Latest / The Bleeding Edge Podcast / Why an AI Assistant With No App Is Worth $10 Billion
Transcript
- Ralph: And welcome to another episode of the Bleeding Edge Podcast. I'm your host, Ralph Benker, and as ever with me in the studio today, Emile Ogier. Emile, how is it on the edge? Em: Awesome, Ralph. Ralph: Well, we've got lots and lots of interesting stories to tell and a couple of deep dives, so let's get into the top five. Anthropic lines up a mid-November IPO and at the same time they have borrowed forty two billion from Broadcom. Basically it's a chip lease and essentially they will basically be showing what their finances look like. and of course they've reduced Prices. they've borrowed this 42 billion. And I think it's left us all wondering whether they can pay the bill. I think we called out in the show last week they were doing 64 billion in turnover. And so it's a good moment really for them to be doing their IPO. They're probably the most successful of the Frontier Labs at the moment. And just to put it in perspective, Deep Seek release their numbers. this week and they are doing a billion in turnover and they are also asking for seven and a half billion in funding. So at 64 billion, that's obviously mega cash. And let's see if they actually do move forward with this IPO in November. But the fact that their chip supplier is lending them the money to basically rent their chips right so here we go again with the circular economy. this is kind of financing at a scale that I think we've rarely seen anywhere other than infrastructure build out. but of course all of this is going to get very closely examined in the IPO because I think it will be the most comprehensive model of the commercial well what a a successful commercial model might look like within a frontier model. And it's certainly not quite there with open AI. Their spending has been rampant. they're only doing around forty billion. And and so I think actually getting a a good look under the covers of what you know Anthropic are actually doing in terms of where they're spending their money, where they're making their money, and whether this thing can survive. so the second story is the frontier price war. Well, I think we covered this a little bit last week. But Claude Opus five point five and GPT six arrived and both labs cut their prices. And so there is a real shift in terms of the cost. certainly if you are using an API to access these models. still incredibly expensive to use. but Opus five point five is such a capable model. I have really struggled. To move away from it over the last few weeks, especially as I've been building out the Bleeding Edge podcast website, which you can now actually find at bleeding hyphen edge hyphenai.com, which I know will make Craig Smith incredibly happy. And you should go away and have a look at that because there is lots of good stuff on that. then the third story Australia says an open AI agent hacked their Medicare portal. And the Prime Minister raised it with Altman at the UN. and so the sources don't say whether a human attacker was directing the agent or the agent behaved on its own. So why is this interesting at all? Well, I mean, we would love to know the details behind it, but I think one of the things that's happening is we're getting a you know, a head of government is basically holding a model provider. Personally responsible for an agent's actions. and not just the model provider. He's going to the CEO of this company and going, Hey, what's going on? not something we haven't seen before when there are big mishaps like this. But just interesting that I think these guys are going to be at the front line and being questioned by heads of government Em: speaking of personal assistance, I love a good David and Goliath story. We're familiar with the big Titans and the products they're releasing, but are you familiar with Instinct, Rolf? Ralph: No. Em: Okay, so well, this is a tiny team with a very big ambition. Instinct is the one that seems to have appeared suddenly in everyone's conversations, and behind it is Noah Shin. He's reported to be only 23 years old His company is called Spear Street Technology. The product attracted public attention in late August, but it has been under development for months. It is a young company becoming visible very quickly rather than a business created out of thin air in a few weeks. The financing story is extraordinary, but the distinction matters. On the 28th September, Instinct announced a 1 billion Series C involving Sequoia Benchmark, valuing the company at 10 billion. A month earlier, 250 million round had valued it at 2.5 billion. 10 billion is a price investors put on the business. It isn't 10 billion in cash raise. The valuation quadrupled between those reported rounds. The appeal starts with a surprising familiar interface. Instinct's own description says you text or call it, and it uses a phone and a computer to do the work. you aren't learning a new dashboard before asking for help. That makes the front door lighter than a dedicated agent application, although there is substantial machinery behind the conversation. Its published examples include arranging airport transfer, finding a handyman and following up on a thread you've dropped. These are small jobs individually, but they are exactly the kind of loose ends that make a busy week feel heavier than it needs to be. For somebody who dislikes managing another app, having the request begin as a message is a real design advantage. The latest funding announcement also describes a phone concierge capability for more involved bookings and administrative calls and a trusted network through which agents can coordinate for friends and colleagues. I would keep that framed as the company's description of what it is building. It suggests a useful direction instead of five people separately negotiating the same dinner plan. Their assistants could do all the coordination for them. There is a revealing growth claim in Shin's September interview with Patrick on Invest Like the Best. The episode describes invite-only growth of 10% a day. That's compounding 10% a day Ralph That is the founder reported figure for an early phase without an independent audit If that rate held, the arithmetic would roughly double the base in a week. That explains the excitement and also why buying enough computing capacity becomes an urgent problem. It introduces a conversation as Shin's first long form discussion in a business. I think this is his first public