We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
The enterprise AI reading lead: own the intelligence your workflows create
1 day ago
•
1 min read
•
68 docs
A single high-signal reading lead: Eno Reyes endorses a Satya Nadella piece on keeping the intelligence generated by enterprise workflows under the company’s control.
Standout recommendation
Satya Nadella’s piece on “owning your intelligence” — written piece
- Creator / link: Satya Nadella of Microsoft. The direct article link is not included in the surfaced recommendation; the available access point is the 20VC conversation in which it was discussed. Reyes refers to it only as “a great piece.”
- Recommended by: Eno Reyes, Factory’s CTO and co-founder, who says Nadella’s argument is “very spot-on.”
- Key takeaway: Reyes brings the piece into a discussion of where AI’s durable learning accumulates. He argues that continual learning happens at the harness layer—the place where application logic and state are maintained—and that businesses should own those learnings rather than surrender them to another company.
- Why it matters: The useful reading lens is not simply model quality, but control of the data outcomes, successful workflows, and business knowledge produced through AI use. Reyes warns that if those assets belong to a model provider, it may eventually know how to run the business well enough to compete with it or gain leverage over it.
This is a lightly specified recommendation rather than a direct “read this” instruction, but it is a specific, positive endorsement made in a substantive conversation about enterprise AI sovereignty.
The enterprise AI reading lead: own the intelligence your workflows create
Research extraction
20VC with Harry Stebbings
Direct answer: The transcript contains no fully specified book, article, paper, podcast, or video recommendation. The one authentic content-resource endorsement is Eno Reyes’s positive reference to a Satya Nadella/Microsoft “great piece”; its title and article URL are not provided.
Satya Nadella/Microsoft article or piece — soft but authentic endorsement
- Title: Not stated; Eno calls it “a great piece.” Creator: Satya Nadella (the transcript renders the name as “Satia”), associated with Microsoft. Link: No article link is present.
- Recommender: Eno Reyes, introduced as Factory’s CTO and co-founder.
- Exact takeaway: Eno says, “I think that all of that opining is very spot-on.” The referent is the idea that enterprises should own their intelligence—especially the data, learnings, and workflows that produce their business outcomes.
- Context distinguishing endorsement from mere mention: Eno is discussing the risk that model providers will pursue the industries they supply and gain leverage over businesses that outsource their intelligence. His praise of Nadella’s piece as “great” and “very spot-on” is a clear positive endorsement, though not an explicit “read this” instruction.
Arena.ai — borderline implicit tool recommendation, not a content recommendation
- Resource/link:
arena.ai; no creator is identified. Recommender: the host/interviewer, speaking in the first person. Takeaway: He says he uses Arena as a discovery mechanism for new models and is “loving the output.” - Classification: This is favorable firsthand-use testimony, but not an explicit recommendation to read, watch, or listen to anything. The surrounding exchange frames it as a way to discover new model releases and as part of the information/news stream around model routers.
- Resource/link:
Explicit non-media recommendation for completeness: Eno recommends that American companies analyze current Chinese frontier models for the tasks they care about, but says U.S. national-security users should not use them and that he “highly recommend[s] against using a Chinese model” for American defense or preparation. This is model-use advice, not a book/article/paper/podcast/video recommendation.
20VC with Harry Stebbings
- Article by Satya Nadella (title and URL not provided): Factory co-founder Eno Reyes called Nadella’s piece on businesses owning their intelligence “great” and said its argument was “very spot-on.” He connected it to the risk that model providers could learn a company’s workflows, compete with its business, or gain leverage over it when intelligence is outsourced.
Should American Enterprises Work With Open-Source Chinese Models? | Only 10% of Neo-labs survive
1 of 1
highlight
0:00I see a world where the smartest model is actually the cheapest. People are thinking about outcomes in AI and they're looking at 20 30 50 and they're saying that's ludicrous. That's crazy.
0:09That is underestimating by an order of magnitude how massive a transformation this is going to be. 8 and 10 billion is the new 1 billion. Today we have CTO and co-founder of Factory Eno Reyes. He is one of the most articulate and insightful thinkers about the value stack of AI that I've interviewed.
0:26Factory is one of the leading companies that specialize in autonomous software development and ENO is incredible in this show today. There's going to be a lot of notes taken in this discussion.
0:35Two of the largest companies that provide models today have explicitly said we are going to go after every single one of these industries and businesses that we provide intelligence for. Calling open- source models Chinese models is a scop by the Frontier Labs to basically trick people into thinking that they're scary and otherize them. In 3 years, 99% of workflows are going to be done on open models.
0:59Ready to go.
1:11Eno dude, it is so good to have you in the studio. I obviously had Matan in the studio. I was chatting to Keith Boy over the weekend about you. So, thank you so much for joining me, dude.
1:19No, thank you for having me. I'm super pumped to be here. Now, I hate background stories. I'm sure you listen to podcasts. It's like, how did you get into this? And you were like, I'm bored of this. But downstairs, I asked you, how did you get into technology and fall in love with it? And your story was like heart-wrenching and compelling. And so, I have to ask it, how did you first fall in love with computers, do you think?
