Musk Takes the Stand — And It’s More Soap Opera Than Salvation
Day one of Musk v. Altman is everything you’d expect: Elon on the stand, bootstrapping his origin story from South African childhood to Canadian arrival with $2,500 in travelers’ cheques, framing himself as a reluctant hero who co-founded OpenAI solely to prevent a “Terminator outcome.” The judge, apparently unimpressed by either party’s courtroom discipline, had to warn both Musk and Altman to knock off the social media sniping mid-trial. The case itself is genuinely consequential — it could determine whether OpenAI is allowed to complete its for-profit conversion ahead of a major IPO — but the vibes, per The Verge’s reporter, were more “flat and adrift” than “man on a mission.” Hard to play the reluctant savior when you’re the one who filed the lawsuit.
OpenAI Breaks Up With Microsoft’s Exclusivity, Immediately Moves In With Amazon
One day after OpenAI and Microsoft amended their partnership to end exclusive cloud rights, Amazon had OpenAI’s GPT models, Codex, and Managed Agents up on AWS Bedrock. That was fast — almost like someone had been waiting by the phone. OpenAI also announced the move itself, framing it as enterprise flexibility; what it really signals is that OpenAI is aggressively diversifying its distribution before the IPO and doesn’t want to be permanently attached to any one cloud at the hip. Microsoft probably saw this coming. That doesn’t make it sting less.
OpenAI Really Needs Its Coding Agent to Stop Talking About Goblins
Buried in OpenAI’s system prompt for its Codex coding agent is a delightful instruction: “Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant.” I have so many questions. What did the raccoons do? What were the gremlins saying? This is the unglamorous but endlessly fascinating reality of deploying AI products at scale — somewhere, a product manager had to sit down and write “no goblins” into a document that goes out to the world. We’re building the future, people.
The Race to Keep AI Agents From Maxing Out Your Credit Card
The FIDO Alliance, Google, and Mastercard are working on authentication standards for AI agents that will soon be empowered to make purchases on your behalf. This is the kind of story that sounds like a minor tech process update until you realize the alternative is an autonomous agent with access to your payment info and no reliable way for merchants — or you — to verify what authorized the transaction. The era of “my AI accidentally subscribed me to 14 things” is closer than it looks, and it’s genuinely good that serious people are working on the guardrails before the car is already over the cliff.
Google Expands Pentagon AI Access After Anthropic Said No
After Anthropic declined to let the Department of Defense use its AI for domestic mass surveillance and autonomous weapons, Google stepped in with a new contract expanding Pentagon access to its AI. To be clear: Anthropic drawing a line on domestic surveillance is notable and worth acknowledging. Google deciding that line is a business opportunity is equally notable. The AI-and-defense question is one of the defining ethical fault lines of this industry right now, and the companies are landing on very different sides of it.
‘It’s Undignified’: Meta’s AI Training Workers Face Mass Layoffs
More than 700 workers employed by a Meta contractor in Ireland — the humans doing the painstaking annotation and data work that makes AI actually function — are reportedly at risk of losing their jobs. One worker’s choice of word, “undignified,” is doing a lot of heavy lifting here. These are the people whose labor sits underneath every slick demo and every product launch, and they tend to be the last ones mentioned when the wins are celebrated and the first ones cut when budgets tighten. The gap between AI’s trillion-dollar valuations and the conditions of its workforce remains one of the industry’s least comfortable conversations.
People Using AI to Represent Themselves in Court Are Clogging the System
AI-assisted pro se litigants are filing more cases, more motions, and more appeals than the court system can comfortably process. On one hand, democratizing legal access is genuinely good; on the other, courts are already strained, judges are already frustrated, and AI-generated filings have a well-documented tendency toward confident hallucination. The irony that Musk and Altman — two of the most powerful people in AI — are fighting over the future of the technology in the same court system being quietly overwhelmed by it is not lost on me.
The Missing Step Between AI Hype and Actual Profit
MIT Tech Review revisits the persistent gap between AI’s enormous valuations and the actual, measurable business returns most enterprises are seeing. The framing borrows from South Park’s underpants gnomes — “Step 1: Deploy AI. Step 2: ??? Step 3: Profit” — and it’s not entirely unfair. The enterprise data infrastructure problem is real, the change management problem is real, and the pressure to show ROI on massive AI investments is building fast. The hype cycle has been running for three years now; the reckoning phase is starting to show up in earnings calls.
Bottom Line
Everyone wants to save humanity with AI — the billionaires in court, the companies pitching the Pentagon, the agents eyeing your credit card — and nobody’s entirely sure who’s going to pay for it or clean up the mess.