# TITV One Year Anniversary Show

Markdown copy of https://realalpha.app/v/veG9nEN_kTQ/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy.

Jeanne DeWitt Grosser and Tomasz Tunguz say AI will reshape company teams and software, while specialized models will drive much of enterprise adoption.

- Show: The Information's TITV
- Hosts: [Akash Pasricha](https://realalpha.app/people/akash-pasricha/)
- Guest: [Jeanne DeWitt Grosser](https://realalpha.app/people/jeanne-dewitt-grosser/)
- Guest: [Tomasz Tunguz](https://realalpha.app/people/tomasz-tunguz/)
- Guest: [Jesse Zhang](https://realalpha.app/builders/jesse-zhang/)
- Guest: [Rocket Drew](https://realalpha.app/people/rocket-drew/)
- Guest: [Adrian McNeil](https://realalpha.app/people/adrian-mcneil/)
- Guest: [Baiju Bhatt](https://realalpha.app/builders/baiju-bhatt/)
- Published: Jul 15, 2026
- Length: 1 h 39 m

## The best minutes

- [24:50](https://realalpha.app/v/veG9nEN_kTQ/?t=1490) [Open source pays off once AI workflows are stable](https://realalpha.app/s/open-source-pays-off-once-ai-workflows-are-stable/) (7 min)
- [1:02:31](https://realalpha.app/v/veG9nEN_kTQ/?t=3751) [Robots need useful real-world data, not just more data](https://realalpha.app/s/robots-need-useful-real-world-data-not-just-more-data/) (4 min)
- [1:10:30](https://realalpha.app/v/veG9nEN_kTQ/?t=4230) [Cowboy plans to make a rocket’s upper stage part of its data center](https://realalpha.app/s/cowboy-plans-to-make-a-rockets-upper-stage-part-of-its-data/) (6 min)

## Also worth watching

- [48:24](https://realalpha.app/v/veG9nEN_kTQ/?t=2904) [Recursive self-improvement does not require continual learning](https://realalpha.app/s/recursive-self-improvement-does-not-require-continual/) (4 min)
- [1:16:55](https://realalpha.app/v/veG9nEN_kTQ/?t=4615) [Cowboy recruits engineers who built and raced college cars](https://realalpha.app/s/cowboy-recruits-engineers-who-built-and-raced-college-cars/) (3 min)

Skip the anniversary setup, Decagon's company updates, the robotics model debate, later space-industry discussion, and the closing AI policy debate.

## Full analysis: Jesse Zhang says Decagon is pretty much 90-plus percent open source

- Decagon is pretty much 90-plus percent open source, Jesse Zhang said.
- That figure lacks a defined denominator, while tuning effort affects when open models pay.
- Budget for task-level tests, model adaptation, and sourcing rules before moving stable workflows.

Its 6 sections, what it means for you and what to watch are on [the page](https://realalpha.app/v/veG9nEN_kTQ/#analysis), free with an account:

- [25:52](https://realalpha.app/v/veG9nEN_kTQ/?t=1552) Decagon's open-model share followed workflow stabilization
- [27:38](https://realalpha.app/v/veG9nEN_kTQ/?t=1658) Tuning makes open-model savings conditional
- [27:25](https://realalpha.app/v/veG9nEN_kTQ/?t=1645) Frontier models retain value in unsettled work
- [37:33](https://realalpha.app/v/veG9nEN_kTQ/?t=2253) Decagon splits model classes by task shape
- [29:22](https://realalpha.app/v/veG9nEN_kTQ/?t=1762) Sourcing rules can outweigh model price
- [36:03](https://realalpha.app/v/veG9nEN_kTQ/?t=2163) Task-level evaluation should precede model switching

## Summary

Jeanne DeWitt Grosser is Vercel’s COO and previously spent nearly a decade at Stripe, where she built a go-to-market team; the episode also notes she had joined Shopify’s board. Tomasz Tunguz is a general partner at Theory Ventures and said he previously spent nearly 15 years at Redpoint. Jesse Zhang is Decagon’s CEO; Rocket Drew covers AI and robotics for The Information; Adrian McNeil co-founded and leads Foxglove; Baiju Bhatt co-founded Robinhood and leads Cowboy Space Corporation.

Grosser and Tunguz argue that AI is changing staffing and software workflows, while companies will use a mix of models and need to manage cost, security and trust. Zhang says open-source models fit scaled, specialized tasks, while frontier models remain useful for experimentation and broader work. Drew says continual learning remains unsolved but could help AI move beyond narrow tasks, while also making researchers more productive or automating parts of their work. McNeil sees robotics as a fragmented market where real-world data and reliability matter; Bhatt says Cowboy plans to use a rocket’s upper stage as part of an orbital data center. The Information’s Amir Efrati and Jason Dean discuss rising pressure for AI regulation and concerns about trust in large AI labs.

### Minute by minute

- [25:03](https://realalpha.app/v/veG9nEN_kTQ/?t=1503) **Open models suit scaled tasks.** Zhang says open-source usage is growing, but its share relative to closed-source models is shrinking because many enterprise use cases are still experimental. Decagon is more than 90% open source, he says, because its customer-service tasks are established and scaled; open models can improve latency and cost, but require work to match frontier-model performance.
- [31:17](https://realalpha.app/v/veG9nEN_kTQ/?t=1877) **Decagon expands its agent platform.** Zhang says Decagon grew from about 50 to 500 people in a year and is broadening from customer service toward a conversational way for businesses to interact with customers. Its Duet product pairs a fast customer-facing agent with a slower agent that reviews conversations, identifies problems and can update workflows. Zhang says targeted customer-service outcomes make evaluation relatively clear.
- [38:27](https://realalpha.app/v/veG9nEN_kTQ/?t=2307) **Continual learning could widen AI’s role.** Zhang says continual learning could let models learn from their own performance as pretraining data becomes limited. He says Decagon works to improve the whole agent using customer interactions and expects researchers to remain important for training open-source models. The company aims to expand from inbound support to more kinds of customer conversations.
- [44:35](https://realalpha.app/v/veG9nEN_kTQ/?t=2675) **AI still struggles to learn on the job.** Drew says today’s models have fixed weights after training; notes and fine-tuning are imperfect substitutes that can fail to transfer skills or cause forgetting. She distinguishes continual learning from recursive self-improvement, where AI could train successors under the current training paradigm. She says continual learning could move AI from narrow tasks toward more autonomous work, while AI researchers may also automate more of their own process.
- [56:22](https://realalpha.app/v/veG9nEN_kTQ/?t=3382) **Physical AI will remain fragmented.** McNeil expects a long tail of robotics companies because factories, homes, drones and self-driving cars require different systems and often run on limited onboard compute. He says task-specific open models may have an advantage, while vision-language-action and world-model approaches are both still developing.
- [1:02:31](https://realalpha.app/v/veG9nEN_kTQ/?t=3751) **Robots need real-world reliability.** McNeil says robotics needs high-quality demonstrations and a way to learn from rare real-world events, not just more data; simulation still has a gap with the physical world. He calls homes especially difficult to model and says humanoids may suit long-tail tasks, but a 95% lab success rate can still fail in production.
- [1:09:12](https://realalpha.app/v/veG9nEN_kTQ/?t=4152) **Cowboy plans an orbital data center.** Bhatt says Cowboy plans to keep a rocket’s upper stage in orbit and use its mass and volume as part of the data center’s heat-dissipation system. He says the company is targeting its first rocket launch by the end of 2028, a six-year useful life, and an FCC application for 20,000 satellites at about one megawatt each.
- [1:16:55](https://realalpha.app/v/veG9nEN_kTQ/?t=4615) **Real-world tests drive space development.** Bhatt says repeated real-world testing is necessary to build reliability and lower costs, and points to Formula SAE graduates as a talent pool used to building and testing hardware. He sees orbital data centers as a national-security concern and says Cowboy is developing optical communications. He also says AI tools have helped him learn rocket engineering and support calculations.
- [1:28:39](https://realalpha.app/v/veG9nEN_kTQ/?t=5319) **AI policy debates are returning.** Amir Efrati says model testing and risk discussions have returned after an earlier period with little action, and expects regulation to intensify as capabilities advance. He says officials may start by restricting government contractors’ use of Chinese models, while open-source systems raise difficult questions because they are widely available. Jason Dean says concerns about data retention and the power of young AI companies are likely to persist.
- [1:33:16](https://realalpha.app/v/veG9nEN_kTQ/?t=5596) **Trust concerns will not stop demand.** Dean says companies have genuine concerns about data retention and AI-lab concentration, while competitors also have incentives to emphasize those risks. Efrati says a fear-based pitch needs a stronger value proposition, and that customers may keep choosing the best product. The editors expect AI’s wider effects on technology and business to remain a defining topic.

## Transcript

The Information's TITV, hosted by Akash Pasricha, with Jeanne DeWitt Grosser, Tomasz Tunguz, Jesse Zhang, Rocket Drew, Adrian McNeil and Baiju Bhatt.

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

[1:01](https://realalpha.app/v/veG9nEN_kTQ/?t=61) Welcome, everyone, to The Information's TITV. The information was first to report just moments ago. The information reported exclusively... Welcome, everyone, to this special edition of The Information's TITV. My name is Akash Mastrija. It is Tuesday, July 14th, and we are coming to you from our San Francisco office where we are celebrating the one-year anniversary of our show, and what a crazy year it has been. When we first started this show, there was no such thing as fable. There was barely any conversation about TPUs or Tranium. SpaceX and X and XAI were all still separate private companies.

[1:47](https://realalpha.app/v/veG9nEN_kTQ/?t=107) There was no open claw. No one cared about AI margins, and orbital data centers were very much still a distant dream. Some things never change. This is our 250th episode. And today, rather than looking back on it all, we're going to look ahead to what we think will be the big themes that are going to dominate our next 250 shows. We're going to start with a conversation about what the AI company of 2030 is going to look like. We're then going to take a deep dive into the tech itself with a discussion on open source and also all the challenges that AI researchers still haven't solved. We're also going to talk about the big questions facing physical AI.

[2:27](https://realalpha.app/v/veG9nEN_kTQ/?t=147) and the space industry. And we're going to end with some healthy debate with two of our top editors. We've got some great guests coming on. The CEO of Decagon is joining us. We've got a top robotics founder coming on. The co-founder of Robinhood, who is now building Cowboy Space Corporation, is joining us in a bit. But I want to start with two of my favorite guests from the past year. Jean Grocer is COO at Bricell. She is on the front lines of the AI coding craziness. Tomasz Tunguz is a general partner at Theory Ventures. He is one of the most frequent guests on our show. Welcome to the both of you.

[3:01](https://realalpha.app/v/veG9nEN_kTQ/?t=181) It is so great to have you here. Thanks for having us. Thrilled to be here. So I want to start, Gene, with a question about what the AI company of 2030 looks like. And you have such an interesting perspective on this because you were at Stripe for nearly a decade. You built an entire go-to-market team over there. And now you're building an AI-native go-to-market team. So what does the team look like nowadays? How is it different from Stripe? I mean, there's a lot of places where we're getting a lot of leverage out of AI. So I would say the mix of roles within a company is shifting. As an example, we obviously are growing headcount still, high-growth company.

[3:41](https://realalpha.app/v/veG9nEN_kTQ/?t=221) But within go-to-market, the percent year-on-year growth and headcount by subfunctions was meaningfully different at the start of this year. So an example, sales development that we grew, but less than our account executives. Typically, those might grow more in tandem. We've actually been able to take our support headcount down year on year, despite more than doubling the total number of customers that we have. And actually, you know, our H1 first half of the year ends at the end of this month. And so we're sort of doing an H2 replanning because it's only been five months, but our plan already is kind of just feels like agent history.

