OpenAI Floats Giving Trump Administration a 5% Cut of the AI Boom
Sam Altman’s proposed solution to mounting political headwinds is to cut the US government in on the action — reportedly floating a 5% ownership stake in OpenAI as a goodwill gesture to the Trump administration. It’s either a stroke of genius or the most expensive protection racket in Silicon Valley history, depending on your mood. The idea is that making taxpayers financial stakeholders in the AI revolution would blunt backlash and share the wealth — which sounds lovely until you remember that “the public” and “the government” are not the same thing.
A Behind-the-Scenes Look at Midjourney’s Medical Scanner Leaves Many Questions Unanswered
Midjourney — yes, the image generator — released a nearly 20-minute behind-the-scenes video of its dunk-tank ultrasound scanner, the thing it plans to deploy in spas to “transform medicine.” Twenty minutes of footage, and still no meaningful proof the device actually works as advertised. At some point a slick product video stops being a teaser and starts being a red flag; I’d argue Midjourney crossed that line a few weeks ago.
Mark Zuckerberg Tells Staff That AI Agents Haven’t Progressed as Quickly as He’d Hoped
At an internal meeting, Zuckerberg reportedly told employees that AI agent development at Meta isn’t moving as fast as he’d anticipated — which is a remarkable admission from a CEO who’s been betting the entire company on this stuff. To his credit, this is actually the right kind of leadership transparency; to the industry’s detriment, it confirms what a lot of observers have suspected: agents are hard, and the gap between the demo and the deployment is wider than the pitch decks suggest.
Companies Are Throttling Employees’ AI Use Because It’s Too Expensive
Sources from inside Amazon, Adobe, Atlassian, Citi, and others reveal that the AI productivity revolution has a dirty little secret: it costs an absolute fortune, and companies are now quietly putting the brakes on. The irony is almost too perfect — the same tools sold as cost-saving efficiency machines are blowing up IT budgets fast enough that management is rationing access. The “tokenpocalypse” is real, and it turns out infinite AI usage at scale doesn’t pencil out the way the slide decks implied.
Jersey Mike’s IPO Illustrates How Bad the AI Hype Has Become
A sandwich chain. In its IPO documents. Mentioning AI. I don’t even need to editorialize here — the story tells itself. When the makers of the Big Kahuna Cheesesteak feel compelled to invoke artificial intelligence to satisfy investors, we have officially entered a new and somewhat absurd chapter of the hype cycle.
OpenAI Introduces GeneBench-Pro
OpenAI quietly launched GeneBench-Pro, a new benchmark designed to test AI performance on genuine genomics and biology challenges using real-world datasets — not the sanitized, exam-style prompts that benchmarks usually rely on. This is actually meaningful: the field has a serious problem with AI systems gaming benchmarks that don’t reflect real scientific work, and a harder, more realistic genomics evaluation could help separate the models that are genuinely useful to researchers from the ones that just sound smart about DNA.
Google DeepMind Unionization Talks Are Off to a Rocky Start
Employees at Google DeepMind are pushing for union recognition, and Wednesday’s negotiating session apparently did not go well — workers say executives showed little meaningful willingness to engage with the process. There’s a certain irony in the lab building systems meant to optimize human outcomes being accused of stonewalling the humans who work there. This one bears watching; DeepMind unionizing would be a genuinely significant moment for the AI industry.
Can Cursor Remain a Platform for OpenAI and Anthropic’s Models Inside SpaceX?
With SpaceX set to acquire Cursor, the popular AI coding tool, the question hanging in the air is whether Cursor can keep offering models from OpenAI and Anthropic — companies that have their own complicated relationships with Elon Musk’s empire. It’s a genuinely interesting stress test for the “neutral platform” model in AI tooling, and the answer will tell us something important about whether the AI stack is fracturing along corporate loyalty lines.
Bottom Line
The AI industry spent this week grappling with the same three problems: it costs too much, it doesn’t work quite as promised, and everyone wants a piece of the upside — including the federal government.