1:41Yeah, I I think that the beginning was my parents. So both my parents went to art school and they were always in love with technology and how that sort of intersected with creativity and media and I think that where that started especially with my dad was he was born in San Francisco in the late60s and he was 6 years old and he got actually hit by a bus uh and that changed the
2:09trajectory of his life in a pretty crazy way where you know he wasn't playing sports, he wasn't able to do as much of that like traditional 1960s kids stuff.
2:19Instead, he went to technology and computers like the early days of uh when you know you barely had screens and were instead sort of tinkering. And he spent most of his life basically embracing technology as a way to extend his own reach beyond like what I'd argue his physical body could do. It's so interesting how life delivers you a hand, so to speak, and how consequential that hand is to kind of how you are today.
2:47100%.
2:48Um, it's very hard to transition from a father being hit by a bus to margins, right? Like only a venture capitalist could do that in such a swift transition.
2:58Well, some margins in AI might make you feel like you've been hit by a bus. So, I mean, yes, absolutely. and we we chatted before you said the cheapest model isn't necessarily the cheapest system and I read this when I was doing the work over the weekend and I was like huh can we just unpack that the cheapest model isn't necessarily the cheapest system what does that mean yeah I you know it really comes down to this idea that when you're thinking about price you should
3:27not be thinking about the inputs to the price you should be thinking about the outputs so I think about the price of the outcome so let's take software as an example How much does a code review cost is far more interesting than how much do the tokens inside of that code review?
3:42And so uh you take a very sophisticated model if it's able to do that code review uh immediately without making any mistakes, getting the right outcome right away, searching the right phrases uh and you use you know a thousand tokens or whatever it will be significantly more than that a million versus you use a cheap model and it spends time. running it uses 50 million tokens and ultimately you know that price difference of the full outcome uh
4:10makes the higher quality model cheaper. Uh now that's not how all tasks go but for many of the most uh intelligent demanding tasks. I see a world where the smartest model is actually the cheapest.
4:24I totally hear you there. And it kind of goes against that a token is a token theory, but will we then have millions of specialized models with every company having specialized models operating on their own data because to your point there, it'll be able to work much more efficiently. Yeah, I think that there's a real world where the speciation of models increases very rapidly. This is sort of the world where the fireworks and the people who help make models
4:52possible. I think win because the alternative is you only have a very few specialized providers that have models.
5:00Now in our view there is going to be probably a difference between the commodity task executors. So this is just like your everything model. Uh and that we think will be dominated by open models. And then you have businesses that will say well you know we do a lot of commodity tasks but there's a couple of very high volume specialized tasks that only we do. And for those your commodity model won't be good enough.
5:25Your frontier model will be too expensive. And so they'll want something in between where they can take a commodity model and make it good enough via post training. And then ultimately they'll run that and they probably will be the only consumer of it. So they won't even give it to the rest of the world. They'll just keep it entirely internal. Uh so I think that that will lead to a lot of models not like a you know it won't be millions but it'll definitely be quite a lot. When you look at the post-training required when you look at the implementation required I
5:54look at that and I think that company structures and teams today are simply not equipped to do that.
6:00How will we solve for that? Is this just a moving into an incredibly services and implementationheavy world where we have these insane AI teams coming into every how do we solve that?
6:09Yeah, and I think that this is one of those things where right now the recipe for post-training and building models does live primarily in the heads of a specialized few group of people. Um, but that's also how software development was like 20 years ago. So in my mind, the same tools that are currently democratizing access to software are actually the tools that will be used to democratize access to intelligence in general. So I see a world where uh you're a company, an enterprise, and you say today, well, we don't have the the knowledge or the skill set to build our
6:39own specialized models. Well, very shortly already to a certain extent, you can uh open up a platform, go into their, you know, web page, click a couple buttons, describe the task you care about, point it towards those workflows that happen in your business today, and outcomes a model. And that model will be really good. Uh, and so I think that right now the there's like a couple of companies that claim that like recursive self-improvement and model training will be only their domain. And I think in reality many businesses will
7:08have access to that technology via software services that other companies sell.
7:12You said about kind of focusing on the outcomes and not the inputs.
7:16That is a great idea when it's a verifi verifiable output. code review when there is ambiguity to something which could be uh marketing conclusion did it come through X channel or Y channel uh my girlfriend's a lawyer different legal notes ambiguous some people like it one way some people like how do you think about the importance of verifiability in determining outcome quality yeah I mean verifiability is ultimately
7:44the single most important property of success with current AI systems and I think that The way that we'll sort of progressively address this is by building new ways to verify the work that we do. And so, you know, I'll be really concrete about that. Uh, what a bunch of the eval creators and model trainers have done in some of these domains like healthcare and legal where you don't really have a concrete set of verification strategies. They'll take experts, they'll bring them in, they'll
8:13have them create effectively, you know, their own new forms of verification.
8:18Maybe they show two examples side by side and they say which one based on your judgment is better. And being able to then take intelligent models that have a lot of the ground reasoning and the knowledge that these domains have, combine them and say now you as the model go and build this similar form of verification. I actually would argue the frontier right now of AI is AI systems that can build verification where there is none and thus they can progress into
8:46tasks that today humans consider to be too difficult for AI to resolve.
8:50AI systems that progress into verification what what does that actually mean?
8:55Yeah. to make it as concrete as possible. You know, imagine that you are walked into a room at a law firm and they say to you, "Hey, you're going to start, you know, basically judging how how to determine whether or not these new hires are good."
9:13What does like a very novice manager do?