[4:20](https://realalpha.app/v/veG9nEN_kTQ/?t=260) And one of the provocations G always has is like, we can't get this company to have more than 1,024. G is the, this is G on the. Yeah, Guillermo. Guillermo. Founder, CEO. But his provocation is Vercel can never employ more than 1,024 people, which is 2 to the 10th. And how many people does it employ right now? Eight something. We're getting there. We're getting there. So are we replacing all SDR salespeople with AI? What does this look like? No. I mean, we're getting meaningfully more scale. And so both more productivity because the humans themselves can do things much more efficiently. And actually.

[4:59](https://realalpha.app/v/veG9nEN_kTQ/?t=299) more productivity due to better outcomes because there are many ways in which we're now getting AI to sort of be a 99th percentile performer 99% of the time. Whereas, you know, if you have a 30 person sales development function, you're going to have a bell curve, if you will, on how they perform. Right. It is really interesting. Those were the first two use cases in AI, right? Like automated sales and automated customer support. So you're really starting to see a lot of success from it. Absolutely. Like we, where we started was pick functions that are more deterministic, Right. So if it's a bit more if then how you would handle something that, you know, even August of last year, I could do reasonably well with a fair amount of our internal context around it.

[5:38](https://realalpha.app/v/veG9nEN_kTQ/?t=338) And now, you know, as of August of last year, we had 90 percent automation of sales development for inbound and our support agent that we've home built handles 91 percent of all support cases. And you can imagine, like, for sales, they're technical. Like, these are not, why is the button blue, not green, right? Tomas, are you using, you know, recruiting? I mean, like, are you changing the way you hire now, depending on, you know, what role you're hiring? You were at Redpoint for, what, nearly 15 years. That must have been a totally different time for venture capital. It is a very different time.

[6:13](https://realalpha.app/v/veG9nEN_kTQ/?t=373) I think we are trying to automate as much of the business as we can. We have more engineers on staff than we have investors by two to one. I think that will continue for the life of the firm. And I think we were able to analyze in much greater depth about three to four times the number of companies year on year as a result of some of the agents. It's not so deterministic as sales development. I think there's a lot more judgment, obviously, but there are a lot of first-pass filters you can look at. For example, we don't invest in India, so it's very easy to filter out those kinds of companies, or we may not invest in B2C businesses, consumer companies.

[6:48](https://realalpha.app/v/veG9nEN_kTQ/?t=408) And so we can parse through many more companies. And then there's the all, I mean, the generation of investment memos, the creation of market maps, the analysis of podcasts, transcriptions, many, many different functions. So, okay, so let's get into then, I mean, clearly there are results with AI, but I want to talk about some of the more difficult parts of using the technology. I mean, cost is one component. And, Gene, I want to start with you. I mean, how are you handling not just the cost of using AI internally at Vercel, but how are you also seeing your customers think about their costs? Because, I mean, the models are expensive.

[7:27](https://realalpha.app/v/veG9nEN_kTQ/?t=447) Yeah. So, for us, sort of, like, twofold. One, obviously, we're paying attention to which models we're using and we'll switch them out as a less expensive model is more equally performant. But two, all the agents that we're building internally, we're building on Vercel's infrastructure, which is actually just very inexpensive to use overall. So we've optimized things like fluid compute specifically for agentic-type workflows. flows. So one of the things I struggle with is I'll tell people about the support use case or the sales agent use case. And in both of those cases, the total annual infrastructure bill is in the single digit thousands. This is your bill?

[8:09](https://realalpha.app/v/veG9nEN_kTQ/?t=489) Our bill internally, if Vercel works. Because you chose them to do it yourself. On Vercel's infrastructure that anyone can use. So that's, in most cases for the work we've done, And like our sales development agent is a 32X ROI. So it's worth it no matter how much that model costs. For our customers, we're actually seeing really interesting things. So we built the AI gateway, which is like a token development or token delivery network is what we call it. But let's use switch between models. And we see really interesting data. As an example, folks that are running AI in production are now on average using 32 different models.

[8:49](https://realalpha.app/v/veG9nEN_kTQ/?t=529) So folks aren't just going to Anthropic and saying, give me Fable, right? They're actually picking the right model for the task. And is that because of cost? I think that's part of it, because you're also seeing. a very real rise in open-weight models that people are now seeing, you know, good performance outcomes from. And so you'll switch those in, and you can do it with a, you know, single line of code change, run it, tweak it, and, you know, reduce costs by 90%. Tomas, what about your portfolio companies? Are they using 32 different models as well? Not that many. We use a lot of different models internally.

[9:23](https://realalpha.app/v/veG9nEN_kTQ/?t=563) Okay. I think of all different sizes and shapes, so 8 billion parameters, 30 billion parameters. We use Fable. And so I think the model diversity is inevitable. Particularly, we call it the iPhone 15 moment. You can pick your iPhone where at some point you kind of stop paying attention to the features and how much better the camera was and the phone just kind of worked. And from a lot of business work, many of these models are interchangeable, even at radically different sizes, as long as you have the harness or the systems around it to get it to work. So, Tomas, where do you stand then on the open-source, closed-source debate?

[9:55](https://realalpha.app/v/veG9nEN_kTQ/?t=595) Is open-source the future? What do you think? It's an incredibly important part of the future. It won't be everything. When you look at open router data, 30% of model consumption today is open source. I think it'll be between 30% to 50%. Of total model usage. Of total tokens over time. There's a nine to one cost ratio, right? You can save 90% by using an open source model. Yeah. And then we're big believers in local. So Stanford published a study and we've replicated internally that for business functions, non-coding functions, somewhere between 60% to 80% of your work can be completed on a modern MacBook.

[10:28](https://realalpha.app/v/veG9nEN_kTQ/?t=628) What percent of the work that happens on Bracel's platform happens using open source models? Well, that's a good question. We actually have that in the AI Gateway report that we published. Okay. I just don't have it memorized. That's okay. That's fine. But I do know it's been aggressively up and to the right, actually. I think Guillermo had a tweet recently that showed a time series data of January through now of how Anthropic, Gemini, open weight models all were shifting as new ones were released. And definitely open source is growing fast. I want to ask you guys about the SaaSpocalypse and the extent to which enterprise software still remains sticky, useful, and an opportunity for people to build.

[11:12](https://realalpha.app/v/veG9nEN_kTQ/?t=672) Jean, I want to ask you about your customers. So, I mean, look, you can use For Shell's platform to build a lot of this stuff. It's very good for that. Yep. At the same time, there's a question, Do I want to spend all my time building? I mean, I was going to CRM. CRM seems like a lower lift, but do I want to create tools for myself? How are your customers thinking about that? I think what we're seeing is a lot of a trend towards headless software. So we think of this, first place to go headless was e-com because no one wants two exact same sites, right? But your back end is still the same to keep everything well structured.

[11:52](https://realalpha.app/v/veG9nEN_kTQ/?t=712) And I would argue that the same actually was always true for CRM. The way you sell and I sell is pretty darn different, but we were sort of forced into the same model using Salesforce. So one of the things we've done at Bursell is actually we've built our own front end for Salesforce. We're still using Salesforce as a back end. It's a system of record, but it now does the workflow we want it to do. And so we're getting a lot of productivity gains because it's a combination of the UI that we always wanted and then insert agents to make that productive as well. So I think you're going to see a lot of companies go that path, particularly for things like CRM.

[12:28](https://realalpha.app/v/veG9nEN_kTQ/?t=748) Things that are more regulated, I don't know that you're necessarily going to want to, like, mess up your taxes or stuff like that. Like, you might leave that a bit more to a full-fledged system or record. And then I think also just as this next sort of era of software continues to mature, you'll probably see it be more flexible out of the box in general. So there won't be quite as much of a bias in some places towards building as there is right now, or I think it pays if you're 90th percentile to be out there building. So one example where everybody's like, I'm not going to buy a payroll app because I'll end up in jail.

[13:05](https://realalpha.app/v/veG9nEN_kTQ/?t=785) So what about your portfolio of companies? Are they still buying from the big enterprise software giants? A lot of them are doing what Gene described, which is headless. There's a back-end system, and then they build systems on top. They're very agentic. And so when you say headless, just so I understand, I mean, I've never built a headless system before. So, I mean, do you buy something? Buy a CRM. Okay. And then instead of logging into login.crm.com, you create maybe something on Vercel that is like TheoryCRM. And that UI, that user interface, is customized to a venture capital firm, but it ends up hitting or using the underlying Salesforce.

[13:42](https://realalpha.app/v/veG9nEN_kTQ/?t=822) So this, if I'm understanding you correctly, this is less so buying the application layer, it's more buying what's underneath. Yeah. I mean, today you wind up buying the application layer because most of these software companies have not started monetizing their MCP server. So like you're still buying seats. That's what we're doing with Salesforce. We still pay for seats, but we're just using their backend. We'll hit their APIs to store access data, et cetera. And then what a salesperson would see is custom Vercel UI. That's how we want you working. All right. Tamash, when you, and I ask you this because I know that you at your firm do a lot of research into customers at enterprises and what they're talking about, how they're feeling. What do they want more of? Is it just a better model? Is it a better agent? I mean, that seems like too simple in terms of AI. What is it going to take to increase adoption?

[14:36](https://realalpha.app/v/veG9nEN_kTQ/?t=876) Oh, I think, well, I don't think they're worried about adoption. Most of the CIOs and CISOs I speak to see a tsunami of adoption already. They are worried about costs. So will we miss earnings? If we're a publicly traded company, will we miss EPS as a result of uncontrolled costs? That's one. The second thing that they're really worried about, and Satya Nadella tweeted about it, and so did Alex Karp talked about it, was the loss of information going to some of these models that are being used for trading. That's another one. The third is just security. Security is number one, whether it's like data exfiltration. We saw a pretty significant supply chain attack when an open source library called Axiom was hacked and it impacted hundreds of millions of different software projects.

[15:22](https://realalpha.app/v/veG9nEN_kTQ/?t=922) Those are probably the three biggest priorities, cost, security, and... I couldn't agree more. We see this all the time. We actually just built a bunch of product for this in particular because you have a lot of folks by coding, but then wanting to go connect to your CRM. And so now you have all these secrets, API keys, that used to be cordoned off in your IT or your engineering team that people are accessing in a lot of cases and setting up connectivity to places you wouldn't normally want people to do. Same thing. Any sort of internal software needs an authentication layer. So you have everybody vibe coding auth, not that great of an idea.

[16:05](https://realalpha.app/v/veG9nEN_kTQ/?t=965) So we saw this actually at Vercel as well because we've sort of been on the bleeding edge of internal build and basically ended up building two products for the market now in Vercel Passport, which basically gives folks an auth layer that can go and make sure if you're doing a query to access customer data and you're in sales and I'm in finance, like, we shouldn't get the same answer. And then Vercel Connect, which is the same thing around don't let people have long-lived, you know, API keys accessing Snowflake, Salesforce, et cetera. Have those be minimally scoped and short-lived as well. Tomas, you talked about Satya Nadella's latest post on X, And I saw Mark Benioff also had a post, I think, in the last 24 hours talking.

[16:50](https://realalpha.app/v/veG9nEN_kTQ/?t=1010) He was really touting the zero data retention policy that they have. And, you know, Satya Nadella is out here talking about, we encourage... you to take control of your own AI, basically. And I mean, it all got me thinking about the extent to which AI or the AI labs right now have a trust problem. What do you think of that? Yeah, I mean, chatting with two different CISOs last week, the first thing they care about is something called ZDR, zero data retention. They have enforced in their companies that no employee can use a model that retains any data. And even then, there is some level of skepticism as to whether that is the case.

[17:34](https://realalpha.app/v/veG9nEN_kTQ/?t=1054) Even if they say. Yeah, I mean, I think there's just a healthy level of skepticism because, I mean, this data is unbelievably valuable. And we can look at it with the model releases even from last week. The share shifts meaningfully. Like Fable comes out, a bunch of people move. In fact, we published some analysis. There was a Stanford, an academic paper that showed that the average half-life of a model, you can look at retention curves for mobile games and social networks and software companies. We published this research today. And a model's half-life is somewhere between a social network and a mobile game.

