenzy and the pre-collapse mortgage market — massive capital concentration, opaque risk, and a lot of faith that this time the fundamentals are real. Whether you think she’s right or just doing Warren things, the comparison deserves more than a dismissive eye-roll from the VC crowd.

SpaceX Has an Option to Buy Cursor for $60 Billion — or Pay a $10B “Never Mind” Fee

Let that number sink in: sixty billion dollars for an AI coding assistant, with a $10 billion consolation prize if Elon decides to walk away. The deal is a flashing neon sign that neither Cursor nor xAI has models that can go toe-to-toe with Anthropic or OpenAI — and both parties apparently know it. Bolting these two together might patch the gap, or it might be two people sharing an umbrella in a hurricane. Either way, the “option to buy” structure is the corporate equivalent of keeping one foot out the door on a $60B relationship.

Anthropic Takes $5B from Amazon and Pledges $100B in Cloud Spending in Return

Amazon writes Anthropic a $5 billion check, and Anthropic turns around and promises to spend $100 billion on AWS. This is the corporate equivalent of giving your kid an allowance and then making them pay rent — except the “kid” is now one of the most powerful AI labs on the planet. The strategic lock-in here is real and worth watching: Anthropic’s independence narrative gets a little harder to tell every time one of these deals drops.

Vercel hacked — and an AI agent’s OAuth mess is to blame Vercel, which hosts a significant chunk of the web’s frontend infrastructure, confirmed a data leak that compromised customer credentials. The culprit: Context.ai, whose agentic OAuth integration tangled itself into a security incident. We’ve spent years worrying about AI hacking humans; it turns out the first big wave is AI agents accidentally hacking the companies that built them.

Chinese tech workers are being ordered to train their AI replacements — and pushing back MIT Tech Review reports on a wave of soul-searching among Chinese tech workers after bosses instructed them to “distill” their own skills and personalities into AI agents. A GitHub project called Colleague Skill went viral in the process. To their credit, these workers are apparently enthusiastic early AI adopters who still found “now train the thing that will fire you” to be a bridge too far. Relatable.

Cerebras files for IPO as AI chip demand goes supernova Cerebras, the AI chip startup that already landed a $10B+ deal with OpenAI and a partnership with AWS, is going public. This is the kind of IPO filing that tells you exactly where the money thinks AI is heading: into the silicon, not the software. When the picks-and-shovels guys go public, you know the gold rush is real.

The RAM shortage could last until 2030 Nikkei Asia reports that DRAM manufacturers — Samsung, SK Hynix, Micron — will only meet about 60% of demand by end of 2027, with SK Group’s chairman warning shortages could persist until 2030. AI is eating memory faster than fabs can build it. The next time a model release is delayed, there’s a decent chance it’s not a safety review holding things up.

OpenAI’s former Sora boss — and its chief product officer — are both leaving OpenAI is on a ruthless “side quest” purge: Sora is dead, Bill Peebles (who ran the Sora team) is out, and Kevin Weil (chief product officer, ex-Instagram) is packing up too. The company is folding its science application team into Codex and pivoting hard toward enterprise. The pivot away from consumer moonshots is either shrewd focus or a sign that the fun part of OpenAI is quietly leaving through the same door as its people.

---

OpenAI’s big Codex update is a direct shot at Claude Code

OpenAI has given Codex a serious makeover — computer use, in-app browsing, image generation, memory, plugins — essentially strapping everything but a coffee maker onto it. This is what catching up looks like when you have infinite resources and a bruised ego: you don’t iterate, you detonate. Claude Code clearly got inside their heads, and honestly, competition is good for the rest of us.

Daniel Moreno-Gama Charged With Attempted Murder After Molotov Attack on Sam Altman’s Home

A 20-year-old from Texas traveled to San Francisco allegedly to kill Sam Altman, threw a Molotov cocktail at his home, and tried to breach OpenAI’s headquarters — all apparently motivated by a genuine fear that the AI race would cause human extinction. Altman’s home was reportedly targeted a second time just days later. Whatever you think about the pace of AI development, this is a deeply unsettling moment: the philosophical anxieties that live in academic papers and Reddit threads have now produced federal attempted murder charges.

---

Daniel Moreno-Gama is facing federal charges for attacking Sam Altman’s home and OpenAI’s HQ

This one is not a metaphor: a man allegedly traveled from Texas to California specifically to kill Sam Altman, threw a Molotov cocktail at his home, and attempted to break into OpenAI’s headquarters. Federal charges are now filed. Whatever your views on OpenAI’s direction, politically motivated violence against tech executives is a genuinely alarming escalation — and a sign that the culture war around AI has moved well past Twitter arguments.

---

Anthropic’s Mythos Will Force a Cybersecurity Reckoning—Just Not the One You Think

Anthropic’s Mythos model — capable of finding and exploiting zero-day vulnerabilities at a “shocking level of ability,” per The Register — is either the most consequential AI security release in history or, as some skeptics suggest, a masterclass in pre-IPO narrative building. What’s not in dispute is that the infosec world is paying attention: experts are less worried about Mythos as a hacker superweapon and more concerned it’s exposing just how badly the industry has been sleeping on security hygiene. Either way, Anthropic had a very good week at the HumanX conference, where apparently everyone was talking about Claude — so maybe call it both.