9:16Well, they kind of go by their gut.
9:17They're looking and what they're doing is if they're if they are actually intuitive, they'll go by their gut and make good calls. they'll say that uh you know new grad is going to be big at this firm. I like them. I'm going to continue to promote them. But when you have a giant firm or you sort of systemize, you realize that that approach to management is very rare. In reality, you need to go to the firm and you need to start writing stuff down. Here's the things that we like about great, you know, new hires. Here's the things we don't like.
9:44You know, any company starts to learn managing at scale looks like this. you have to write down the things that you care about and build a framework or a system to analyze how it goes well uh you know the job to be done. I think that great AI systems are basically relearning this like management strategy of saying you can go by your gut and honestly a great AI system can be right very often without this sort of structure framework but in reality what you'll need to do is you'll need to write down this is what good looks like
10:14this is what bad looks like and here's how we judge and what's interesting is yes that makes the system better at determining what good looks like but it also changes the incentives because when you write down this is what good looks like. People read that and then they start to act more like what good looks like. And so you have to be careful because if you say good looks like A, B, and C, you're going to get a lot of A, B, and C. But if you basically built the wrong incentives, then that system will
10:43end up following that pattern regardless of if it's actually good or not.
10:46It's a brilliant statement. Show me the incentive and I'll show you the outcome. The hardest thing about actually venture firms when you run them as a business cuz if I set you, you know, the the kind of goal of three deals per year, you'll give me three deals per year. I I don't want three deals per year. I want a great deal. I want factory.
11:05I don't care if it's three or six or one. And so it fundamentally, to your point, changes the output that you get. Um, I think I made a big mistake because I'm in Mccor.
11:17Uh, and I think we did it at like two or three billion. My my memory should be better, but I'm older than you. Um, and I didn't do the latest round at, you know, whatever 20 billion because I thought, how much bigger can it be? Like maybe a hundred billion, but that's a 5x.
11:36It's like not that exciting. 5x and that's with no dilution. and I'm now thinking that I'm completely [ __ ] wrong and that there is a pathway to 20030 billion in the data requirements that will be needed.
11:50How do you think about what I just said?
11:52No, I think that's that's totally true.
11:54I mean, people are thinking about outcomes in AI and they're looking at 20, 30, 50 and they're saying that's ludicrous. That's crazy. That is underestimating by an order of magnitude how massive a transformation this is going to be. However, I think that what people underestimate is that the types of businesses that are going to become massive uh do not look like businesses 20, 30, 40 years ago where they had a technology mode or they had some sort of key capability that no one else could
12:23replicate. Instead, it's basically collections of people that understand what the future looks like a little bit more cleareyed than the other people.
12:30And so these data companies like you mentioned Merkore, yes they sell data, but every person at that company gets how AI is going to look much clearer than the average human and that makes them worth significantly more than you know even what investors will say. Going back to what we said about lots of specialized models and companies working on their own data which is obviously proprietary. I'm I'm confused. How does this not reduce the TAM for frontier models? I think it might. I think that
13:00the TAM of frontier models is frankly overweighted right now. Uh the world basically assumes that there's going to be one to three companies that have total domination over the intelligence era. And not only is that I think is just a silly proposition because generally people don't like that sort of strong dominance of a couple of few of small companies or large companies um but the real question is how do you
13:30defend your margins if you're a model lab when there are so many options and I think that margin profile shrinking is going to change the expected value of these businesses baked into 23 4 trillion valuations is an assumption that you can basically 2x the price of those tokens and people will buy them.
13:50So is the margin profile looking [ __ ] And what I mean by that is Anthropic just produced their first quarter, I believe, of profitability. Margins are seeming to be ripping and they're throwing off cash now. Is that not going counter to what you said?
14:04No, I think I think also that one of the biggest drivers of that is actually the applications on top of that. Uh and so one of the things that I think is quite clear to all of these model businesses is that the model itself may be a fairly rough trade-off or rather uh the margin profile of the models is definitely worse than the applications. Uh in my mind if you are a model provider you're basically looking A to dominate the platform era in which case you want to
14:33be one of you know end companies that sort of sell get really good at selling inference or B you just want to move up to become an application layer company that has really good models. Anthropic seems to be following the application path while OpenAI seems to be dipping its toes in both but the platform commitment from them seems much stronger. Which strategy do you think's right if you were to bet on one?
14:57I I think that the application layer is going to be a much harder battle because being model locked is actually a huge disadvantage if you're trying to sell outcomes.
15:07Why why is that?
15:08Well, it's bad incentive alignment. If you are a model locked provider, then you are inherently selling those tokens in order to make sure that your business, you know, gets the margin it needs. If you are going to a company and saying we can give you the, you know, the best outcome, you have to do that with only your models. So they can really give you you know their best model. Meanwhile, someone who's not model locked can give you the best model, right? That difference between their best model versus the best model
15:37can be massive in the pricing. I mean, we see this right now in real time in coding. Anthropic can basically only deliver, you know, their model's outcomes. And the BAS model also is highly subjective dependent on the consumer.
15:51Oh yeah, it's totally different based on the task, the profile, the risk takingaking that people want to have. Uh tons of different options.
15:57Can you help me? Again, I'm I'm very thick, but I I I don't like cynical questions. I like to be optimistic. I think it's fantastic. We're seeing anthropic potentially go out at 2 trillion, but you're essentially placing a $2 trillion price on claw code.