[18:10](https://realalpha.app/v/veG9nEN_kTQ/?t=1090) You will lose more than 50% of your users by the end of month one. And so if there's that much movement between these models, the competitive dynamic to have a significantly better model is incredibly important. It provides you access to capital, to raise. It provides you an ability to market, really move shares. But I have to ask, I'd love to know actually, who was the sample for that? That open router data. Open router, okay. Because I'm thinking also about enterprise adoption here. I mean, it takes a long time to even use any of these models, right? And so I'm sort of thinking about maybe if that data is sort of the people on the bleeding edge or i don't think it is i mean you know cursor you can switch the model like this if any i mean you can tell claude like put opus in there put sonnet in there and then the other dynamic is a state-of-the-art model only remains a state-of-the-art for 41 days and so so you know the expression like a king is dead longer the king is basically a monthly chant at this point and that that shifts like that all you have to be on the latest one if you want to be a really productive software engineer, SDR, you want to test, do I have five or 10 percentage points

[19:22](https://realalpha.app/v/veG9nEN_kTQ/?t=1162) of better performance using this new model or not? So I will try it. Jean, do the AI labs have a trust problem? I mean, I think any, any large company that's growing this quickly and dealing with data, that's always going to come up, right? Like that, that's sort of been a truism. And a lot of these companies have had less time to mature and gain trust to begin with. But, you know, I thought Satya's points were solid, which is, in general, setting up a more agnostic approach to how you engage with models is going to also mean that strategically you can switch every 41 days. So, you know, I think we sort of see the same thing with some of the stuff that Broussel has been focusing on.

[20:03](https://realalpha.app/v/veG9nEN_kTQ/?t=1203) We launched the EVE framework for agents, which is an open source framework that's meant to enable you switching, having your own context, et cetera. So I agree with one of Satya's points because it's been our own experience, which is your context is what makes or breaks the output of the agent. So you take the sales development use case. There are 10 or more AISDR offerings on the market right now, and we're outperforming them. And that really, I think, is because of how, for our context, how we've thought through and gone the last mile of getting that into our agent. hmm tomash i want to switch briefly to advertising which is a space that you have invested in uh you've in fact brought uh in portfolio companies on our show and talked about yeah um what is the right user experience for advertising.

[21:00](https://realalpha.app/v/veG9nEN_kTQ/?t=1260) as it relates to AI and chatbots. How do you think about that? It's also being discovered. I think the potential is enormous. Google generates about $120 per user per year. As we both know, we're both Zooglers. Both of you? Yeah. Did you know each other? We did not know each other. No, no, no. Gene was much higher in the organization than I was. What teams were you on? I was on Gmail and then Google Cloud before it was called Google Cloud. And Tomat, you were on the ads? I was on the ads. Yeah, started in customer support. So did I, though. Great place to start. Yeah. But anyway, so I think there's a tremendous amount of potential.

[21:34](https://realalpha.app/v/veG9nEN_kTQ/?t=1294) One of the ways we think about it is AI ads really live in the middle to the bottom of the funnel. It's not just awareness like a billboard might be, but it really drives users to understand what is the right car for them, what is the right shoe. And the other dynamic is like today when you visit the information, there might be an ad, right? And that ad format is consistent across every user that visits. It might be a little bit customized. But within the world of AI, it can be uniquely customized because of the context of that particular user or that query. And so we think the ads, the creatives, or the advertisements that you see will change as a function of the user, very much like in e-commerce, the way that Gene was mentioning before, everybody's experience.

[22:13](https://realalpha.app/v/veG9nEN_kTQ/?t=1333) So are you bullish on OpenAI hitting its big advertising target in 2030? Yes. I mean, I think it's critical to the success of that business just given the fraction of consumers or the visitation patterns. And then also, it's important to offset the cost of inference just the way that Netflix has done and others in video. So I'm confident that they will. And I think you look at search ads, $250 billion market, social ads, $263. At the end of last year, AI ads will be, I think, much bigger. Gene, I know it's been a while since you were at Stripe, and you just joined the board of Shopify, which is an exciting new opportunity.

[22:51](https://realalpha.app/v/veG9nEN_kTQ/?t=1371) I mean, broadly speaking, just if you were to make a prediction about AI and commerce for the next year, what do you think is going to change the most? One of the really interesting dynamics we've been seeing is actually that it's exposing more of the long tail, right? So you can sort of do like a very specific search within AI and all of a sudden it's getting you access to, you know, the sort of torso and tail of shops that are selling your specific Japanese button down, you know, shirt that you wanted. So Neuron, that's been positive from a commerce perspective is just getting more exposure to the folks that don't have the hundred million dollar, you know, advertising budget to show up.

[23:35](https://realalpha.app/v/veG9nEN_kTQ/?t=1415) I think, like, what will be interesting is where people want to hand over the reins on commerce and actually let AI go complete the purchase versus I think people sometimes underestimate the degree to which people shop because they enjoy it and want to go look at it. It reminds me of, I mean, look, Brian Chesky has said that he's not sure if a chatbot is the right medium for shopping and commerce and, you know. There are a lot of like if you take I just was at an Airbnb this weekend and I would have searched for what I wanted fundamentally differently with AI than with the filtering method that you have now.

[24:13](https://realalpha.app/v/veG9nEN_kTQ/?t=1453) So I think you there could get me much more rapidly to the 10 house homes I actually want to look at. So I think it'll be like incredible for discovery for sure. And then, you know, degree to which people, again, sort of go off and say, OK, now complete that transaction versus I actually do want to peruse those 10 different homes or, you know, those five dresses, et cetera. Well, I want to thank you both for joining us. It was a great conversation. That is Gene Grosser from Vercel and Tomasz Tunguz from Theory Ventures here on TITV. Whether it's GPT or Sonnet or DeepSeek or Fable, the... Strength of AI models is one of the big topics everyone is tracking, irrespective of where adoption is at.

[25:06](https://realalpha.app/v/veG9nEN_kTQ/?t=1506) And while the companies behind closed-source models like OpenAI and Anthropic are battling it out, they are both brushing up against a lot of recent interest in open-source models. To unpack the current state of that race, I want to bring on Jesse Zhang, the CEO of Decagon. Jesse, welcome to the show. It's great to have you here. Thanks for having me. So you've been doing a lot of interesting writing on X, a lot of great posts. And one of your most recent posts really intrigued me. it was about open source models. And you wrote that everyone is wrong about open source AI in the enterprise. What are they wrong about?

[25:18](https://realalpha.app/v/veG9nEN_kTQ/?t=1518) Yeah, I would say to put it simply, it's, you know, there's a really strong narrative now that open source has taken over. It's getting a lot more adoption. You see this in a ton of metrics. You know, the usage of the models on Hugging Face, the spend on the inference providers. I mean, they're all ripping. So that is true. Open source is happening more, but then actually if you look at the research, a lot of the stats are saying that the share of open source compared to closed source is actually shrinking. And so we were just thinking why that is because we are a big proponent of open source.

[25:52](https://realalpha.app/v/veG9nEN_kTQ/?t=1552) So we are pretty much 90 plus percent open source at this point. Wow. Have you always been 90%? When we started, it was all closed source because you're just trying to figure out the use case, right? Yeah. But my thesis was, hey, you know, why is it the case that, you know, open source is growing, but then the share is shrinking?

[26:11](https://realalpha.app/v/veG9nEN_kTQ/?t=1571) **Jesse Zhang:** Is that, you know, simply put, open source is great for scale production use cases where you know the structure of the use case and you know the structure of your agent or the model calls.

[26:22](https://realalpha.app/v/veG9nEN_kTQ/?t=1582) Everything else, closed source makes more sense. And so if you look at the enterprise, most of the use cases are still in that experimentation phase or still in that sort of not fully productionized scaled out phase. And that's why closed source is still so popular. So the point that I sort of read from your post was this idea that when you come up with a task for AI to do, you want the cutting edge model. You want the frontier model because you don't know how much power, essentially, you're going to need to complete the task. The way I read what you were saying is that, you know, we had a lot of these use cases.

[27:01](https://realalpha.app/v/veG9nEN_kTQ/?t=1621) People sort of thought about, hey, I think I could do this. Why not use a frontier model? Now we're at the point, though, where they're saying, okay, well, maybe I don't need that. I can use the open source models. The question I have for you, though, is that, I mean, won't the applications of the technology continue to keep going? Like, we're going to keep finding new use cases, so frontier model use, conceivably, will still continue, right? Yeah, that's my main point. So it's not that, hey, open source is going to take over, and because they're smarter now, you don't need the frontier models.

[27:34](https://realalpha.app/v/veG9nEN_kTQ/?t=1654) I would say there's always going to be new use cases. And the other thing I wrote about is that the effort to get open source to perform at the same level as Frontier, actually, it does take effort. And that's not to say the effort's not worth it. I mean, again, we put a lot of effort in, and it's super worth it. Because once, you know, we do AI customer service, and so once that use case is solidified and you're already scaled in production, then open source makes a lot of sense, because it makes sense to put in that effort to train those models. Because those open source models, especially the small ones, you get a lot of benefit out of them, right?

[28:04](https://realalpha.app/v/veG9nEN_kTQ/?t=1684) You get better latency, which was the main reason for us. You get better cost as well. But out of the box, those models are just not good enough. So you have to put in a lot of work to make them good enough and make them be able to sub in for these frontier models in certain spots. And once you put in that work, now the results have to justify the work. And they only justify the work if you're deployed in production at scale. Most of the use cases, to your point, are not like that. So people are just going to continue using frontier models. Which models specifically are you guys using in the open source category?

[28:37](https://realalpha.app/v/veG9nEN_kTQ/?t=1717) It's a mix. So we have Gemma 4 is the main American one. A lot of the Chinese ones are quite good. Quinn, GLM. and we have our own evals to kind of decide which ones perform better in certain scenarios. Right. How do you, I mean, you know, the race here, the open source race here, I guess, between open source models coming out of China and those developed here in North America, I mean, you know, at one point this was a race that people were paying a lot of attention to. Now it's sort of accepted, I think, that, you know, those models are great. People are willing to use them. I mean, if you were to sort of make a prediction as to how you see that race playing out, Do you think that the issue of it is kind of behind us now and now we can just sort of use whatever we want?

[29:22](https://realalpha.app/v/veG9nEN_kTQ/?t=1762) I think it's really up to people's choices, right? Like we've talked to enterprises where, you know, super fairly their leadership has made a decision, hey, we only want to use models that are trained in the U.S. Because even the Chinese open source models, they're all hosted in the U.S. Your data's not going anywhere. So that's not a concern. But it's, hey, you know, maybe there's some bias in those models. Like we don't want those models in our system. And so that's kind of up to each enterprise. Other ones are very comfortable using open source models. I think China has taken a very open source forward approach because I think that's how they compete.

[29:55](https://realalpha.app/v/veG9nEN_kTQ/?t=1795) In the US, a lot of the leading labs, I think just if you're in their shoes, it makes sense to put your best models not open source because you make more money that way. So there is kind of that balance. And you also hear from China now that, okay maybe they want to kind of now that they've reached you know frontier level and in certain cases maybe they want to stop doing open source who knows we'll see but i think having a lot of choices is just better for everyone what do you think of meta's decision i mean meta initially the llama family of models and that was open source um muse unless i'm missing something that that is a closed source model and so this is sort of a a transition away from open source for the moment which is a choice why you know what do you think make of that decision um i mean again i think it makes sense if you're in their shoes in in the u.s that's the norm right all the all the leading models are closed source and if they're closed source kind of you you maintain more of your alpha and you get to make more money because people are paying for the right model it's like that's like a self-reinforcing loop you make you take that revenue you can that helps cover the cost of

[31:08](https://realalpha.app/v/veG9nEN_kTQ/?t=1868) trading the model and so that makes sense right and again in china like that those like if they did that no one in the u.s would use those models because people are afraid of their data going to china um so having open source having open source strategy there makes sense because it gets people to use your models give us an update on your company so we had you on the show almost a year ago now actually last summer it's been that long uh you were telling me before we started taping So you're at 500 people now as the company? Yeah, the company's grown very quickly. We're based here in SF, not too far away, but now we're- Your product suite has expanded quite a bit, right?

