Zuck's Manifesto, Grok Gets Cheap and an AI Royal Comish.

Ah, yet another big week in AI. This is The AI Exhale, hosted by Vibe CFO with Jez and Cam. Five minutes, one thing that matters, or maybe two. We ignore the rest so you can too.

Jez: So in the news this week, we had good old Zuck release a 6,200 word manifesto. I remember going to uni and doing the 1,200 word count and creeping up to that. Quite positive and utopian in its outlook, but maybe there's a little bit of a sales plug in there. Who would have thought? So he's pushing open models, open weight models, with Meta's open source latest model to back it. The core idea is that AI capability shouldn't sit only with a handful of giant companies, and he's framing it as something that empowers small businesses and individuals, our sweet neck of the woods, and not just big enterprise.

But access to intelligence broadening fast isn't the same as businesses knowing how to use it. What do you reckon?

Cam: I reckon Zuckerberg, the utopian giver to the people, give me a bucket. The amount of damage that Zuckerberg has done in 20 years. He only cares about share price, and this is a share price move. And it might end up being good for the people, but it's a different thing going on here. The good thing about what Zuckerberg is doing, I mean, there's always this saying, if you are on the frontier, right, and winning, then you want closed AI.

And if you're losing, then you want open, right? And Zuckerberg and Meta, for all the money they've spent, there's no doubt that they've got an endless bucket of cash that they can throw at this, but they're losing, right? So they're going down the open weight path. And his point here is actually pretty clean. I mean, the greatest example was recently when they brought out Claude Design, right? This was sort of like this litmus test of what big business are worried about.

It's that Claude lets you in, steals all your data, then creates its own product and steals all your customers, right? And that's the story that's been told with Figma. The story is that Figma came in, did all the training, Claude got to know design, and then Anthropic launched Claude Design to take market share off it, right? And that's what other bigger businesses are worried about.

That's why they are pushing so hard for this open model, because they want it to be their own IP, their own intellectual property that they build and train on their own data, and it's proprietary and it's their power. Which makes more sense than it being closed and this knowledge being in the hands of a couple of players. Like, it makes sense, doesn't it? Open source then allows people to have their own proprietary models, their own trained data, all of their IP, all of their knowledge.

It fits nicely with AI becoming part of the world, you know, part of the internet, like just fitting in and everyone getting to have access to it.

Jez: It wasn't a crazy news week in AI, but obviously there's always something to speak of. So we'll keep it, as we do, short and sweet. Grok. It's good to see Elon's back in the AI game, or at least news-wise, outside of his rockets. Grok dropped a new model this week and the pitch is frontier level intelligence at roughly half the price of the other top models.

Cam: But for us, we want more competitors. You know, the more frontier intelligence, if we could bring Google up to speed as well, and we've got four. The more frontier intelligence, we want them in the price war. We want them getting us access to cheap intelligence. This is good news for us. Bring it on. Go Elon.

Jez: Get in there. Go Elon, indeed. All right. And we are Aussie founded, and more specifically out of South Australia. So let's turn to a bit of a story out of SA, and that is the Royal Commission into AI. It kicks off in October, with the final report not due until next July, in 2027. So you put that next to models making enormous jumps in a matter of weeks, and the obvious question is how any government inquiry can actually keep pace.

What are we thinking there? Is it worth tracking this?

Cam: It doesn't change anything Monday morning. I mean, it's good. We can't bash the government and say they've got no idea about how fast this is going, and then they're attempting to put together a research piece on what's happening and how they can govern with it. How can we put security measures in place and protect the people, but also let AI startups and the ecosystem flourish at the same time?

I mean, we don't know a lot with this, other than it looks like a good decision, but it doesn't change anything right now. Let them go. We'll see next year what comes out of it. I'd also put, you know, this is the difference, right? We just talked about Grok, right, and how fast. That was like one to two months for Grok 4.5 up to 4.6. I might be getting the numbers wrong, but in short, the level of intelligence jump that it did, this is the ratio.

It takes 10 months for this Royal Commission, and then one to two months for this intelligence jump. I mean, that's the argument people are making, that this will be outdated before it's even relevant.

Jez: Yeah, so definitely a disparity between those two dots, the innovation pace and good old government keeping up with the risk lens.

Cam: It did feel like there was a bit of a European approach, protection of the people and forget about innovation. It was both, but I would say there was more of a slant towards making sure that it's safe rather than letting us innovate on it.

And that's your five-ish minutes back. Vibe CFO, Australian founded. An AI CFO for small and medium businesses, out of the box, governed and secure. vibecfo.ai. Now you know.