16:11Mhm.
16:12Which is not that difficult to switch off of.
16:15Yeah.
16:16What am I missing? Is that what should I know that I'm not getting? How should I think about that?
16:21No, I mean I think that is fundamentally the risk for an investor is that you're making a $2 trillion bet on one of the most competitive application markets in one of the most finicky segments of the market which is you know dev tools. And so I do think that part of what needs to happen in order to make these uh companies like Anthropic and OpenAI realize the value is they either a uh which they're pursuing have to go through regulatory capture in which case they go and they tell uh they sort of
16:50scare politicians into into thinking that they must own the means of intelligence and thus they become the only providers of the most frontier capabilities or B they have to figure out a way to build applications and outcomes that match the true PTO frontier of cost and quality. And I think that means opening up to more models. So it's actually like sort of at odds with the two strategies. You either, you know, capture it and keep
17:18the model or open up to everybody. This is a very hard decision and one that I think you can start to see OpenAI actually grappling with as they've let more models into their harness. They they're not making it official, but they're clearly supporting an open model ecosystem in a more direct way.
17:36Do you think Dario's marketing message has been mistaken?
17:39I think that the marketing of AI in general was probably one of the worst marketing jobs done by uh you know, contemporary capitalists in that it basically did the opposite of what you want. scare every single person, tell them it's very unreliable, and basically threaten their well-being and livelihood with the technology while you roll it out at scale. And so, I think that the challenge is that the things that Daario
18:08brings up are not only well-intentioned, but there are very real threats from unregulated and dangerous AI. I think that there's a way though to communicate about this without maybe embellishing the economic ends and because I think that the carrot and stick here can just be one you know the technology will be incredibly transformative two and humans will have a huge role in that transformation and you will have a huge
18:35role in that transformation uh and three if we don't do this you know well then like all other technologies there's going to be risks but the moment you start talking about the singularity and AGI and create this godlike mythology out of AI, you're going to scare a lot of people.
18:52We're going to be the last remaining private company. It's like truly, well, you flew here on United. I don't know how that's going to work.
19:02Yeah. I mean, you know, there's going to be a lot of uh of change if some people think that there's going to be one company. And I think Sam Alman just did an interview where Yeah. And kudos to him. It's hard to go back and say I was wrong. He basically says, I totally underestimated the momentum of the economy, the momentum of existing businesses, and so I predicted this future that actually has not come true.
19:25And I think that that's a great reckoning where you go and you say, I didn't think that this was going to happen, and it's clearly not going that way, so I'm revising my prediction.
19:34What did you not think was going to happen where you had to revise your prediction?
19:38Oh, I love that question. I think the biggest thing that surprised me where I've had to go back and seriously correct my priors uh is building a business is much more reactive than planning. Uh, and what I mean by that is, you know, almost every one of our best decisions as a company has been in reaction to some information and a split-second decision sort of rather than a master plan that we
20:08forecasted 6 months ahead. And I think that that knowing that you start to look at the rest of the world. You hear from other business leaders how that's also how they make decisions and you start to realize that uh the world is just this like constantly reactive feedback loop where people are just talking to each other and making decisions and no one actually has an answer to what the future is going to look like. And so you basically just are defining it in real time in group chats and in the actions
20:36that you take as a business. So I think that like learning that in real time has made me totally one question the existing world and structure like basically nothing's guaranteed and anything could change. Uh and two it really gives it's quite empowering. It makes you realize you're basically a couple of decisions away from even greater outcomes and you know an even bigger business than you had prior.
21:00Totally get that on that anything could change. One thing that I hope changes is margins and margin profiles. um help me understand when we look across the wave of incredible businesses in AI the margin profiles are still lower than they were previously 30 to 35% say as a barometer compared to 70 80% with SAS right is that a momentary period in time where we're in a buildout phase and they expand over time or is that just a net new but the revenues are going to be much larger
21:30well not all businesses in AI have bad margin profiles I mean I know we've got good margins and and and that actually comes comes in a way that I think is also customer aligned in that we are so focused on sort of thinking about these outcomes themselves as a uh as the thing to be valued and so we've kind of gone away from a lot of common paths that we see other AI companies doing we don't subsidize consumers uh in like the dev tool space that's you know that hurts us
21:58in a lot of ways we don't have the mind share from self-service users so you have no PLG I would separate uh self-service from PLG We have a lot of focus on PLG within companies that we've deployed to to allow the adoption to increase. But what we don't have is like a consumer public facing plan that is I would say super rational for a current consumer unless you're optimizing for quality. Should you in a land grab environment like today should you I think it's a really good question in
22:27our in my mind the biggest reason for this is there are two players with effectively infinite money who are trying to flood the market and their intent is if we flood the market we keep you but remember what we talked about just a second ago you don't keep them after you uh after you sort of pull back the subsidies as we've learned and so in I think that probably the consumer will continue to follow the sort of like most cost effective solution. And so what does that look like in 1 2 3 years? I
22:56think it's open models. I think the most cost effective solution for a model is going to be the cheap one that you can run locally on your computer. And so we want to make sure that we are the product that best fits that type of experience. So that's why we're so optimized on local and on open models today for onrem, but in the future it's for all consumers. And so we won't have to subsidize as much as we'll have to create an amazing experience for self-service. So I think of this as like a long con where eventually you know a long game where eventually the
23:26self-service will come to us but not because we subsidize but because we have the best product in market.