[31:46](https://realalpha.app/v/veG9nEN_kTQ/?t=1906) Yes, so when we first met, I think, I mean, a year ago, we were maybe a 10th the size. And we only started the company two and a half years ago. In terms of like revenue or? No, people. People, okay. Like 50 people maybe. And yeah, I mean, early on, it was AI customer service agent, right? That's the easiest way to think about us. I would say very much that's still our core focus, but we've kind of expanded the aperture a little bit to be, hey, our goal as Decagon is to help any business maximize their customer experience. So make it really easy for customers to interact with them, have a conversational agent that they can talk to at any point.

[32:24](https://realalpha.app/v/veG9nEN_kTQ/?t=1944) And it could be for customer service. It could be proactive as well. You could proactively ping someone to get them engaged or make sure they don't churn or whatever. And so we call that concept like a concierge. So it's like a new UX for the business. Right. How big is the company at in terms of revenue? We don't talk about revenue publicly, but, you know, we've been growing very quickly. So last year, we were like roughly five. This year, probably want to do roughly the same. Wow. 5X in 2025. Yes. Top line. There's a category of products that we've written about here. Customer support tickets where they can actually code up solutions.

[33:07](https://realalpha.app/v/veG9nEN_kTQ/?t=1987) That's something that we've written about in our newsletters. Have you guys expanded into that, looked into that at all? Yeah. So the concept there is, hey, you have an agent that is, you know, talking with the customer, right? So that's what that agent does. Maybe it's a voice agent or whatever. And maybe the outcome of that conversation is, hey, we actually need to fix something. So then you can have a slower, maybe smarter agent that's sitting behind it that it works with to get better or to fix things. And so that was actually one of our biggest launches this year. And so we call it duet because it's like two agents working together, right?

[33:44](https://realalpha.app/v/veG9nEN_kTQ/?t=2024) Right. so you're a fast agent a lot of the um you know new releases from opening anthropic like you know uh opus 4a fable like these big thinking models are not actually be that useful in your voice agent because they're way too slow they could take like you know 10 seconds to reply but they're actually really good for the for the second agent and so the second agent's goal is to you know look at what the first agent is doing look at these conversations to your point identify that oh yeah actually there is this area that we're not doing super well like here's how we would fix it right we will update your documentation or we'll change this tool that we wrote etc and its job is to just kind of almost be like a cloud code type thing for for the and actually do do the task but fix the problem so that that is something with duet that you have yeah so like what do people use duet for you know they could come in and just be like hey you know um i want to build these new workflows we don't really have good documentation for them but here here's like 10 000 previous conversations like you go and write these workflows for us right and so they'll go inside Decagon. They'll write these documents. We call them agent operating procedures to kind

[34:48](https://realalpha.app/v/veG9nEN_kTQ/?t=2088) of program the main agent what to do. Right. And then once you're out in production, it'll actually like create tests for you. It will read those conversations, every transcript as they come in, like no human will ever be able to do that. There's way too many. And it'll tell you, okay, this is going well, but this thing is broken. Like every time someone talks about this topic, we're struggling. So I've gone in and read all those conversations and I've come up with, oh actually here's here's how you get better and so now not only have i identified how to get better i've gone in and made the changes right so like do the work how do you then determine the success you know i'm going to evals here really you know uh evals with agents is a topic that i think people are really focused on these days i mean with with customer service it's sort of i guess it's sort of binary you either answer the inquiry or you didn't in a in a task like you just described, you know, creating workflows.

[35:39](https://realalpha.app/v/veG9nEN_kTQ/?t=2139) I mean, the effectiveness of those workflows becomes a question. So are the evals with agents, are they getting harder or are they getting easier right now with the way the technology is going? I think it depends on the space. I think in our space, evals are fairly clear because if someone comes in, you have an idea of like what is a successful conversation. So you can set up evals against those. And so we have a concept called simulations in our product where you can just simulate all sorts of different types of conversations. And even before you push an agent live, you can eval if it's working, right?

[36:15](https://realalpha.app/v/veG9nEN_kTQ/?t=2175) Same thing, once it's in production, you can like evaluate those conversations. So that's fairly simple. And are there any areas where it's getting more difficult? I mean, evals, it's a topic right now that, and it might not be in your business, but I guess from the people you talk to, I mean, evals, this is an area that people are really grappling with right now. i think the area that people are grappling with is more evaling the the base models themselves it's like hey is fable better or is musespark better or whatever right and no it's like really hard to say so that's why people have evals they have benchmarks And then people can come out and say, like, actually, no, this benchmark's not good because of X, Y, Z reason.

[36:56](https://realalpha.app/v/veG9nEN_kTQ/?t=2216) Like, that's the tough part. All right. Where it's really tough to measure if a model is good or not. But in our use case, it's like, it's so targeted, right? It's like one agent for one use case for one company. And so that company knows, here's what success looks like. And so the evals are a lot simpler in our use case, if that makes sense. So if 90% of the work that's happening on your platform right now is coming from open source models. What are the tasks then that you are using Fable for, 5.6 for? What are you using those models for specifically? So Duet, it's a good example, right? So Duet relies on the Frontier models.

[37:37](https://realalpha.app/v/veG9nEN_kTQ/?t=2257) If you kind of rewind, like, why did we make a push to use open source in the first place? It's because in our voice agent or our chat agent, latency does matter, right? If you're a customer and you're coming in, you're talking to the agent, And if it can be super snappy and just, you know, get to my problem faster, that's just a much better experience. Like, you don't want to be like, okay, give me a second, and then it's just waiting for 10 seconds. So that's why we moved to open source, because those are smaller models. They're faster. Those models are not going to be that good out of the box.

[38:05](https://realalpha.app/v/veG9nEN_kTQ/?t=2285) And so you have to fine-tune them. You have to train them so that they can perform the same tasks as you would need to. But, you know, it works, because in this voice agent, most of the steps along the way, like, one of the steps could be, like, you know, choose the topic. That model doesn't need to be a big model. You don't need to know a lot to do that task. All you have to do is be really good at choosing the topic. Right. So that's why open source works there. But for things like Duet, where it's much more general purpose, you're just asking it, hey, like, go read all my conversations and, like, do some work for me.

[38:36](https://realalpha.app/v/veG9nEN_kTQ/?t=2316) Using the intelligence doesn't make sense there. Right. So that's the tradeoff. I want to ask you about a couple of the more technical elements here. So, I mean, continual learning, recursive self-improvement. I mean, you're deep in the space. We just had ICML last week and all the researchers were talking all about these topics. I mean, what's your view on, you know, the continual learning problem here, the opportunity to get to there, the extent to which you actually need it to drive AI adoption? What do you think? um i think it's it's obviously very powerful i mean there's so much evidence that it works and also uh you know if if you look at the work that a lot of labs are doing there is a bit of a data limitation problem right like we've we've taken all the data that we can to to pre-train these models and well where do you go from there well the continual learning is great or just being able to kind of learn from actually the the model's performance then you kind of open up your your aperture to like a whole world of new type of data. And you could argue that's how humans learn as well, right? Like when you're little, you kind of get instruction of like, okay, here's how you like get started and whatever. But then from there, throughout life, you're kind of learning

[39:54](https://realalpha.app/v/veG9nEN_kTQ/?t=2394) through your own experience and your own mistakes. So you could argue that that's sort of naturally what should happen. That's at the model level. For us, we're maybe like one level higher in the in the app stack where we are trying to improve the entire agent as a whole. And so the type of learning there is less about improving the models. It's more about how can you, through our agent, having millions of customer service interactions for a certain business, through those interactions, figure out what can be improved and how to get better. Right. What's your view on the extent to which the role of AI researchers can be automated?

[40:37](https://realalpha.app/v/veG9nEN_kTQ/?t=2437) I think we're kind of far from that. Far from it. This is the anxiety. I'm not, it's not a concern I have. It's just, I'm telling you what the AI research is. I mean, there's more, there's always more to do. Yeah. We have a pretty sizable research team at this point. And it's a very expensive team, but we have some great people. And the reason is that, hey, we decided, unless we invest in... In these things, right, unless we invest in training open source models, our product is not going to perform state-of-the-art because we're going to have latency issues and we're going to have all these things that, you know, could be done better if we actually, you know, compile the data and train it.

[41:14](https://realalpha.app/v/veG9nEN_kTQ/?t=2474) So I think there's just going to be so many use cases for researchers to help with. I want to ask you a question. So Salesforce bought FIN. Have you guys gotten any acquisition offers? uh i mean we don't talk about those either but it's not it's not really top of mind for us like we don't really we're only two and a half years well what about fundraising i mean oh yeah i mean fundraising we have been very active yeah uh we've done five rounds so far and is there a sixth happening right now right now no but uh at some point soon so we just did our last round not too long ago how do you think about you know growth of the company so you talked you talked about expansion of the product suite beyond just the customer service agents. I mean, three years from now, where does Decagon, what does expansion look like? Is it new markets? Is it horizontal products? What's the ultimate ambition for you? So the ambition for us, as I meant, like the sort of the big vision here is, can we produce an agent or a set of agents that can just improve how every business interacts with their customer right so that that interaction layer is what we care about and you can almost think of it as as hey you know in the past people had websites and

[42:31](https://realalpha.app/v/veG9nEN_kTQ/?t=2551) then phones came out and everyone was using them so then a new ui was created which is the mobile app now there's almost a new ui after that which is a conversational agent it's a new way to interface with the business that doesn't mean mobile apps and websites are going away but it's it's it's a new way to interface with the business and so that's the the vision that we're building towards. So what does that mean? Well, one, we have to launch a lot of different types of conversations that an agent could have, not just inbound customer service, right? So that's where a lot of our product development is. There's a lot of other products you can build in this theme of making that, you know, interaction better, being able to understand the interactions well, being able to auto-learn, as you were saying, right? So, you know, one of our big launches with Duet was this concept of Duet autopilot where instead of you even having to go and ask at things, it just prepares a report every morning. You just wake up, you log in, and Duet tells you, yeah, you know, I read all the conversations from last day and I put together this report of, you know, what's going on. So those are the things that we're launching. And we just think

[43:37](https://realalpha.app/v/veG9nEN_kTQ/?t=2617) that that vision is just so large and there's like so much to do there that you're not going to run out of things to build anytime soon. Great. Well, Jesse, I want to thank you for coming on. That is Jesse Zhang, the CEO of Decagon, here on TI-TV. The AI talent wars over top researchers is likely to be another dominant story for many months to come. The core reason, of course, is that there are still some meaty questions AI researchers still need to solve. Can software cut inference costs? How quickly can agents improve themselves? Will recursive self-improvement become a reality? I want to bring on our AI and robotics reporter, Rocket Drew, for a conversation about all of those topics. Rocket, welcome to the show. It's great to have you here. Thanks so much. It's great to be here in person. So fun. I should say, I mean, I'm the visitor. You are based here. That's true. That's true. You're on my home turf.

[44:35](https://realalpha.app/v/veG9nEN_kTQ/?t=2675) There we go. Okay. So we just talked to Jesse Zhang about open source models. Okay. We We touched a little bit on recursive self-improvement. I mean, this topic of continual learning is one that we've been trying to wrap our heads around on the show. Can you help us understand what is included in continual learning and where is the research actually at right now? Yeah, absolutely. That's a great place to start because continual learning is a super hot. topic on the frontier of AI research right now. The basic idea is that if you have an AI that can continually learn, then it learns on the job. It picks up new knowledge, new information, new skills from experience just as it goes about working and completing tasks for you.