interview due to his age as well. That supports the feeling that the founder has only recently stepped into public view. the more useful point is that we finally have account of what he's trying to build. for now, access is still described as early and invitation led only. And the public evidence is thinner than the valuation headlines. I see instinct as a promising choice for personal rants and coordination if you can get in. Then judge it on a handful of real jobs. Does it return a confirmed booking? Does it know when to ask? Does it explain a failure? instincts bet is that useful software can disappear into the conversations you already know how to have. If that works, the winning interface may simply be the one you remember how to use. The harder part will be staying reliable at a small, carefully managed, early audience turns into a much larger one. I think the most interesting number to watch next is how often a job finishes cleanly. A one billion check can buy capacity. Earning the right to handle somebody's everyday life takes repeated, uneventful success. And they he frames this as a race over the future software itself, and Shin describes having only months to complete. With the world's largest companies. That is striking tension. A very small newcomer is trying to become our default assistant before the companies already on our phones get there. he's twenty three years old and he's raised ten billion dollars. That's quite the story. Ralph: No, it's insane because how do you get to basically you know a billion a near nearly a billion in revenue is is what I'm reading in some of the news items. I don't know if that's quite right. So let me just see here. So the company raised two hundred and fifty million at a two point five billion eva valuation in August, bringing its total raised at the time to three hundred and fifty million just this week. They raised an additional billion in a in a C series, which is what you were talking about. Now Em: Correct. Ralph: it's valued at ten billion. Okay. and if I look at the news it's saying Instinct is a free personal assistant powered by AI. So Em: it it is free, it's invite only, it's in beta, and that's also most probably one of the reasons why it's performing so well, So I think we need to keep an eye on it, but as things are working, people are reporting good things and you know, saying it's getting the job done. Ralph: So it's it also goes on to say instinct is now approaching one billion in annual transaction volume. So I don't that that's a it's a bit of a funny wording. It doesn't sound like that's one billion in revenue. so are they actually making any money at all? But you know, interesting line. We spent zero dollars on marketing so far. So just incredible story to be twenty-three years old and you know. Heading up a company that's worth 10 billion. I did find it also quite interesting, just catching up on the news about instinct, Emile. Alexander Wang, who you had mentioned from Meta was apparently making fun of this particular tool. and that was just ahead of him actually launching Amuse. So no doubt that actually attracted some attention. he probably that was probably his marketing. So very interesting story. I love that. It's always great to be inspired by young founders. You know, my son is fourteen years old and he keeps telling me I'm just fourteen and I keep saying, But that's just a number, you know, literally you could behind the next billion dollar AI company. But anyway, I won't hold it against him. Let's see, when he's twenty three. If he hasn't done it then I'll judge him. Em: But it's inspirational nonetheless. hearing that you can literally come from nowhere and be the next big thing because there's so much capability at your fingertips now. There's nothing holding you back apart from a good idea and putting in the gruntwork. And even there the gruntwork is less now, arguably. Ralph: It's so fascinating and obviously this space in particular, you know, these personal assistants, I think that's where it's gonna grow up. let's start talking about mainframes. you know, I did see a really interesting TV show the other day and it did make me think that maybe mainframes are going to come back. And so one of the things this the the show it was just a a T V show. And somebody was trying to steal a secret. And the company actually had an old mainframe that was locked behind a steel door. And when the hacker arrived to try and extract the data from it, they found that this machine only took those floppy discs. You know, those so basically it had a total storage capacity of about one point four megs. And obviously they were thwarted because they literally couldn't get any data out of it. What was so brilliant about the show was they said, okay, so this old technology is making a comeback for companies to store their biggest secrets on because they're completely air gapped. They're not connected to the internet. They don't even know what the internet is. and so they're completely secure. They're the most secure systems on the planet. And I just thought, wow, what an amazing use case to bring back some of that old tech. So if you've got an AS four hundred sitting in your basement somewhere, let's start a secure vault for somebody's biggest secrets. Emile Em: I c I cringe. I used to work on AS four hundred systems. I never want to go back there, thanks. Ralph: Well, I loved it because it was my first experience of reading RPG and I felt like a real hero. Em: Mm. Ralph: in fact my claim to fame, which not is not a claim to fame at all, but you know, our listeners might remember that amazing moment that was the year two thousand and I spent the best part of a year literally going line by line through code with developers. making sure that the system that we were looking at wasn't going to break. and it was an absolute baptism by fire in in terms of architecture and code. and I took an entire retail company apart line by line to basically find the 26 events that were gonna cause us some problems and we found a bunch of problems. it was probably not the most interesting work for a lot of people, but I really enjoyed it. but as you say, I never want to go back there again. let's rather talk about AI and how we're gonna leapfrog absolutely everything, including the existing operating models. But with that, I think that has been a really exciting episode. Emile, thanks for a a great show. Em: and thanks to all and please like and subscribe. Take care. Ralph: See you in the edge.