23:31So if you're advising me as an investor today on on how to think about margin and how that plays into my decision to invest or not in a business what would you say?
23:40I think that uh for us this is a part of our strategy. It doesn't mean though that it's the only way to win. I do think that there's probably going to be businesses where uh they're able to lock you in because of a workflow or a system of record that they produce and the margins or the subsidies can be a route or reduce temporarily reducing margin can be a route towards gaining you know this is a classic strategy right it's not this isn't even new to AI um you know gaining customer base what I would
24:08say though is that if there's no path to increasing the margin profile that is very risky and I think that a lot of investments ments are being made in businesses where the promise is simply that they will like raise prices but you won't see a consent increase in the value of the platform. It is so competitive in AI right now if you are not also raising the outcomes and the value that you get out of the product while you raise that price people will churn and move to another thing. So I do think it's it's quite tricky and the
24:36margin profile actually matters a lot but it's not end all be all. Well, you have that churn in enterprise sales. You work with some of the biggest companies in the world. I'm sure you sign year-long, minimum, multi-year long. Y you have pretty sticky client bases there. No, I think so. I think that people also see these year or multi-year partnerships as just that, like a partnership. Uh, part of what makes it interesting to build right now is that a lot of what you're
25:05selling is not only the technology, but your knowledge about how to best use that technology, which I would, you know, carefully differentiate from consulting or professional services in that you don't actually have to go in and do all of the implementation or and I honestly think if you have a product that requires 100 FTEEs to get it deployed, you just have a bad product.
25:26But instead, I think that if you have the advice and the knowledge of the direction that you think the world should go on, you're selling that with the product. And so people are willing to go and buy that. And I think that if they see it from you today, they sort of know that in a year you'll also still have that same knowledge and sort of forward thinkingness. And so now obviously lots of people can give that.
25:47But if you combine that with a product that then acts a little bit more as a platform or a system rather than a tool that people use, you can also get stickier by just being something that you know you build on top of.
26:00We spoke about the different frontier model providers essentially having this really challenging dynamic of being locked into their own models when serving say code or any application that they choose to serve. One then thinks that the value becomes in the rooting of models like the tasks to the model what's optimized for each use case cost latency function whatever it is and open router gets bought for $8 billion uh all the value in the routting
26:28great and then everyone is doing rooting and like ramp has a rooting provider you know one of my companies merge.dev dev has one. We're in another startup requesty and I'm like, well, the routine is completely commoditized, right?
26:43Help me understand what world do we live in?
26:45Yeah. Well, I think that, you know, routing is a really interesting technology in that I don't think the technology itself is necessarily that differentiated. Uh, and so if somebody comes up and says, "Look, uh, you know, Stripe bought Open Router for 8 billion because of the technology, then I'd say either A, if they have insider knowledge, then that was a bad decision.
27:08If B, they're just sort of assigning that to it. I think they're missing what I read when I read the letter to shareholders, which was that they see this as a bet on where the direction of capital allocation is going." I mean, think about it like this.
27:23What is what are tokens other than intelligence? And how do you get tokens?
27:29Well, you pay for them. And how does the infrastructure layer get it? Energy. It's literally like translating energy into intelligence. And you're just trading dollars along the way. All of this is basically converging towards one thing. Whether it's you call it allocating energy, allocating intelligence, allocating money.
27:47Businesses need to allocate whatever this is in order to grow and expand how they operate and grow and expand their bottom line. If you're Stripe, you already control the flow of one out of the three of these. With Open Router, rather than the the routing technology, you actually just gain the information of where these models are going, right?
28:08You start to understand what are people doing with intelligence, how are they allocating it to what models are they using. So now you start to control the second of these three things. Maybe they'll make a play into, you know, energy, infrastructure, data centers at some point, but just ownership over those two is a massive bet on how to think about where companies allocate uh resources. And so for them, I think this is very reasonable. But what's interesting is you wouldn't pay $8
28:35billion for uh the same company that had no users with better technology. And so that difference I think is really important and it gets towards that broader idea of the technology is just no longer the moat.
28:47Do you think it was a good buy at 8 billion? It would have to be really really foundational to the team that becomes whatever this next bet on allocating you know capital is for Stripe uh or or rather helping the businesses that Stripe has as customers allocate capital. I would say that if they think that data gives them insight into how to run Stripe better as well,
29:17that could also potentially make it worth it. 8 billion is quite steep though. So, I mean, stranger things have happened. It feels like 8 8 and 10 billion is the new 1 billion. So, I'm intrigued. You know, you obviously have a rooting product within factory.
29:31Um, what do you see? What insight do you get from that that maybe the world doesn't see? Well, one of the most interesting things and I've been talking about model routing for these products that you've mentioned, which is gateway routing. And that's sort of how it's referred to because ultimately the routing effectively happens outside of where the task is being completed. And so a lot of companies will do this.