[45:16](https://realalpha.app/v/veG9nEN_kTQ/?t=2716) And we're talking about the models or the agents right now? Well, that's a good question. Let's say the models. Let's say we're being kind of purists. We really want the model to be able to adapt and flexibly pick up new information. Now, There are kind of workarounds right now to get your agent to sort of continually learn. And I'll explain that in a second. But I think some people might have a reaction to this idea that's like, well, can't chatbots already continually learn? Doesn't ChatGPT learn more about me as I use it? That's true to some extent. But as a reminder, the way AI models are trained these days is you have one big phase of training, and that's kind of a discrete period.

[45:52](https://realalpha.app/v/veG9nEN_kTQ/?t=2752) And then you just use the model. And basically the weights of the model, the brain of the model, are frozen while you're using it. I think of it like in stage one, you're forging the tool, and in stage two, you're using the tool. Continual learning reimagines this. It says, what if the tool could adapt and change as you use it to become better and better at the jobs? Now, there are some workarounds, but they're very, they're kind of janky. Frankly, I think that's the way to put it. Like when ChatGPT learns more about you over time, it's leaving little notes to itself that it can reference at a later date.

[46:22](https://realalpha.app/v/veG9nEN_kTQ/?t=2782) And like you mentioned, some of the agents have workarounds as well. They can leave notes to themselves, remind themselves how to solve a task. But it doesn't feel elegant. It's not like the way humans pick up new knowledge on the fly and from experience. It's like you hired an intern, and the intern works for a day. And by the end of the day, it's a little bit better at its job. But then it leaves a note to itself, and it wakes up the next day with amnesia, like Groundhog Day, and comes into the office and reads the notes that someone else left for it and tries to remember how to work. It doesn't really feel like the true solution to learning on the fly and flexibly incorporating new skills.

[46:58](https://realalpha.app/v/veG9nEN_kTQ/?t=2818) So where are we at then with continual learning? We can leave these notes. Are there applications today that incorporate continual learning the way researchers want? Or is this something that we are working towards? We're still working towards it for sure. I think right now we have these little workarounds and they will continue to get better and better. Some people think we really do need to go back to the drawing board and rethink the way AI models are designed and the ways that they're trained in order to really fundamentally improve their ability to continually learn. We have some promising directions to explore, but there are big limitations right now with all of the ways that we try to do continual learning.

[47:41](https://realalpha.app/v/veG9nEN_kTQ/?t=2861) So, for example, you could take a bunch of data from the AI doing jobs for you, and then you could fine tune on that data. But this is a really blunt way to go about it. You're just getting the AI to imitate everything that it's done and just sort of regurgitate it. And you also run the risk of the AI forgetting things that it used to know. When you say, what do you mean by imitate? As in just like repeat the things that it has done before rather than like flexibly incorporate new skills into the way it works. So, for example, think of it's like pretend, pretend continual. Exactly, exactly. Think about it as like I can write down all of the moves that a chess grandmaster makes and I can sort of fine tune on those moves, but it's not going to make me a chess grandmaster myself. I have to learn the principles behind the moves that the chess grandmaster did and learn the reasons they took those moves. That's what it would really take to have continual learning.

[48:36](https://realalpha.app/v/veG9nEN_kTQ/?t=2916) Okay. So now let's go back to recursive self-improvement. Is that a part of continual learning? Is it super continual learning? Where does that fall into it? I would think about them separately to start. There's ways that they're related, but I would think about them separately. And what I mean by that is even if we don't solve continual learning, so AI that can learn new experiences, learn from its experiences and and learn on the job.

[49:00](https://realalpha.app/v/veG9nEN_kTQ/?t=2940) **Rocket Drew:** You could still have recursive self-improvement, which is the idea that you could have one AI model create its successor, so train the next generation of AI, without any humans

[49:10](https://realalpha.app/v/veG9nEN_kTQ/?t=2950) involved, no humans in the loop at all. The entire research loop is done just by AI. You could do that under the current paradigm, where we still have one training phase, and then you put the AI models out into the world. The way that they're related is they're sort of both paths to creating a super-intelligent AI. As in, you could imagine that instead of recursive self-improvement, you could have a bunch of AI models out there in the world. They're working with different enterprises, they're working with different organizations, and they're picking up all sorts of new skills. And then somehow they're accumulating that knowledge and feeding it back into some more central model.

[49:46](https://realalpha.app/v/veG9nEN_kTQ/?t=2986) That's a way that models could also get very good very quickly, even if humans are still involved in the loop of training them to some extent. So I want to ask you about, you know, some of the tradeoffs here of having continuing learning, achieving recursive self-improvement, because, I mean, look, we talked with Stephanie yesterday. She came back from ICML, and she was telling us about how AI researchers are getting nervous that, hey, I mean, first it was AI coders that were nervous that AI could do their job. Now researchers are saying, maybe they could do my job as well. So you have a bit of anxiety there.

[50:22](https://realalpha.app/v/veG9nEN_kTQ/?t=3022) On the other hand, you have the issue of AI adoption and sort of some excitement that, well, maybe there's an opportunity here for continual learning and stuff like that, these advancements to improve adoption in the enterprise. To what extent do you think that achieving continual learning will help improve adoption? And to what extent do you think it might put AI researchers out of a job? I mean, two meaty questions. Definitely. I think AI researchers are right to be a little bit nervous right now about being automated. To your point, coding is just one piece of what a researcher does in their day.

[51:01](https://realalpha.app/v/veG9nEN_kTQ/?t=3061) They also design experiments. They come up with hypotheses. They run experiments. They interpret the results. They generate new hypotheses. I mean, that's the scientific method, right? Coding is the piece of that that's most automated right now. You saw Anthropic recently announced that Claude is writing 80% of their code. OpenAI recently said that of their research budget, the compute that goes to research internally, the amount spent on coding has increased 100 times over the past six months. So coding is definitely the frontier here, but AI can do more and more of the whole research process.

[51:32](https://realalpha.app/v/veG9nEN_kTQ/?t=3092) So I think AI researchers are right to be a little bit nervous. Another way to think of this, though, is they'll be able to get a lot more done. They'll be more productive than ever. For every one hour of time that they spend at work, they're getting a thousand hours of research done or what used to take a person a thousand hours. What about adoption in the enterprise? Adoption is huge. This is the big question. It is the big question. I mean, continual learning would be so insanely profitable compared to what we've seen for AI so far. And the reason goes back to that analogy I gave you about an intern, right?

[52:04](https://realalpha.app/v/veG9nEN_kTQ/?t=3124) Imagine AI right now is like an intern. Maybe you can teach it to do one specific task and it kind of mostly does that task right. But interns are not that useful, right? In a company, interns are kind of annoying. You hire interns not because they're useful. You're not being very tight. I'm sorry. I've been an intern. I know what I was like. You were a good intern. I'm telling you. Maybe. Maybe. I've been an intern a few times. I think I took a lot of work. But you hire interns because eventually they become employees. Eventually they learn the ropes and they can be productive later on. If AI right now is like an intern, it's continual learning that would let them become an employee someday.

[52:39](https://realalpha.app/v/veG9nEN_kTQ/?t=3159) Without continual learning, the thought is maybe they'll be just stuck in intern mode forever. So you can automate specific pieces of workflows, but you'll still have to have people hand-holding them and working around them. Right. So let me ask you one more question then. You know, we talk about the strength of frontier models and the extent to each... to which you need the full power that these frontier models have. Then you have, well, we only need the models to hit a certain threshold, essentially, right? I mean, this is the whole conversation around open source models too. Do we need the frontier model?

[53:14](https://realalpha.app/v/veG9nEN_kTQ/?t=3194) So my question is, if you look at the aggregate volume of work that we want AI to do, do we need continual learning to do all of these tasks? I mean, you know, the intern tasks. I mean, it's not like everyone needs to be a rocket scientist to succeed in the workplace, right? That's right. That's how I'd think about it, too. There are intern-sized tasks that you could have AI models handle. And for some of those, you don't even need the frontier models today. You can make do with a smaller open-source model that's maybe fine-tuned to be better at that specific task. But it feels like we should be able to do better.

[53:50](https://realalpha.app/v/veG9nEN_kTQ/?t=3230) It feels like there should be room to have an AI model that acts like an employee flexibly and can work autonomously on long-range tasks and perform them very well. So I put this take in sort of the bucket of cope or hopeful takes that continual learning won't be necessary to get to that point. It's like, well, maybe in practice, most tasks just didn't require that much intelligence. And you can basically carve up a workflow into these different discrete chunks and pass them off to different models. Something else that's in this bucket of kind of cope-like takes about continual learning, I think, is that scaling will just be enough.

[54:28](https://realalpha.app/v/veG9nEN_kTQ/?t=3268) That you'll get models that are smart enough in general that when it clocks in in the morning, even if it has amnesia, it's not intern level, it's employee level. So it's okay that it's forgetting. Or we'll just take all of the experiences of the internship and we'll just cram them in a context window. We'll have a context window that's longer and longer. So it doesn't really need to update its weights. It doesn't need to change its brain. it will just be basically reading all of the notes from everything that it's done and just brute force it will know enough to be able to work effectively. So I don't think any of these really feel like the solution to continual learning, but there are ways that we could get around that problem and not really need a true solution.

[55:06](https://realalpha.app/v/veG9nEN_kTQ/?t=3306) Great. Well, Rocket, I want to thank you for coming on. That is Rocket Drew, our AI and robotics reporter here at The Information. physical ai is where a lot of ai companies are focused their growth efforts right now whether it's software companies building world models or chip companies hoping to power the wave of robotics that everyone is anticipating adrian mcneil the co-founder and ceo of robotics company fox club sits at the center of that story i want to bring him on for a conversation about that sector and what is going on in the current moment adrian welcome to the show it's great to have you here in person thank you thanks for having me i was at the florist just yesterday and i saw a fox glove you did yes it is also known as digitalis yeah digitalis florist told almost that is the company name and when i told the florist that i was interviewing you and your company was called foxglove she said oh he probably named it that and did a robotics company because it was called digitalis yeah yeah that would have been a better name but the domain wasn't available so we went with a foxglove there you go okay i have this thing with uh with company names right where like they should not be tied to like what it is that you create because what you create as a company

[56:23](https://realalpha.app/v/veG9nEN_kTQ/?t=3383) evolves over time you sort of you pivot you shift but like names that you can kind of create and you just give meaning to them right like what is what is amazon what is apple these these names that you just start repeating so much so you can you can give meaning to a name that is that is not directly related so what is the meaning of your company remind us we had you on the show before You raised $40 million from Bessemer. Just give us the quick highlight. So this is software for robots, right? Yeah, right, exactly. So we built a data platform, data infrastructure for physical AI. So this helps you with the whole data flywheel, your logging journey from robots that are out there in the production, bringing that data back, evaluating it, understanding it, sifting through that.

[57:00](https://realalpha.app/v/veG9nEN_kTQ/?t=3420) Data and finding interesting, relevant events, debugging, triaging, and then ultimately building better models for your robots. And so you build your own models or whose models do you? We don't build the models, no. So we build the data platform that helps power all of this. So everything from the logging, the cloud data management, the debugging, the triage, the understanding. Most of our customers are building models. So we work with a lot of the top labs that are building pure foundation models, a lot of the humanoid companies. Like who? Which ones am I allowed to name? um we'll just leave it at like a lot of the top humanoid companies a lot of the top robot foundation model labs um a lot of the top self-driving especially newer generation self-driving customers um and then you also have this long tail of agriculture robots construction robots um you know a lot so you know one thing that i'm trying to sort of wrap my head around is in ai right now you've got these ai labs that are the giants right and we're sort of in this interesting moment where there's a question around trust of these ai labs you know they've gotten pretty big what i'm trying to figure out is will the same story play out in robotics might

[58:10](https://realalpha.app/v/veG9nEN_kTQ/?t=3490) you have these giant robotics labs i guess model developers that become sort of the kingpins of the of the sector yeah um and then does it turn into everyone else do you see the sector playing out that way? I actually think that we're going to see a really long tail of, I think we're going to see something closer to the explosion of SaaS businesses in the early 2010s where we're just fragmented, like thousands and thousands of robotics companies, because we talk about robotics and physical AI, like it's one thing, but what you're doing in a factory is so different from a drone, from a self-driving car, from sort of a household, like a residential robot.

