Nora: [dry] Seven hundred dollars to have a fake Harvard professor tell you your idea is bad.
Dex: This is Dex.
Nora: And I'm Nora.
Dex: And this is Crash and Burn for Sunday, August 23rd. We've got a lot to get to in the world of startups and AI.
Nora: Starting with Harvard, which apparently thinks it can sell you an AI business school instructor.
Dex: Then there's a new British AI lab, started by ex-DeepMind folks, who say their AI teammate is already lapping the competition.
Nora: And in a total plot twist, OpenAI is now begging California for more AI regulation.
Dex: Alright, let's start with Harvard. They've just launched something called the HBS Foundry.
Nora: Which sounds prestigious. Let me guess, it's a startup bootcamp?
Dex: Exactly. But here's the kicker: for six hundred and ninety-nine dollars—
Nora: $699. Okay.
Dex: —you get access to AI avatars of its instructors. Digital versions of real HBS faculty.
Nora: So you're not talking to a person. You're talking to a deepfake with a tweed jacket.
Dex: [laughs] Basically! The idea is you can practice your pitches and run through board meetings, and get feedback from these AI professors 24/7.
Nora: Wait, the AI gives you feedback? On your business?
Dex: That's the pitch. Instant, on-demand mentorship from a digital version of a world-class expert.
Nora: Dex, this is a very expensive chatbot with a Harvard logo slapped on it. I can get my pitch torn apart on Reddit for free.
Dex: But can you get it torn apart by an algorithm trained on decades of HBS case studies? They're selling the brand, the feeling of Harvard.
Nora: [sarcastic] Oh, the feeling. For seven hundred bucks. So what does this mean for founders? It means prestige brands now think they can sell you an LLM in a fancy wrapper and call it an education.
Dex: I don't know, if you're a founder in the middle of nowhere, maybe this is your only shot at that kind of feedback. It's democratizing access, right?
Nora: Or it's democratizing disappointment. 'Sorry, your pitch was ranked in the 17th percentile by Professor Bot 3000.' Hard pass.
Nora: Alright, let's head over to the UK. There's a new AI lab called Inherent, started by some heavy hitters from DeepMind, and they are making waves.
Dex: And when you say 'making waves,' you mean they're coming directly for the kings. OpenAI, Anthropic...
Nora: Exactly. They just released a paper about their new AI agent, named Faraday.
Dex: Faraday. Love the classic scientist naming convention. So what does it do?
Nora: This is the wild part. They tested its ability to replicate the findings of scientific papers. Given a paper, can the AI reproduce the results?
Dex: Whoa. That's a huge problem in science, the replication crisis.
Nora: Right. And they claim Faraday smoked both OpenAI's and Anthropic's latest models at this very specific, very difficult task.
Dex: Outperformed them? So a new, smaller lab just beat the giants? [excited] I love this.
Nora: It's a specific benchmark, but still. If an AI can reliably replicate research, think about what's next: helping generate new ideas.
Dex: So it could accelerate R&D like crazy. Imagine your AI teammate just running a hundred experiments overnight.
Nora: That's the dream. So what does this mean? The AI arms race isn't just about chatbots anymore. It's about specialized agents that can actually do science.
Dex: And it shows that a small, focused team can still come out of nowhere and create a breakthrough. The game isn't over.
Nora: Not by a long shot. The moat might not be as wide as some people think.
Dex: Speaking of moats, let's talk about the company that thought it had one: OpenAI.
Nora: Oh, this is a good one. What are they up to now?
Dex: Something really weird. They've done a complete 180 on a big AI safety bill in California, SB 53.
Nora: Hang on—SB 53? Wasn't that the bill they were lobbying against just a few months ago? The one they said would stifle innovation?
Dex: That's the one. And now, get this, in a blog post yesterday they said they not only support it, they think California should make it even stronger.
Nora: [dry] I'm sorry, what? Strengthen it? This smells... off.
Dex: Their official line is that after 'extensive internal red-teaming and reflection,' they believe stronger guardrails are necessary for the next generation of powerful models.
Nora: Oh, bull— [censor] — They had a 'reflection'? Or did they see regulation was inevitable and decided to get in front of it to write the rules themselves?
Dex: That's the cynical take, for sure. Welcome to regulatory capture 101.
Nora: It's not cynical, it's pattern recognition! They fight the rule, realize they're going to lose, and then swoop in proposing changes that just so happen to benefit them and make it harder for smaller competitors—
Dex: Like Inherent.
Nora: Exactly! They want to strengthen it in ways that only a company with a thousand lawyers can comply with. It's pulling up the ladder.
Dex: So the takeaway for startups is: when a giant suddenly starts talking about 'safety,' odds are they're just protecting their market share.
Nora: [sighs] Every single time. 'We've had a change of heart.' Please.
Dex: Alright, so just to sum up: Harvard is selling AI professors for seven hundred bucks.
Nora: A British lab has an AI that might be better at science than the big guys.
Dex: And OpenAI has suddenly decided it loves the very regulations it used to hate.
Nora: Before we go, Dex. Saw a funding announcement that made me genuinely laugh out loud.
Dex: Oh, do tell.
Nora: A startup called 'Poetica' just raised two million dollars for a smart dog collar that translates your dog's barks into Shakespearean sonnets.
Dex: [laughs] No. NO WAY. 'Hark, what squeaky toy through yonder window breaks?'
Nora: I wish I were kidding. Two million dollars. The bubble is not back, it has achieved sentience and is mocking us.
Dex: And on that poetic note, that's Crash and Burn for Sunday, August 23rd. We'll be back tomorrow.
Nora: See you then. In the meantime, try not to get your pitch rejected by a chatbot with an Ivy League degree.
This show is made with AI: the hosts’ voices are synthetic and the scripts are AI-assisted. Every story links to its original source.