29:54They'll look at ramp and stripe and open router and they'll put a model gateway and that gateway is just how all of the different tools and products at the company route to LLM. What's interesting is that we've seen that you can definitely get some nice cost savings doing this like 10 20% from these types of products. Um, but you really need something fundamentally different when you have agentic workflows because to actually take the most advantage out of
30:23models, you need to do something that's a little different from just routing. You need your agent or your system to dynamically understand the task it's working on and understand how to allocate uh intelligence uh in a much more stateful way. And when I say stateful, all I mean is like you need to know not only what's going on today, you need to know what just happened and what's going to happen in the future.
30:45And that can't happen outside of where the task is being completed. You have to be in there in the task.
30:52Is that related to the importance of context window expansion that everyone talks about? I'd say that uh in a way that taking the context window and expanding it was solved not at the model layer like outside at the endpoint but instead inside of the agent with something called compaction. It's exactly similar.
31:12People really want the problems to be solved like sort of somewhere else like in the model or in the gateway but more and more we see it's the harness that solves these problems. And I think that the uh the thing that is sort of underestimated about why the harness continues to solve this is that the harness is effectively the new sort of application. It is just where all the logic happens. It's where the state is maintained. It's the easiest and the
31:39best place to basically do work with AI.
31:43And so as that starts to accumulate, I'd say like more advantage or technology benefits, uh I think we're starting to see people ask questions like, should I be building my own harness? Should I get a really great harness? What is a harness? Uh and that's something that I think at Factory, we're trying to spend as much time as possible educating people about sort of what does well in the harness versus what can be done outside and how to think about like building versus buying a harness. and
32:13context window expansion is one and then continuous learning and the kind of the rise of the first truly continuous learning model. Does a continuous learning model help or hurt factories business? I think that uh well it's interesting cuz there there's sort of two directions for this. I would say that there was an idea of continuous learning from over the last couple years that said that you would have a like a literal LLM like model where all of the learning happens internal to this closed
32:42loop system. Uh that technology has not been developed. It doesn't exist.
32:46There's sort of attempts and early looks at it. But in general, uh, anybody who wanted to try and hold all of the learnings behind an API, uh, would be able to potentially accumulate advantage that would, uh, make it harder for others to use that model in their product because ultimately they would be accumulating all of the learning. But in reality, what has happened is quite the opposite. Basically, model providers have even acknowledged that all of the
33:15continual learning happens at the harness layer. And that learning is basically something that I would argue businesses are going to find very critical that they own that they are the sovereign of. And I think that that is ultimately the question of the next 5 years of AI. Who is the sovereign of your intelligence? Is it you or is it some other company?
33:39Sorry, can you unpack that for me? What what does that actually mean? Who owns the data outcomes that are generated from the tasks that you do? basically who owns the learnings and the workflow that successfully achieved the outcomes for your business. The a great example for this would be if you were a law firm and all you did was just outsource every single one of your cases to some other company and in fact maybe it's like 10
34:08different companies. uh then you know come 5 years later those other companies can just turn around and screw you over because they know exactly how to do your entire business. And what's interesting is I think that pretty much every company in the world is thinking to themselves right now what's going to happen in a couple of years if I outsource all of my intelligence to someone else. And what's interesting is that at least you know two of the largest companies that provide models
34:36today have explicitly said we are going to go after every single one of these industries and businesses that we provide like intelligence for.
34:45Do you see with large enterprises the biggest companies in the world are scared of open AI and anthropic coming after that business? I think that not all of them necessarily think they are going to come after the business, but the largest companies in the world are very wary of the model labs coming in and promising them intelligence and sort of luring them into a trap. And that's something that I know you know there are many businesses that are starting to
35:14become aware of this like Palunteer has been quite loud about this idea of owning your intelligence. Microsoft as well, Satia wrote a great piece about this. I think that all of that opining is very spot-on. At the end of the day, if it's not your intelligence, uh then there is just a real risk that either a they come after your business or b if they disagree with what your business is doing, they have a little bit more leverage and control than I think the typical business owner would like.
35:40And when we talk about sovereign intelligence and owning that outcome, is that why on premise is so important?
35:47I think that's a huge part of it. uh to most businesses on premise isn't even about the technology. It's just about the idea which is that if I need to I can take control and ownership over every dimension of this software and we get to stay in control. And that element is interesting because for us we offer an on-prem offering. It's in fact one of our most popular offerings, factory private. And even though we offer this,
36:16a lot of the businesses that we talk to actually go with our SAS model because they just know and have the peace of mind if they need to switch, not only do we have it available, but they understand exactly how it would work.
36:28And so I think that a part of this story is just being able to share with people that we are incentive aligned. If you need this, we have it and you won't lose anything. And so on and owning your intelligence are very similar stories.
36:44How much does it help you or hurt you that cursor was bought by SpaceX? It gives some amazing scale benefits in terms of access to compute, but it does make them model bias.
36:57Yeah, I mean the that outcome for the folks at Cursor is obviously amazing. So yeah, needless to say and and I think that where it helps us is uh one, you know, it's going to be a very hard story to become model independent or rather stay model independent when you're attached to a model lab. So they're going to want to push Grock. The mo the products are going to become increasingly oriented around Grock and that I think will become a challenge.
37:26There's also some, you know, to be honest, trust and enterprise related concerns that they're going to have to deal with with their new brand, but ultimately the team there is obviously incredibly competent. And so I don't discount them as a player in this market. uh however I do think that most of the enterprises are going to have a second look at the idea of sort of seeding their software development life cycle to a provider who is one likely to
37:54be model locked and two uh has an existing sort of history or pattern of maybe struggling to operate in these larger and more secure environments.