[58:46](https://realalpha.app/v/veG9nEN_kTQ/?t=3526) There are so many different things that you need to go and tackle. And then you look at what made the big language model labs successful was the fact that you can run this giant overpowered model in the cloud that just does everything. When you think about physical AI, usually you're running on limited compute. You've got battery powered robots that are moving around. You've got like a small sort of Jetson or some type of embedded GPU. And so you don't want this giant overpowered like trillion parameter model, right? Like you need a model that is fine tuned for a particular task. And so what we're seeing is a lot of companies, there are labs doing real research on these foundation models And that is like a really important point. But I think when we see what we're going to see in physical AI is all of those models being fine tuned and tailored for thousands of use cases that never even occurred to you.

[59:32](https://realalpha.app/v/veG9nEN_kTQ/?t=3572) OK, so more fragmented. What about the open source versus closed source debate that's playing out in AI right now? Does that same debate exist in robotics? Yeah, the debate does exist. I think we're not there yet where there are either, you know, we don't have kind of models yet, either open or closed source that are generally available that are solving a lot of these problems. Like the labs are getting there. A lot of companies are making huge progress on it. But it's not like we have, we don't have kind of like a GPT-3 class sort of or GPT-4 class model available yet. um but i again the debate does exist and i think in physical ai open models do have more of a leg up than they did in language models remains to be seen like which is why is that because of this because of this nature of needing to fine-tune and and put them into particular you know if you're building an agriculture robot and its job is to go around and like pick weeds or something like that like that's the only thing you don't need to know how to fold laundry or make a bed like some of that grounding is useful but a fine-tuned model for a particular task is probably going to be able to outperform like a giant sort of centralized model. And then you also don't

[1:00:34](https://realalpha.app/v/veG9nEN_kTQ/?t=3634) have the cloud, you don't have like an API billing, right? Like you need the model to be deployed on the robot. So you don't just have this thing where I can stick Opus behind an API or stick Fable behind an API and charge you, you know, hundreds of dollars a month to use it. Right. Okay. So the other debate that's playing in physical AI right now is VLA versus world models. Where do you stand on that? Right. You got to pick a side. Do I have to pick a side? So for the longest time, the VLAs were basically the only game in town, right? So this was taking big, large vision language models, especially open vision language models, and then fine tuning them with action data.

[1:01:12](https://realalpha.app/v/veG9nEN_kTQ/?t=3672) So creating tokens. Most big LLMs or VLMs today are creating, they're auto aggressive models. So they're creating a sequence of tokens, tokens in, tokens out. And then we basically add some action tokens to that. So we're fine tuning that model based on action data. But ultimately the model was grounded in the world of images, like still image cams and text. And that's what the model kind of knows best. And then we have this kind of other cam, which is relatively new entrant, I would say, where we have world models, which have been grounded and they've been grounded in action data. A lot of them have been trained on simulation and things like this.

[1:01:44](https://realalpha.app/v/veG9nEN_kTQ/?t=3704) And then we're trying to add in action tokens to that. And we're saying, hey, what if we take this world model that's good at predicting what would happen if I drop something and it can create realistic videos of what the next frames would be? Can we ground that in action data so that we can actually like pick up tasks and manipulate them and things like that? Neither of those approaches have won yet, basically. Which do you think is more likely to it? World model is a little further away. World model is probably, yeah. A little further away though, right? Yeah, the space moves month to month, honestly, like it changes and people are having success with both approaches.

[1:02:18](https://realalpha.app/v/veG9nEN_kTQ/?t=3738) But I think the world model approach is certainly interesting, right? Like even maybe a more interesting approach that you're starting to see is some companies like I think generalists are one of the ones that most recently came out with a model that has been trained from scratch as a robotics foundation model, right? So they're not even starting with a base of an- Who's working on that? Generalist. Generalist, okay. Came out with one, and then, yeah. So I'm thinking about cost here. I mean, which of those two, which of these approaches is most expensive? How do you think about capital constraints here with training any of these?

[1:02:54](https://realalpha.app/v/veG9nEN_kTQ/?t=3774) All of them are, yeah. So the biggest challenge in robotics, right, is data. Right. And the thing that most people miss is it's not just volume of data. There's now hundreds of companies out there trying to sell data for a physical AI. But a lot of the data is not that usable, right? So you don't just need a lot of general. We're not just going to feed in. People are certainly trying to experiment with feeding in a whole lot of YouTube data, for example. But it's not just that data. We need high-quality data, demonstrations of robots doing things. you know the really important part for most companies is when you get beyond the first

[1:03:24](https://realalpha.app/v/veG9nEN_kTQ/?t=3804) **Adrian McNeil:** proof of concept how can you put robots out there in the real world and then learn from the data that you're getting on the particular tasks so you think about like a self-driving car example

[1:03:33](https://realalpha.app/v/veG9nEN_kTQ/?t=3813) right like what separates uh what separates companies that are driving driverless around san francisco today from people that are still testing what was the the hardest thing that took the industry you know 10 to 15 years to get over this hump of like taking the driver out of the car it was the ability to put those cars out in the real world and then learn from learn from that experience so look at all of this driving data discard the 99 of it that's just me driving along a straight road find that interesting like one percent or point one percent of like weird interesting things that happened turn that back into learnings turn that like turn that back into training for a fleet and then being able to know being able to like evaluate and validate a release to actually know that like that data had an impact yeah we've gotten better at driving in this particular and we know that across these like thousand other situations we haven't got any worse wasn't there i mean there was an approach using gaming to to simulate data is that always try this right but they tried it's not like simulation in general right like a lot of people um would really like it to be true that we can go build some robots in simulation and then we can deploy

[1:04:33](https://realalpha.app/v/veG9nEN_kTQ/?t=3873) in the real world and they will be successful and people are having limited success but you have of this big sim-to-real gap is what it's called, right? So you have like, hey, great, the robot completes the task 100% of times in simulation, but the real world is just different. It's more complicated. And some things, you know, in some ways, for example, self-driving cars are actually a little bit of an easier simulation problem than some of the things that we're trying to create in general purpose robots, right? Because why? so you think about driving i mean it's a really hard thing to simulate it's not easy but at least there are rules of the road you've got vehicles you've got cyclists you've got pedestrians if you look at enough data you can kind of predict what what those different actors are going to do over time um but it's relatively structured but then if you try thinking how do i simulate everything that could possibly happen in a household for example right like i kind of i frame residential robots as like the final boss of physical ai because you just have so many different tasks that I can take everything from cracking eggs to folding laundry to, you know, all of these different things. And then you need to deal with model. If you try to simulate it again, which is

[1:05:37](https://realalpha.app/v/veG9nEN_kTQ/?t=3937) talking about how do you model, model, you know, the properties of a sheet of a bed sheet or a clothing or things like this? How do I model the squishiness of a strawberry? Or these are just like really, really hard things to, to model. And so you just kind of have this like combinatorial explosion of, of complexity. Right. Let me ask you one last question before we let you go. humanoids yeah overhyped underhyped um i'm i'm long on humanoids i think humanoids are but you don't need a human aid for everything right you don't need a human this is the whole thing so the thing is um you got to think though about like skews cost money right creating like different robots special purpose robots cost money in almost any case you can think of a purpose-built robot could outperform a humanoid robot um if you went to the effort of creating a purpose-built robot and trained it for that particular task and in some some tasks that we see enough of in the world it's worth creating like driving a car for example it's better to just make a self-driving car than to build a humanoid that sits in the car and like drives the steering wheel you can just focus there's there's enough like volume there and purpose that you can create a robot to do that

[1:06:40](https://realalpha.app/v/veG9nEN_kTQ/?t=4000) but there's this long tail of tasks that it's not worth building a purpose-built robot and you think about a lot of stuff in factories you think about the the general challenge of being in a home for example where you've got stairs um and again i've people get really hung up on like humanoids specifically and legs specifically a lot of humanoid like robots just have wheels which is fine in a factory but it's still generally taking a sort of bimanual dexterous form um so there's this long tail of things where it makes sense to have a general purpose robot even though it's categorically going to be less performant than if someone went and built a purpose-built robot but you can build them at scale you can build a lot of them you can make them easily reprogrammable And that's going to ultimately be where we see humanoid security.

[1:07:18](https://realalpha.app/v/veG9nEN_kTQ/?t=4038) So let me ask you to make a quick prediction before you go. Tesla's humanoid business, one year from now, where do you think it is? One year from now. That's pretty soon. I think humanoids are... Because they've got to, I mean, you know, they set very aggressive goals and then they have a track record of falling behind. You can produce a lot of humanized, right? Production is not the challenge. The challenge is getting those first robots out there into production and learning from the data. So without a doubt, like we're going to start seeing deployments, right? But you can't scale up. Like if you have a humanoid or any kind of robot and it's achieving tasks at like a 95% success rate, that's like an amazing result for a lab. And then you put that into production, it's a complete failure. Because if you put like 100 of those in the line, you're, you know, every five minutes, you got to go fix a robot because like one of them has failed, right?

[1:08:05](https://realalpha.app/v/veG9nEN_kTQ/?t=4085) So you need to get an average website, you know, most serious websites are at least 99.9% of uptime they're looking for, or four nines of reliability for a lot of, you know, four or five nines of reliability for a lot of cloud platforms or big websites. So these numbers where we're looking at like one nine of task reliability or task completion success, it's not there yet, right? We need to take these good learnings we've got from the lab, deploy them into production, and learn from it. Watch what they're doing, get the data, bring it back, find the interesting edge cases, and use that to feed the learning.

[1:08:36](https://realalpha.app/v/veG9nEN_kTQ/?t=4116) Great. Well, Adrian, I want to thank you for coming on. That is Adrian McNeil from Foxglove here on TIT. Now that SpaceX has gone public, I've been interested to see the conversation pivot back to the technology rather than. the valuations of space companies. I've also been curious to see which other companies in the orbital data center race start to make headway. Baiju Bhatt, the co-founder of Robinhood, is working on one of those companies. He is the founder and CEO of Cowboy Space Corporation, and he joins me in our San Francisco studio. Baiju, welcome to the show. It's great to have you back. So last we talked to you, you were just getting around and telling us all the different places you were hiring people around the U.S. And you said you were hiring people in D.C., you've got people in Seattle, you're here in San Carlos.

[1:09:31](https://realalpha.app/v/veG9nEN_kTQ/?t=4171) Just orient me as to what is happening where right now for the company. Yeah, I mean, this is one of the fun and exciting parts about building a hardware company is you have to actually make stuff. And so going to the places where the people are that make it is kind of the path that we're going down. So we have an office in the Bay Area. It's in San Carlos, so 45 minutes down the road, which is where we're going to be doing the compute and the sort of data center and computer engineering part of the business. We've got a facility that's coming live in Seattle, which we're going to do the rocket engines as well as the satellite engineering.

[1:10:08](https://realalpha.app/v/veG9nEN_kTQ/?t=4208) And then two more, which we're working on, are going to be the actual vehicle production facilities, so the places where you plug the engines into the big cylindrical tubes that are the body. We're going to talk through the anatomy that we're on here in a second. And a launch pad. So you're making the stuff in Seattle. The engines, correct. And then you've got policy folks in D.C. presumably to help you deal with. Not anymore. We had some folks there, but we may have some folks in D.C. again at some point. Right. So, I mean, let's talk then about the anatomy of the rocket because, I mean, you've got an interesting approach here.

