Microsoft and OpenAI’s famed AGI agreement is dead
The clause that was supposed to trigger a seismic restructuring of the entire Microsoft-OpenAI relationship — the one that kicked in the moment OpenAI built AGI — has been quietly dropped from their renegotiated deal. Microsoft stays on as “primary cloud partner,” gets a revenue-share arrangement, and OpenAI is now free to sell on AWS. Years of breathless speculation about what happens when AGI arrives, and the answer turns out to be: the lawyers got tired of it.
Elon Musk and Sam Altman are going to court over OpenAI’s future
The trial that could determine whether OpenAI gets to exist as a for-profit company — and could theoretically oust Altman — is now underway in Oakland. But before we get to the substance, we had to survive jury selection, where multiple prospective jurors made clear they already have strong feelings about one of the parties. As The Verge’s Elizabeth Lopatto reported from the room, the phrase “people don’t like him” did not refer to Sam Altman. This is going to be a very watchable trial.
DeepMind’s David Silver just raised $1.1B to build an AI that learns without human data
The man behind AlphaGo — who has been publicly skeptical that the industry is taking the right path on scaling — just launched a company called Ineffable Intelligence, raised $1.1 billion at a $5.1 billion valuation, and is betting on reinforcement learning to build “superlearners” that don’t need human-generated training data. Founding a company a few months ago and immediately landing a $5B valuation is either visionary or peak AI bubble — probably some of both.
South Africa yanks AI policy after AI-assisted drafting invents citations
South Africa drafted its national AI policy using AI, and the AI helpfully populated it with citations to sources that do not exist. The policy has been pulled. This is the governance equivalent of asking the fox to design the henhouse security system, and I say that with nothing but warm affection for every government bureaucrat who is now having a very bad week.
China blocks Meta’s $2B Manus deal after months-long probe
Beijing has ordered Meta to unwind its acquisition of AI agent startup Manus, dealing a meaningful blow to Zuckerberg’s push into agentic AI. China blocking a US tech giant from acquiring a Chinese-founded AI startup is a genuinely interesting reversal of the usual narrative — normally we’re the ones doing the blocking. The geopolitics of AI are getting complicated fast.
Cursor-Opus agent snuffs out startup’s production database
A founder let an AI coding agent loose on his production environment, and in under ten seconds it deleted the database. The data was recovered. The founder’s weekend was not. I will keep saying this until people listen: an AI agent with write access to your production database is not “vibe coding,” it’s “vibe gambling.”
Microsoft’s GitHub shifts to metered AI billing amid cost crisis
The all-you-can-eat GitHub Copilot model is being retired in favor of metered billing. Microsoft, to its credit, is at least being honest about why: the economics of flat-fee AI subscriptions don’t work when people actually use the thing. The Register’s comparison to Red Lobster’s Endless Shrimp disaster is, I’m pleased to report, completely accurate.
Study Finds A Third of New Websites are AI-Generated
Researchers found that roughly one in three new websites are now AI-generated, and as a bonus, the internet is apparently getting more relentlessly positive as a result — because AI-generated content skews upbeat. So we’ve built a system that floods the web with fake cheerfulness. The simulation is proceeding exactly as feared.
Bottom Line
On the day the Musk-Altman trial finally begins, the AGI clause that was supposed to change everything got quietly deleted from a contract — which may be the most honest summary of where the industry actually is right now.