Episode — 2026-04-16

**Episode title:** The Chip King's Warning, Tesla's Brain, and the Consultants Who Noticed

The man who sells the shovels in the AI gold rush just told everyone to slow down and talk to each other. That's new. Let's get into it.

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.

**Story one.**

Jensen Huang — CEO of Nvidia, the company that makes the chips powering almost every AI system on the planet — said something unusual yesterday. He argued that Anthropic's Mythos model, the one that found thousands of zero-day security vulnerabilities and was quietly shelved for being too dangerous, proves that the US and China need to actually sit down and talk about AI safety.

Now, if you've been listening this week, you'll know Mythos has been the ghost haunting these episodes. Monday it was about Anthropic's PR machine versus what it actually builds. Wednesday it was the drone race. And now here's Jensen Huang — not a safety researcher, not a policy person, the CEO of a $3 trillion chip company — saying that this particular model is serious enough to warrant diplomacy.

The bit that matters: when the person who profits most from AI acceleration starts calling for international guardrails, that's not a casual opinion. It's a signal. Whether the US government listens is a different question entirely.

Worth keeping an eye on: Jamie Dimon at JPMorgan also weighed in this week, saying Mythos "creates additional vulnerabilities." Two of the most powerful people in American business in two days, both saying the same thing about the same model. That's a pattern.

**Story two.**

Tesla tapped out its AI5 chip. If that sounds like jargon, here's the short version: the chip design is finalised and has gone to the factory. Mass production is expected by 2027.

The numbers are striking. Ten times the performance of the current chip, five times the memory bandwidth, somewhere between 2,000 and 2,500 TOPS — which is a measure of how fast it can crunch the kind of decisions you need for real-time driving or robot movement.

Tesla is planning to put this chip in its self-driving cars, its Optimus humanoid robots, and its own data centres. That's a vertical stack that most companies can only dream about: you design the chip, build the robot, train the AI, and run the cars — all on your own silicon.

Why does this matter beyond Tesla? Because it's another piece of physical AI infrastructure locking into place. This week we've seen Google's embodied reasoning model in factories, Nvidia's partnership with Boston Dynamics, and now Tesla's own chip for its own robots. The software layer is maturing. The hardware is following. Physical AI is starting to feel less like a concept and more like a construction site.

**Story three.**

And this one's a bit more personal, depending on where you work.

Bloomberg reported yesterday that McKinsey, BCG, and Bain — the three biggest strategy consultancies in the world — are quietly changing how they hire at entry level because of AI. The traditional model was: hire very smart young people, give them tedious analytical work, watch them grow. The problem is, AI now does the tedious analytical work faster and without needing a salary or a graduate scheme.

So the firms are apparently rethinking what they actually need from junior hires. Less emphasis on raw analytical horsepower — AI does that — and more on client skills, communication, judgment, the stuff that's genuinely hard to automate.

Now, healthcare communications is a bit different from management consulting. But the underlying logic applies anywhere that junior roles were built around structured analysis, research, and synthesis. If your agency is still hiring on the same basis it was five years ago, this is worth a conversation.

The honest take: nobody knows yet whether this means fewer jobs or different jobs. But "we need people who can do things AI can't" is becoming the new "we need people who can use Excel." It was always true, but now it's urgent.

That's your lot for Thursday. A chip king calling for diplomacy, a chip that'll drive your car and run your robots, and the world's most prestigious firms quietly rewriting their job descriptions. The week ends tomorrow — and there's a thread running through all of it that's worth sitting with. See you then.