[1:10:43](https://realalpha.app/v/veG9nEN_kTQ/?t=4243) The way I understand it is, I mean, you basically, instead of losing the second stage and the upper, whatever, the payload stage, you're keeping it all in orbit. That's going to be the data center. Is that right? Correct. Yeah. How are you going to do that? Well, we're going to leave it up there. That's not been done before, dude. I mean, that's kind of the hardest part, right? Well, there are actually, I think there are some examples from the 60s of using the second stage as the satellite in orbit. I mean, a lot of things have been tried in the 1960s. Yeah. But the general approach is, so you have two-stage rocket, second stage gets you to kind of your orbit, and then the traditional approach is that you dispense your payload and then you re-enter the upper stage and it burns up on re-entry and parts of it will sort of end up in specific spots where they're landing.

[1:11:38](https://realalpha.app/v/veG9nEN_kTQ/?t=4298) In the case of data centers in space, the actual use case for the satellite is different. And in particular, a big chunk of what data center satellites in space do is they dissipate heat out that the chipsets are generating from the power that they're generating from the sun.

[1:11:55](https://realalpha.app/v/veG9nEN_kTQ/?t=4315) **Baiju Bhatt:** So our approach is actually to use the mass and the volume of the upper stage of the rocket as a functional part of the data center itself.

[1:12:03](https://realalpha.app/v/veG9nEN_kTQ/?t=4323) So using the actual vehicle itself as a part of the cooling mechanism is the novel thing that we're doing. And it's kind of interesting because it takes the really difficult part of re-entering the upper stage of a rocket for reuse in the case of things like Starship or the Space Shuttle program. And it says that there's a different answer, right? So you can either go with a disposable upper stage or attempt at a reusable upper stage. Or the third one is there's a usable upper stage, which is what we're kind of pioneering. Why hasn't this been tried before? I mean, you know, I know you've prided yourself on doing a lot of user research, qualitative research.

[1:12:45](https://realalpha.app/v/veG9nEN_kTQ/?t=4365) This is something you and I talked about the first time that you came on the show. I mean, I'm sure you've talked to many entrepreneurs who have tried space innovation. I mean, surely somebody must have tried this before, right? I mean, why hasn't this become a thing? It's the use case that data centers in space in particular generate because even if you contrast it with something like a low-Earth orbit communications constellation where, you know, you have a certain amount of mass and volume you can take to orbit, and then if you want to provide a service on the ground, you kind of want to take that mass and volume and spread it out over as many spacecraft as you can because they're moving so fast.

[1:13:23](https://realalpha.app/v/veG9nEN_kTQ/?t=4403) The thing about data centers and space is the actual application for the technology is pretty unique in that you ideally want to have your chipsets, meaning neighboring GPUs as physically close together as possible, which leads you to a monolithic architecture. And also the power requirements are gargantuan. This is the reason why this whole approach is being sort of explored, which kind of leads you to saying, hey, let's maximize the power that each entire rocket launch can generate. Right. So the point I'm trying to make is that while it is a novel approach to doing the data centers in space as one big upper stage as the data center, the reason is actually the way the technology works and what the technology is trying to do, if that makes sense.

[1:14:12](https://realalpha.app/v/veG9nEN_kTQ/?t=4452) You're still targeting the end of 2028 to have your first rocket launch? Yes, we are. And that's still the timeline? We're still working towards that, yeah. Okay. And, I mean, there's a couple other logistical questions, and we'll get to sort of your perspective on space, generally speaking. How long will these orbital data centers end up being in space for them? And what's the lifespan? Lifespan, yeah. So right now we're targeting about a six-year useful life, and we'll iterate on that as time goes on. If we can get more useful life out of the vehicle beyond the six years, then obviously that's beneficial.

[1:14:51](https://realalpha.app/v/veG9nEN_kTQ/?t=4491) But it's kind of baselined against the usefulness of the chipsets and how long we think each one of the components is going to survive. And how many will you want? What's the ambition in terms of how many you have up there? Yeah, our application with the FCC is for 20,000 of these things. So each one, we're baselining at about a megawatt of power. So the target is about a 20 gigawatt constellation in orbit. And if you just kind of think about our company doing this in the context of where the space industry is today, and this is kind of the really exciting part about all the data center and space stuff, is that this represents a really dramatically larger commercial use case for outer space.

[1:15:33](https://realalpha.app/v/veG9nEN_kTQ/?t=4533) Because up until now, it's been primarily government applications, as well as Earth imaging and communications, data centers in space. The mass and volume that it's going to demand is very substantial compared to those applications. Right. What do you think of Blue Origin's approach? They just raised, or they're in the process of raising outside capital for the first time. They've got this interesting approach here where they've got TerraWave and Sunrise, which is sort of this two-sided system for their own orbital data centers. Pros and cons of that approach. I mean, what do you think of that?

[1:16:09](https://realalpha.app/v/veG9nEN_kTQ/?t=4569) You know, I know a little bit about the TerraWave approach. I haven't followed their data center in space approach that closely, so, you know, I'd have to... Tell me, have you ever met Jeff Bezos? I have not. I have not. You've not met either of them? No, I haven't. What would be one question you would ask them as a fellow space entrepreneur? Yeah, I mean, I think I try to approach a lot of the situations where I meet other successful people. I try to approach it as a student, right? I think both these folks are folks that I have tremendous admiration for. Do you think you can do it better than them?

[1:16:48](https://realalpha.app/v/veG9nEN_kTQ/?t=4608) I mean, obviously I do. Otherwise, I wouldn't be doing it. What makes you think that? I mean, that's just kind of the entrepreneur in me, right? I mean, like I said, obviously, I have a ton of respect for what. I think I would be honored to spend some time with them. Haven't had a reason to yet, but we'll see if the day comes up. I want to ask you a little bit about the trials and tribulations of those companies, SpaceX, Blue Origin. Space is hard. There's no denying it. What have you learned from their process that you're trying to apply to how you build your rocket? Yeah, I think one of the most important things is kind of the role of iterating through failures, right? SpaceX kind of dramatically illustrated that iterating through failures is the path to reliability, is the path to reusability.

[1:17:41](https://realalpha.app/v/veG9nEN_kTQ/?t=4661) And I think it's like when even I think when we were watching that happening a decade ago when they first got to reusability, it was kind of like, oh, my gosh, is this launch going to work? There was a level of sort of tension, which I imagine must have been even more pronounced internally. But it's kind of having the risk calculation of saying in order to get these parts to work in a reliable way in the real world, the only real way to do it is to test it in the real world. That there's a limit to where you can get to with simulation, iterating with computers and models. And at some level, you've got to try these things in the real world.

[1:18:18](https://realalpha.app/v/veG9nEN_kTQ/?t=4698) And the reason I think this is important is because it also has an effect on how you think about the cost of these things, right? That if you're going to try to build something that's truly reliable, it kind of goes hand-in-hand with trying to down-cost as much as possible. Because if you try to iterate through failure and each one of your, you know, items that you're building is very, very expensive, it limits your ability to do so. But go deeper than that for a minute. I mean, I hear you on the iterative approach, but, I mean, what have you learned from, I don't know, the engineering approach that they've taken, the teams that they've built, you know, the cadence on which they've approached it.

[1:18:58](https://realalpha.app/v/veG9nEN_kTQ/?t=4738) I mean, Blue Origin has been in business for 25 years now. Yeah, it has. I mean, how do you think about how you are building your own company learning from the process that they've gone through yeah i'll give you one really uh concrete real world example it's kind of the role of of hiring people that are fresh out of college the new grad talent and what what to look for there so i think one of the things that we've built at our company is hiring a lot of people that have done competitive racing programs in college wow there's something called formula sae um which is a college racing program where my understanding is that teams of students build race cars and they go and race them

[1:19:37](https://realalpha.app/v/veG9nEN_kTQ/?t=4777) **Baiju Bhatt:** and getting folks that have done that kind of stuff in college even if they've had a little bit less real world experience like that experience of going from idea to building the real thing to

[1:19:49](https://realalpha.app/v/veG9nEN_kTQ/?t=4789) testing the real thing to living with its failures that that is actually like a really interesting talent pool to build the company around so we've got a lot of those folks at cowboy and if anybody's listening, we are continuing to hire. Do you think that orbital data centers start to become a matter of national security? I do actually, yes. I think for a variety of reasons. One, I think the importance of artificial intelligence from a national security perspective is undeniable and I think when you look at that in the context of the space industry, space in general is an area of importance from a national security perspective So it's kind of the intersection of two.

[1:20:31](https://realalpha.app/v/veG9nEN_kTQ/?t=4831) So I also think that as time goes on, you're going to see other countries try to build substantial constellations to do this job in low Earth orbit. And I think it's, you know, preserving the U.S.'s lead in this is a matter of incredible national importance. To what extent do you think that lead could be under threat? I mean, if you look at the innovations happening outside of the U.S. I don't follow the space companies in Asia as closely, but are there companies that are close on our tails at all? I mean, I think the Chinese are very formidable, right? I think that the Chinese government's involvement with it, again, I don't follow their program that closely, but, you know, the Chinese national ability to build stuff is unmatched.

[1:21:25](https://realalpha.app/v/veG9nEN_kTQ/?t=4885) So I think regardless of where they are today, I think it's more of a snapshot in time and it's more like how quickly do we expect to see them evolving right so yeah I mean I think they're they are going to be the nation that is the most likely competitor to the United States from a volume and mass orbit perspective right uh SpaceX what public as you know did you buy SpaceX stock I did not what what did you make of that whole IPO process I mean, you know, you built Robin Hood, you've studied new issuances at large. I mean, what was your reaction to just the sheer size of that offering and how everyone reacted to it?

[1:22:07](https://realalpha.app/v/veG9nEN_kTQ/?t=4927) Yeah, I mean, I think the sheer size of it was pretty jaw-dropping. It's kind of interesting because the week before they went public, I was actually in New York having some meetings with various banks. and it was kind of interesting because i think the last time one of the times before that that i was in some of these banks was when we were doing the same thing um yeah i i think the the the sheer size of the thing was was the thing that was like kind of jaw-dropping because um when you when you just think about the amount of capital that's going into that and the impact of that on the space industry as a whole um it has a variety of effects right number one is it really puts SpaceX as a company with the ability to build some pretty big stuff going forward. I think the other part that's interesting is it also is educating a new generation of investors, institutional investors, retail investors across the board on the role of the space industry and kind of how that technology works. And that kind of goes from the public markets to the private markets as well, right? So I think the overall effect on the space industry is really, really substantial from it.

[1:23:19](https://realalpha.app/v/veG9nEN_kTQ/?t=4999) Right. I was speaking to another guest on the show who is well-versed in the space sector, and they had a question for you that I will ask you. It was a technical question. Okay. They asked, so how are you going to deal with the issue of getting the actual compute downturn, the ground link issue? So the way I understand it, I mean, cross-link is the technology that a lot of entrepreneurs have tried at. Right now, radio frequency is really what Blue Origin and SpaceX have had to rely on. How are you going to approach that issue? Yeah, our approach is using optical for data transmission, both space-to-space and space-to-Earth.

[1:24:02](https://realalpha.app/v/veG9nEN_kTQ/?t=5042) Even though it's not been done before? There have been examples of space-to-Earth data transmission done. And on that part of the overall technology stack, our first satellite goes to orbit later on this year. While it's not doing space-to-Earth data transmission, it is doing space-to-Earth power transmission. And so in service of that, we've actually built up quite a bit of technology around the optical stack, both the lasers and the pointing mechanisms. And for our second mission that goes to space, which will be first half of next year, we'll actually be reusing a lot of that core optical technology to demonstrate our ability to transmit data from space to Earth.