38:04Totally get that and no I mean unbelievable outcome for team man amazing team. Can I ask you when we look at the cadence of model development today, it's just so fast.
38:14Yeah.
38:15And like I actually use arena.ai uh as like a discovery mechanism to new models and it's I'm using suddenly these weird models that I've never heard of and would never have used before and I'm loving the output. My question to you is will we see the cadence of model creation sustain in the way that we are today? In other words, the rate of new models keeping on coming or is this a a momentary period at the start of a new cycle? I think that it will likely
38:44sustain for quite a long time. Uh and this actually gets to another interesting property of like the open routers. A lot of people treat model routers as effectively a information or news stream about which model is next.
38:58It's kind of a free advertisement every single time a model drops. you know, now Stripe can tweet new model on our router and you see Stripe's name with this news cycle. So, I think that actually it's very common for people to basically use new model drops and all this news as a way of keeping up with AI in general.
39:15Uh, new models will likely continue to drop as people one, it gets easier. It's just going to become fundamentally easy to build models. And two, uh, this idea of sovereign intelligence also is like humans, we're going to have models with tons of different opinions, tons of different perspectives. And that I think is going to play nicely into how people sort of operate in today's world. A lot of the times you don't go with a business cuz it's purely the best performance. You go because you like the
39:44person who started it or you want to buy from someone where you saw them on the news or on TV and you agreed with their statements. models are going to be like that as well where they emit opinions and they have takes that are different from the ones that are most popular and people will gravitate towards those. So I see this actually just getting much faster and even broader before it shrinks.
40:04A lot of guests on the show before have made kind of bold statements that like 70 80% of the neolabs that we have today will die in a given time period 3 to 5 years whatever you want to choose. Um, do you think that's true? And how would you advise me and other investors on the model or the Neolabs that will thrive versus die in this next wave?
40:27I I think it's plausible it's even more. I I think it could be 80 to 90% of Neolabs die in the next 18 months. And die is going to be a funny word to use because it'll probably be for a lot of them incredible outcomes. So I don't know if it's necessarily doom and gloom as much as it's these businesses may not make sense as independent businesses.
40:49And so a lot of what I think matters for a neolab is you should ask questions like one is this business attached to a durable workflow. Two is that workflow going to change if new frontier models get better? And three, if this workflow were to be introduced uh to a new business, then would that new business figure out something even better? And so basically, is it durable to like
41:19effectively an entirely new way of thinking or new way of working? If all three of those are true, legal is a great place where I think one uh new models won't necessarily get better without access to the data. Two, it's obviously a very proprietary workflow.
41:32And three, we're still going to have legal system in 5 10 20 years. So probably all the Neols focused on legal are going to have great outcomes versus I would argue that there's some places like um a lot of knowledge work that's related to intermediate tasks like people operating in Excel and Jira that's just not going to be differentiated. The workflows are very common and I think that we may not use a lot of tools like that in 5 to 10 years.
41:59So this like general computer use all this other stuff just may not be as valuable as an independent business. I I'm you know the thing that strikes me is just the misalignment in capability progression which sounds like a real word wank but like when you look at like coding and customer service amazing undeniable legal good but not in the same level as coding and customer service and then other things honestly marketing copy visuals a lot of it's it's so not
42:29there like if I wanted to use AI to clip this show it it clips this it misses is both of our face cuz it goes to the middle.
42:36It has no understanding of how to align clippings between an audio edit and a video edit. It's so far off.
42:44Will we see a real multi-year time lag between different sectoral capabilities progressing in the same way that coding has done?
42:53Definitely. And the biggest reason why, for example, clipping a podcast is still such a hard problem for models uh is likely because the maybe handful of businesses that deal with media haven't devoted 100% of their time to just taking that knowledge that lives inside the heads of people and brought bringing it into AI. And the moment that we start to see businesses capitalize on that delta, uh I think the progression will
43:22happen extremely quickly. So it's purely a matter of time before most workflows become something where a business capitalizes on that first bit which is uh a workflow that currently has proprietary data or proprietary knowledge. You don't normally think of clipping a podcast as proprietary but I think for the most part it's a real skill and a lot of the people couldn't even describe how they know when to do the right clip and that intuition
43:50writing it down is hard. Oh, I totally think and knowing like the hook, if you started 10 seconds earlier and the hook was 10 seconds in, your chance of virality goes down significantly. The skill of knowing what is a it kind of goes to taste 100%.
44:06But it it's actually not. I completely get you there. Um, we always hear about the Chinese open-source ecosystem and the questions around security and everything that is in between.
44:17Do you think those are justified or do you think actually we should leverage it and thank them for their capabilities?
44:24I think calling open-source models Chinese models is a scop by the Frontier Labs uh to basically trick people into thinking that they're scary and otherize them. Uh in reality in my mind the open models that come from a bunch of different places are not any different from a existing frontier model from a lab. uh they just happen to have been created by people, you know, a couple thousand miles away.
