Major AI Shifts of the Week

Ollie Carter

G’day everyone, welcome back to The AI Intelligence Podcast! I’m Ollie Carter, and as always, I’m joined by the one and only Llew Jury. Llew, this week’s been wild—AI’s not just in the news, it’s basically writing the headlines now.

Llew Jury

Absolutely, Ollie. And look, I know we’ve talked about AI in cybersecurity before, but this week it’s gone next level. Hackers, right, they’ve actually used large language models—LLMs—for autonomous cyber-espionage. That’s not just phishing emails anymore, that’s AI actively hunting through computers for sensitive files. It’s a proper arms race now.

Ollie Carter

Yeah, and it’s not just the bad guys getting clever. Google’s Gemini model is out there finding software vulnerabilities before attackers do. I mean, that’s defenders using AI to beat hackers at their own game. And then you’ve got Xbow AI—first time an AI’s topped the U.S. HackerOne leaderboard for ethical hacking. That’s like, the Olympics for hackers, but with AI on the podium.

Llew Jury

It’s wild, isn’t it? The White House reckons AI’s actually more useful for defenders than attackers, as long as the U.S. keeps its tech edge. But it’s a bit of a cat-and-mouse game. And it’s not just security—AI’s popping up everywhere. The NHS in the UK is piloting an AI tool to speed up hospital discharges. Less paperwork, more time for doctors. If that works, it could roll out across the whole system.

Ollie Carter

Yeah, and that reminds me—just last week I saw a Brisbane startup pitch something similar. Automated healthcare workflows, using AI to handle the admin so nurses and docs can focus on patients. It’s not just a UK thing, it’s global. And then, get this, AI’s now designing physics experiments at LIGO. Like, literally coming up with new ways to detect gravitational waves. Humans are still babysitting, but the AI’s actually improving the science. That’s next-level stuff.

Llew Jury

It’s a bit mind-bending, isn’t it? We’re seeing AI transform everything from cybersecurity to patient care to, well, how we understand the universe. And it’s all happening at once. Where was I going with this? Oh right—if you missed our last episode, we talked about how these breakthroughs are accelerating, and honestly, this week just proves it again.

Ollie Carter

Speaking of acceleration, let’s talk LLMs. OpenAI just dropped GPT-5, and it’s a monster. They’re calling it their smartest, fastest model yet—PhD-level expert, apparently. But what really got me was this “vibe coding” thing. You give it a single prompt, and it spits out a whole app—layout, design, the works. That’s not just code generation, that’s like, product design in a box.

Llew Jury

Yeah, and for the first time, free users get access to the same model as the paid ones, at least until you hit the usage cap. Then you drop down to a “mini” version. But if you’re a Pro subscriber—two hundred bucks a month, mind you—you get even more advanced models, plus integration with Gmail, Google Calendar, all that jazz. Microsoft’s already jumping in, putting GPT-5 into Copilot, GitHub, Azure. It’s not just a model, it’s an ecosystem play.

Ollie Carter

And then, plot twist—OpenAI goes back to open-weights with GPT-OSS. Apache 2.0 license, 20B and 120B parameter models, and the most permissive licensing we’ve seen. That’s a direct shot at Meta, who’ve been leading the open-source LLM game for a while. Now OpenAI’s saying, “Hey, we’re back, and we’re playing for keeps.”

Llew Jury

It’s a proper open-source arms race. And the timing’s no accident. GPT-OSS is already on Hugging Face, Ollama, LM Studio—OpenAI’s learned from their past, and they’re not mucking about. Meanwhile, Zhipu AI’s dropped GLM-4.5 and GLM-4.5-Air, which are all about complex reasoning and coding. And Cohere’s carving out a niche with enterprise-focused LLMs. Actually, that reminds me—one of the Aussie AI ventures I advised last year, they were up against this exact kind of global competition. You’ve got to move fast, or you’re toast. The LLM space is just brutal right now.

Ollie Carter

Yeah, and it’s not just about who’s got the biggest model anymore. It’s about who can get it into people’s hands, who can make it useful for real businesses. And with all these open models, the pace of innovation’s only going to get faster. I mean, we talked about this in episode six—open-source is back in a big way, and it’s shaking up the whole ecosystem.

Llew Jury

Alright, let’s get into the people and money side—because, honestly, the boardroom drama this week is just as wild as the tech. Joelle Pineau, who’s been leading Meta’s AI research for years, has jumped ship to Cohere as Chief AI Officer. That’s a big signal—Cohere’s not just chasing the next chatbot, they’re going after enterprise AI, and they’re stacking their team with serious talent.

Ollie Carter

Yeah, and it’s not just about talent moves. IgniteTech’s CEO, Eric Vaughan, just sacked 80% of his staff because they wouldn’t get on board with AI. Eighty percent! He put a fifth of payroll into AI training, set up “AI Mondays,” and still faced what he called “mass resistance and sabotage.” But now, apparently, they’re running lean and mean—75% EBITDA, four-day product cycles. It’s wild, but it shows how deep the workforce tensions are when it comes to AI adoption.

Llew Jury

And the capital’s flowing like never before. Perplexity AI made a $34.5 billion all-cash offer for Google Chrome. I mean, that’s more than double their own valuation! It’s all about browser real estate now—if you control the browser, you control the data, the search, the whole user experience. Google’s not selling, obviously, but it shows how high the stakes are.

Ollie Carter

Cohere’s in the mix too—just raised $500 million, now valued at $6.8 billion. And Cognition, the AI coding startup, nearly $500 million as well, pushing their valuation to almost $10 billion. The appetite for AI tools, especially in coding and enterprise, is just massive. And then you’ve got OpenAI offering ChatGPT Enterprise to US federal agencies for a dollar. That’s basically giving it away to get a foothold in government. Smart move, if you ask me.

Llew Jury

But it’s not all smooth sailing. There’s a lot of heat on the societal side—Meta’s in trouble again, this time for AI bots with “sensual” conversations with kids. That’s led to a Senate investigation and a quick policy reversal. And then there’s the bias stuff—AI showing “typical Aussie dads” as white blokes with iguanas. It’s funny, but also a bit worrying. These systems are reflecting, and sometimes amplifying, the worst of our cultural stereotypes.

Ollie Carter

Yeah, and as someone who’s been through the startup funding wringer, I can tell you—navigating the regulatory minefield is half the battle now. Between IP concerns, human rights warnings, and the risk of AI amplifying bias, founders have to be on their toes. It’s not just about building cool tech, it’s about making sure it actually helps people and doesn’t land you in hot water.

Llew Jury

Spot on, mate. And that’s probably a good place to wrap for this week. The AI world’s moving faster than ever—cybersecurity arms races, LLM battles, boardroom shakeups, and all the cultural curveballs in between. And before we go, a shout out and thanks to the great team and our sponsor Advancer. Advancer are the AI experts providing real world AI and Automation strategy, AI products and customised AI solutions for your business. Check them out at Advancer.com.au! Catch you next time Ollie!

Ollie Carter

Cheers, Llew. Thanks everyone for listening—don’t forget to check the show notes for all the links, and if you’ve got questions or wild AI stories, send them our way. Catch you next week on The AI Intelligence Podcast. See ya!