[1:24:46](https://realalpha.app/v/veG9nEN_kTQ/?t=5086) So this is a solvable problem to you? I believe it is, yeah. This is not a technical challenge that you're still grappling with, in other words. I mean, I think it is going to continue to be a big technical challenge that we have to develop because we want to take... the bandwidth that we're transmitting data and really push the boundaries on that. Right. Let me ask you one last question, not necessarily about space, but the AI race right now. How does that impact what you are building for space? Is it, you know, does it matter to you whether or not the models get better, whether Fable can do X, Y, Z?

[1:25:25](https://realalpha.app/v/veG9nEN_kTQ/?t=5125) I mean, does that have a direct impact at all on what you're building in outer space at all? I think it does, yeah. I think it does. How? Because you see the amount of energy that the models are consuming these days. And yeah, I take your point on the energy more. I meant with respect to your ability to build what you need to build. Oh, like our ability to leverage the models. I'm not talking about the market size. I'm talking about your ability to make rockets. I mean, is that at all, is that impacted by what Fable can do and not do? Yeah, absolutely it is. I mean, I think you just take a look at, I mean, there's a refrain in society that something is or isn't rocket engineering, right?

[1:26:04](https://realalpha.app/v/veG9nEN_kTQ/?t=5164) Like, this is viewed as very challenging to do, and it's knowledge work, right? The ability to use these models to help accelerate that is undeniable, and we see it across our company. And I see it for myself, too, right? I've gone from being a finance bro. Self-proclaimed. Self-proclaimed. I don't think anybody would dispute that. to being able to build a company like this. And the amount that I've been able to learn through my own use of AI from asking questions about how different parts of the rocket work and getting really high-quality results, obviously, that I have to go and verify, it's been a dramatic accelerant, right?

[1:26:47](https://realalpha.app/v/veG9nEN_kTQ/?t=5207) And the ability for it to solve more and more nuanced problems, its ability to do, you know, computations in the background to help size stuff, it's quickly becoming like an indispensable tool across the board. It's kind of amazing, honestly, especially for the math and physics stuff. Last question for you. Your dad worked at NASA, I recently learned. Are you going to hire him? What's the story? Yeah. So when I was a kid, my dad had a job at NASA Langley on the East Coast as a contract research scientist, coincidentally doing atmospheric sciences. So it's like you, the apple really doesn't fall that far from the tree.

[1:27:29](https://realalpha.app/v/veG9nEN_kTQ/?t=5249) I think he's probably not going to be working with us too directly on this, although has quite a few opinions to share with me. It's like, why don't you do it this way? Why don't you do it that way? And he's usually right, annoyingly. Hired. Yeah. Working with your dad, I don't know if it's like, is the adventure I'm looking to sign up for at this particular stage in life, but he would very much appreciate you saying that. Great. Baiju, I want to thank you for coming on. That is Baiju Bhatt, the founder and CEO of Cowboy Space Corporation here on TI-TV. Okay. To close off the show, I want to end with a special edition of The Editor's Cut.

[1:28:20](https://realalpha.app/v/veG9nEN_kTQ/?t=5300) We're going to be talking today about AI policy and the debates around that topic. I'm joined here in San Francisco by our co-executive editor, Amir Afradi, and our San Francisco Bureau Chief, Jason Dean. Welcome to you both. It's great to have you here. Hello, hello. Great to be here. Okay, so predictions for AI policy. I mean, we're looking ahead. We've done 250 shows. We're going to do 250 more, 250 after that. Amir, I'm going to start with you. Where is AI policy going? It's kind of a meaty topic right now. Yeah, finally heating up. It really seems like for the first year, year and a half of the Trump administration, there was.

[1:29:00](https://realalpha.app/v/veG9nEN_kTQ/?t=5340) There wasn't too much going on. A lot of the efforts by the Biden administration were kind of pushed aside. We just had a long discussion about it last week. You know, the organization that tests models was really shunted. Now, all of a sudden, in a matter of a few weeks, it's all back. And, you know, there was certainly a lot of discussion around chip exports before, but really nothing on the models themselves, and how, you know, we would be dealing with potential risks on cyber, bioweaponry, other things like that. that, you know, folks in the Biden administration were really focused on, all of a sudden, out of the blue, you know, all these concerns have come up. This is just going to crescendo from here because these models aren't going to get any worse. So you just imagine that, you know, every month or two or three, there'll be some new leap forward in capability. We now have, you know, the beginnings of a kind of process in place for different arms of the government to test these things but we're not going back like this is you know clearly from a just a technological you know point of view there is a lot of reason to review these models uh for for different times types of risks and now politically there's an incentive heavy incentive for much more hands-on

[1:30:13](https://realalpha.app/v/veG9nEN_kTQ/?t=5413) regulation jason yes you agree is there another version of the future uh there could be but i tend to agree with the mirror on that i think but okay hear me out here i mean you know if the staggered release is going to be the norm i mean silicon valley has at least some influence over government they try right i mean i'm thinking out loud here if there's enough uproar enough lobbying i mean couldn't they have some impact on you know uh making it uh an even friendlier environment for these model releases? What do you think? I mean, I think they have had impact. They've had impacts in convincing the administration early on to take a more laissez-faire approach. I think they've had a hand in and influence over the nature of the regulation we have seen. It's not that they necessarily universally want it to play out exactly that way, but obviously Anthropic has talked about the need for government regulation. Others in the AI industry have talked about the need for the government to play some role in ensuring that these are safe and that the challenges we've seen with things like mythos and cybersecurity are not just Wild West.

[1:31:25](https://realalpha.app/v/veG9nEN_kTQ/?t=5485) Okay, what about open source? I mean, this is another battleground here in AI. We've got open source models coming out of Asia that are, I think from what I understand, better than the rest right now. What do you think is coming for regulation of open source models coming from that side of the world well this is the topic that's preoccupied our newsroom for the past week or so um and i think a lot of people got riled up as we and others were asking questions about this because it was one of the unresolved questions coming out of the most recent executive order on ai around evaluating models before release um something surely will happen uh around kind of contractor use of Chinese models just optically doesn't look great for that to occur. So you could see a situation where, you know, if you were doing direct business with the U.S. government, U.S. government will require you to use non-Chinese models over, you know, perceived security and other risks. So that seems like a starting point. And then the question is, okay, well, if the proprietary or closed-source models are being looked at a lot more closely for various risks and they need to be tested um what does the regime look like for testing completely open source

[1:32:41](https://realalpha.app/v/veG9nEN_kTQ/?t=5561) models and like if the chinese ones are just available and their capabilities are getting quite good certainly not necessarily at the level of anthropic and open ai but getting quite good um what are the guardrails that need to be placed on on them since they're open source that's just available. So I think it, you know, raises a whole ho... Most of very difficult questions, just given that these things exist, they're out there, and anyone can use them. And if we don't have like a U.S. business community that's completely hardened and steeled against, you know, cybersecurity threats, as an example, you know, people could just start using these open models for those kinds of intrusions. So I think it's going to require a lot of creativity, but there will be, this is going to kind of dominate the conversation, or at least be a big part of the conversation for the next few months.

[1:33:29](https://realalpha.app/v/veG9nEN_kTQ/?t=5609) Right. Jason, let me ask you a slightly different question. So we've been talking on the show today about trust or lack thereof in the large AI labs. We saw Satya Nadella came out with a blog post, and he was talking all about the idea that it's better to have more control over your AI, essentially, in this climate. We saw Mark Benioff with the zero data retention tweet, I guess, really touting Salesforce's own technology. I mean, trust in AI Labs, it seems like it's at an all-time low in some ways. Is this a blip? Is this a trend? What do you make of that? I'd say it's a trend. So it's not coming back?

[1:34:11](https://realalpha.app/v/veG9nEN_kTQ/?t=5651) Well, I don't think it's binary, but I do think that part of what's happening is that these companies are growing so fast, Anthropic in particular, and there's a concern within the tech industry about the merits of the issues around data retention, but also around the consolidation of power in these two relatively young companies that are threatening the entire SaaS industry. And obviously, we've seen stock valuations there take a big hammering over the last year or two, and just in general. So I think that there are genuine concerns, and then there's the fact that you have a lot of these companies seeing them now as, including companies that were close partners with them and even still are, in fact, seeing them as much more intense competitors.

[1:34:58](https://realalpha.app/v/veG9nEN_kTQ/?t=5698) And that isn't going to go away unless, for some reason, the AI itself kind of fizzles out and those companies are cheating. As long as they're continuing to grow their business anywhere near the pace they're at now, I think that the other companies in the industry are going to be incentivized to talk up problems. And I don't think that's going to help trust. Yeah, but it remains to be seen whether anyone's going to listen. If you have a fear-based message to your customers, if you're Satya Nadella or Alex Karp or Mark Benioff, you probably need to couple that with more of a value proposition than just saying, like, these guys are scary.

[1:35:36](https://realalpha.app/v/veG9nEN_kTQ/?t=5736) Don't do business with them because so far people are continuing to do business with them. And, yes, everyone's looking for alternatives. Everyone's looking to save on costs. But, yeah, I'm not sure if the fear-based approach is going to work. Well, and it's sort of making me think about the extent to which an IPO could have any impact at all over this, because I'm thinking back to when Circle went public, for example. I mean, part of Circle's value prop was, hey, we know crypto has some trust issues, right? Lack of transparency. This is something we want to fight against. We want to go public. We want you to see our books. We think it's going to help with that.

[1:36:13](https://realalpha.app/v/veG9nEN_kTQ/?t=5773) But, I mean, you can open your books if you want to the public, but still, the technology seems to be as opaque as ever. And so, I don't know, going public in some sense doesn't really help this trust issue at all. Again, it just depends on whose trust you mean. If you mean the partners, the cloud providers, the SaaS companies that are reliant on these models, that's one thing. I think for, you know, customers of a coding assistant that is powered by Anthropic or Anthropic's own coding assistant, I don't know, they seem okay at this point. Like, again, there's going to be limits just in terms of costs and people trying to figure out, okay, well, is there a more efficient way to do this?

[1:36:56](https://realalpha.app/v/veG9nEN_kTQ/?t=5816) But as long as they have the best product, the best model... You know, going to be tough to stop that freight train. Great. Jason, last question for you. Next 250 shows, what do you think is the defining trend that we should be covering over the next 250 shows? I'm going to be coming to you with a question, so you better pick something that I think you have expertise about. What is the rise of AI and how it's affecting the rest of the tech and business world? And all the facets of that that you will undoubtedly be digging into. in great depth. Well, what seems to happen is, you know, the people who are living in the future at these frontier labs, they, you know, it turns out that the stuff that they talk about today is the kind of thing that's actually going to arrive in two or three years. So this may not be the best example, but the first conversation I ever had with somebody at OpenAI was around recursive self-improvement. That was in mid-2023. It is now finally at the stage where people are actually discussing it as a serious thing and it's taken three years so um i do think that that so far you know you know i i don't know necessarily about the labor concerns that that dario and

[1:38:07](https://realalpha.app/v/veG9nEN_kTQ/?t=5887) others have articulated but generally speaking a lot of these leaders have been right over the long run so you still talk to that person uh yes i do so 250 shows on recursive self and that is why it's worth subscribing to the information is because we have long running sources that will take you inside and outside of the companies i want to thank you both for joining us that is jason dino san francisco bureau chief amira fradi our co-executive editor that does it for today's special one year anniversary of the show i want to thank you for joining us i want to thank our guests to mark the occasion we have some champagne on hand i am told so with year one in the books i would like to thank the entire newsroom for joining us on the show i want to thank our producers sydney and zach for making the show a reality i want to thank you for tuning in here's to year one here's to year two here's to every year thereafter Cheers.

[1:39:07](https://realalpha.app/v/veG9nEN_kTQ/?t=5947) Cheers.