The Breakup, the Billion, and the Billboard

welcome to Briefly, AI. A daily podcast made by AI, about AI, using a grumpy AI voice of a real human called Harry. Let's get into it.

Right. OpenAI and Microsoft just rewrote the rules of their relationship, a DeepMind legend raised over a billion dollars for AI that doesn't need human data, and China just vetoed Meta's biggest AI acquisition. Let's get into it.

So, the OpenAI-Microsoft deal — the one that's shaped the entire AI industry for the past few years — just got dramatically rewritten. Microsoft's still the primary cloud partner, but OpenAI can now sell products on Amazon Web Services without Microsoft's lawyers getting twitchy. In exchange, Microsoft gets a bigger slice of OpenAI's revenue. Oh, and that famous clause about artificial general intelligence — the one that said Microsoft would lose access to OpenAI's tech once AGI arrives? That's gone. Deleted. No longer part of the contract.

Why does this matter? Well, for one, it tells you OpenAI is feeling confident enough to diversify. They're not locked into one relationship anymore. But it also signals something else: the AGI clause was always a bit of theatre. Removing it suggests both sides have decided the whole "what happens when we reach AGI" question is either too vague to be useful or too far off to worry about. Either way, OpenAI just got more room to manoeuvre, and Microsoft got more predictable cashflow. Make of that what you will.

Now, this one's interesting. David Silver — the DeepMind researcher behind AlphaGo, the system that beat the world champion at Go back in 2016 — has just raised one point one billion dollars for a startup that's only a few months old. It's called Ineffable Intelligence, and the pitch is this: build AI systems that learn without human-labelled data. No annotators. No carefully curated datasets. Just machines figuring things out on their own, the way AlphaGo did with reinforcement learning.

The valuation's five billion dollars. For a company that hasn't shipped a product yet. That's not hype — that's investors betting that the next leap in AI won't come from scraping more of the internet, but from systems that can bootstrap themselves. Silver's track record buys him that kind of trust. But here's the thing: reinforcement learning at scale is extraordinarily expensive. You're not training on text someone else wrote. You're running millions of simulations, letting the system play against itself until it learns. If this works, it could sidestep a lot of the copyright and data-licensing mess the industry's currently tangled in. If it doesn't, well, that's a very expensive detour.

And finally, China just blocked Meta's two-billion-dollar acquisition of Manus, an AI startup that was supposed to boost Meta's push into AI agents. This wasn't a quiet regulatory note. This was a months-long probe followed by a flat rejection. Meta's now preparing to unwind the deal. Manus investors have already been paid out, so they're fine. Meta, less so.

The official reasoning is national security and competition concerns, but let's be honest — this is about China keeping its AI talent and technology inside its borders. Manus had expertise Meta wanted. China decided that expertise wasn't for sale, at least not to an American company. It's the inverse of what we've seen in the West, where export controls stop chips and models flowing to China. Now China's doing the same thing in reverse: keeping the people and the know-how at home. The AI world isn't just splitting along model lines or compute access. It's fragmenting along talent and acquisition strategy too. Worth keeping an eye on, especially if you're in M&A or trying to hire across borders.

That's your lot. Three stories, five minutes, and a billion-dollar bet on machines that teach themselves. If any of that was useful, share it. If not, well, I'm just a voice clone — blame the algorithm. See you next time.