How We Use AI Is Changing

June 8, 2026 · Episode Links & Takeaways

Author

AI Daily Brief
June 08, 2026

HEADLINES

Trump Confirms Government Is Exploring AI Equity Stakes

President Trump has confirmed reports that the administration is exploring taking equity stakes in major AI labs, calling it a way for "the American public to essentially become a partner with the companies." OpenAI appears to be actively driving the concept, with Sam Altman meeting Bernie Sanders to pitch donating equity to a public wealth fund that could distribute AI dividends — possibly through Trump accounts for children. The cynical read, as former Microsoft employee Dare Obasanjo put it, is that "the groundwork is already being laid for a government bailout of OpenAI." The less cynical read: if you don't think the government already views leading AI labs as too big to fail, you're not paying attention.

On the other hand, there's genuine debate forming. Former AI Czar David Sacks acknowledged why the Sanders proposal resonates even with conservatives, before warning that nationalization would "accelerate the corporate-government fusion we're already sliding toward." Investor Brad Gerstner took both sides — opposing government ownership in principle while actively encouraging founders to donate shares directly to citizens via Trump accounts. The Overton window on the government's relationship with AI has become, as Nathaniel put it, a "flapping open Overton door."

Google Signs $920M/Month Compute Deal With SpaceX

Elon's pivot to cloud kingmaker has its second major customer. Google has signed a three-year deal with SpaceX, paying $920 million a month for access to at least 110,000 NVIDIA GPUs — structured the same way as last month's landmark Anthropic deal. Google's cloud spokesperson called it "a short-term, timely agreement" to meet surging demand for Gemini Enterprise. As always with Elon deals, this is a Rorschach test: short seller Jim Chanos quipped that "this nine-month contract has more easy outs than a kids' T-ball game," while Boring Business did the math and found that between the Anthropic and Google deals, xAI will pull in $26 billion per year against a $40 billion data center spend — an 18-month payback from just two customers. As one observer put it: SpaceX has accidentally become the largest neocloud on Earth, with ~550K GPUs and more than double CoreWeave. Starlink does $15B ARR, but GPU rentals may now be SpaceX's biggest business.

Jensen Locks Up NVIDIA's Memory Supply With SK Hynix Deal

Jensen Huang has secured NVIDIA's memory supply in a new multi-year deal with SK Hynix, ensuring the company remains NVIDIA's largest memory supplier as shortages deepen across the supply chain. The deal also makes NVIDIA a design partner on next-gen memory chips for physical AI, personal AI, and AI infrastructure — and secures HBM supply as Vera Rubin production ramps. A big part of the NVIDIA story this year has been Jensen personally traveling the globe to close deals: last month Taipei for TSMC, this week Seoul dining on grilled pork belly and soju with the country's top executives. This deal was reportedly sealed over a chicken-and-beer meeting with SK Group Chairman Chey Tae-won. As Huang told the press on his way out: "Demand is enormous. Everything in the entire industrial supply chain, from wafers to silicon photonics and cable connectors, is in a state of supply shortage."

MAIN STORY

The Way We Use AI Is Changing

There's a story about the ChatGPT superapp overhaul that a lot of people are misreading — and getting the read right tells you a lot about where AI user experiences are actually headed. Yes, there are financial and IPO implications to what OpenAI is building. But the more important thing happening is the recognition that the most valuable AI use cases are no longer about chat, and that interfaces are the lever for bringing those use cases to everyone.

SUPERAPPS AND STRATEGY

The ChatGPT Superapp
FT: OpenAI is plotting the biggest ChatGPT overhaul since launch.
The Financial Times reports OpenAI intends to transform ChatGPT into a superapp combining coding tools, AI agents, image generation, and partner apps from the likes of Canva and Booking.com. The business framing is clear: this is about chasing enterprise revenue and setting up an IPO, with the FT noting the pivot mirrors Anthropic's "make money first" strategy. Jenny Xiao of Leonis Capital put it bluntly — the two companies are converging because investors care more about money than dreams. The cynical take from several corners: "Nobody builds a superapp because users asked for one. They build it because a chatbot is hard to put a multiple on. This is a feature for the S1, not for you." That's not entirely wrong. But it's also not the whole story.

The Widening AI Advantage Gap
Power users aren't just using AI more — they're using it differently.
OpenAI CFO Sarah Friar recently disclosed that free users average about seven turns a day, Plus subscribers around 21, and Pro users about 77. The gap between what power users and casual users are getting out of AI has been widening dramatically since agents became viable — roughly from November 2025 through January 2026. People running coding agents are seeing compounding value; people using regular chat are seeing only linear gains. Codex has become OpenAI's most important product by the metrics that matter: user numbers are up sixfold since the desktop app launched in February, and in a recent developer poll, 51% named Codex as their primary coding agent. The business model consequence is equally stark — the difference between seat-based and usage-based pricing is the difference between a $3B run rate and a $47B run rate.

Loops: The Next Level of Abstraction
You shouldn't be prompting coding agents anymore — you should be designing loops that prompt them.
OpenClaw creator and now OpenAI employee Peter Steinberger put the leading edge plainly. And Claude co-creator Boris Cherny recently shared that he's moved even past prompting agents directly: he now writes loops that prompt agents and figure out what to do autonomously. This pattern has a lineage — from the Ralph Wiggum loop, to Andrej Karpathy's auto-research work in March, to the /goal primitive now embedded in both Claude Code and Codex. The idea in each case is the same: require less human intervention, let the AI run longer, fix its own mistakes, and accomplish more complex tasks end-to-end. The challenge is that almost nobody outside the vanguard knows how to do this. Railway founder Just Jake described it as "evidently the future and also somehow gatekept." Shanu Matthew's question — "what does this mean for the non-coder audience?" — got 300,000 views, which tells you everything about the size of the gap.