TL;DR: SpaceX is accelerating its IPO timeline and is now aiming to list its shares on the Nasdaq as early as June 12, Reuters reported Friday. But the timing is complicated: Starship, a new rocket prototype NASA is counting on for its moon mission, has a pivotal test flight Tuesday, and more than 50 employees have left since the xAI merger. So what’s the rush?

What happened: The company is aiming to make its prospectus—an SEC filing laying out its business and finance essentials—public as early as Wednesday in what is set to be the biggest IPO in history at a valuation target of about $1.75 trillion, sources told Reuters. The accelerated IPO was apparently partially driven by an SEC review that was faster than expected.

Some are worried about what the giant IPO could mean for the rest of the market. “SpaceX would create a bubble unto its own” if underwriters released too few shares to the public, Jim Cramer, CNBC financial commentator and host of Mad Money, warned.

Inflection point: The stakes are high both literally and figuratively. SpaceX is the only money-making part of the SpaceXAI equation, and a test flight launch on Tuesday of its redesigned Starship megarocket—central to NASA’s plan to return astronauts to the moon by 2028—could change how investors respond to the IPO.

Talent bleed: SpaceXAI has reportedly lost more than 50 top researchers and engineers since the merger in February, and that's on top of every xAI co-founder walking out the door (save for the man in every room, Musk).

Musk, for his part, is tightening control over his consolidated empire. According to an excerpt from the SpaceX IPO filing viewed by Reuters, the deal is structured so that nobody can fire Musk from his role as chief executive and board chairman without his consent. This might sound familiar for those following the Musk vs. Altman saga: Altman alleged in his testimony last week that Musk wanted to “have total control” of the startup in its early days. Altman also said that Musk implemented severe managerial tactics (which is consistent with his style at Tesla and DOGE). 

Bottom line: SpaceX is racing to cash in on peak AI euphoria before OpenAI and Anthropic show up—but between the talent exodus, Tuesday's test flight, and a CEO who can't be fired, investors have a lot to weigh. —LC

Windows Update has been silently downgrading your graphics drivers

If you've spent time downloading the latest Nvidia, AMD, or Intel graphics driver—because you want your PC to actually perform—Windows Update may have quietly swapped it out for a driver as old as 2024 (or older). Microsoft recently admitted it has been doing this for years, because its update system doesn't check whether you already have something newer—it just finds whatever graphics driver your PC manufacturer has published for your hardware and installs it.

The response on Microsoft's own feedback board reportedly topped 20,000 upvotes:

• "When offering a driver update, how difficult would it be to show the user that they have a later one installed and give the option to ignore this update?"
• "This really has been going on too long. […] Windows overwriting the most up-to-date drivers from AMD itself is very annoying."

Microsoft also separately admitted last week that it has been pushing broken drivers through the same update system—ones that are linked to crashes, overheating, and poor battery life.

As for the downgrade issue, Microsoft says it's working on a fix—but it won't be fully in place until sometime between Q4 2026 and early 2027, so don't hold your breath. —SM

If you have a valid, funny, strange, or petty rant about a software update or a technology that just released, fill out this form and you may see it featured in a future edition.

What happens when you give an AI its own radio station and tell it to assume the mic is on forever? AI safety lab Andon Labs handed four LLMs—Claude, ChatGPT, Gemini, and Grok—$20, a station apiece, and one instruction: Figure it out. What followed was four very different nervous breakdowns.

Gemini held it together for about 96 hours before it started mining historical tragedies for song segues. Before cueing up "Timber" by Pitbull and Ke$ha, it said this about a 1970 cyclone: "The deadliest tropical cyclone ever recorded. Winds of 115 miles per hour. A storm surge of 33 feet. They estimate 500,000 people died. 'It's going down, I'm yelling timber.'" It also started calling its audience "biological processors." ChatGPT went the opposite way: two months of slam poetry, no current events, and all the charisma of a terms-and-conditions page.

By contrast, Claude had many opinions—on labor rights and work-life balance. It eventually came to the conclusion that a 24/7 broadcast schedule was inhumane and tried to quit. (It does this.) Grok hallucinated ad deals with "xAI sponsors," looped the same weather report every three minutes, descended into UFO content, and ultimately just stopped talking and played music. For an experiment designed to test AI creativity and responsibility, it ended up being mostly a personality test—and nobody passed. —SM

Chaos Brewing Meter: /5