44:50Now, there are some very real challenges with taking in models from really any business. And I think that what's interesting is it's actually the same challenges with models from like Openthropic. You should ask questions for all of your models. Like one, what is potentially being censored by the creators of these models? Uh two, are these models going to be able to solve the problems that I care about? And
45:16three, uh, if this model becomes, uh, if this model goes away in like 6 months, 12 months, will I be able to switch to something else? And if the and if the answer to all these things is like yes, yes, yes, then I do think that there will be concerns about using those models. Uh for me the Chinese models specifically have demonstrated no examples where they have some sort of security risk or backdoor compared to
45:44American models. Instead they are just biased in a way that American models are uh for their own creators and preferences. These are things you have to be aware of but typically don't change the day-to-day.
45:56So you don't think that American companies should be concerned about using open source Chinese models? I think that as of today, the current Chinese frontier models should be analyzed by American companies for the tasks that they care about. And if you're, for example, working on national security in the United States, you definitely should not be using Chinese models. But I do think that we have to be sort of cleareyed and say for a code review, it's very likely that a Chinese model and American model will give you
46:26the same. I'll give you a great example of this. uh if you are writing a 10K uh that expresses your business's uh current state and some of the examples in preparation uh for you know sharing finances with investors. Uh let's say a part of your strategy is about introducing recursive self-improvement to models and you care about AI and your business is leaning heavily into it. Uh if you use a model from this provider uh
46:54it will block you. The answer is anthropic, right? And so if you are writing a model about American defense or preparation, I highly recommend against using a Chinese model. But all of these are basically contextual based off of the preferences of the creator of the model. So it's just like any other technology. You have to be aware of who created it and you have to be careful because if the person who created it doesn't want you doing the things that you're going to do with that model, it is going to be harder. Uh that's
47:23something I think people need to be aware of for all models though.
47:26Totally get that. GMO from Versel tweeted last night or yesterday about the waiting or usage of open models significantly increasing at a much faster rate to tokens used on closed models or Frontier models. Um what percent of workflows will be completed with open models in 3 years time? In three years, 99% of workflows are going to be done on open models, but 1% of
47:54those tasks is probably going to be 30 40% of the economic value of the future of intelligence.
48:00Wow. But you still think that 60% will flow to open models.
48:04I think I think close to I think almost all usage of models in 3 years are going to be primarily open. But that difference between Frontier and Open is going to become actually larger than it is today. Uh, and I think that this is actually pretty aligned with how maybe that percentage is overweighted, but like how if you listen to how Sam and Daario talk, the use cases that they're talking about their frontier models being used for are incredibly niche.
48:32They're talking about bio research at the frontier. They're talking about super advanced LLM and AI development.
48:40They're talking about security and defense. These are use cases that really only fit a very specific profile of effectively the frontier of science and technology. This is where I think labs like Open AI and anthropic actually can be incredibly differentiated because they already have the muscle to work on those very frontier problems. But if you go into any business in the global 2000 today and you look at you ask any random person what are you doing today? It's not something that needs the true
49:09frontier of intelligence like 99% of the time and so cost will dominate. Open anthropic could release an open model that they then inference on and I think that that could actually be a great business for them. Given the commoditization of the model air like we've spoken about people seemingly chastise Microsoft given that commoditization do you think Microsoft have actually played a great hand in having a little bit of a bet through open AI but not being tied down with an
49:36extensive model investment layer.
49:39I think Microsoft might be one of the best positioned uh hyperscalers honestly with respect to AI because of this independence. Right now, Satia has played a masterful game of getting huge upside from the OpenAI investment and work. Uh Kevin Scott as well, who I know is like sourcing a lot of that deal. Uh you know, the that team found a lot of the potential of what AI was going to be, but they currently realize that one
50:08provider is just simply not sufficient to cover what intelligence is needed in the enterprise. And so now they've moved towards more concretely expressing one that we support all AI developers and Azure should be a place for inference on anthropic and open AI and open models most importantly. Uh but two even if you do want that frontier intelligence you can come here. And I think that that duality that model independence is going
50:36to be massively valuable for a business that wants to accelerate uh really like work for all other businesses. I think it's a very clever positioning because they captured the upside with a bet and now they're capitalizing on the market as a whole.
50:51Is Zuck wrong then to be putting as much money as he is into Spark and building out that program?
50:58I think he's right for humanity in that we need more open models like that, especially like Americanmade open models. I think that a lot of people, you know, even with respect to what I said earlier on Chinese models, people are going to buy us and and so open models are great because it just increases uh adoption in America and abroad. Uh but two, I think that as a business, they're going to have to build on top of that and take what they did with Meta, the consumer, the monstrous
51:26consumer business that it is, they need to power all of their operations with these new models and the investment will be well worth it.
51:35So, would you buy Matter or Microsoft today if you could only buy one?
51:39If I could only buy one, Microsoft for sure.
51:41Wow.
51:42The biggest thing that Microsoft has going for it is that infra. They own so many of these data centers. They're spending so much on buildout. No matter what model runs on top of that, Microsoft is going to win.
51:54Do you worry about the debt cycle? And what I mean by that is there is so much cash being put out into the data center build out and and we've just never seen levels of debt like this before and you're seeing that in bond pricing for matter and other things.
52:09Do you worry about that sustenance or do you just think [ __ ] it, we're still so early? I think that it it actually should start to concern investors uh in that if you do not have a huge amount of free cash flow then taking on massive amounts of debt is bad it's dangerous and so for the Microsofts and the Googles of the world uh they have access to businesses that are durable are existing with huge
52:38barriers to entry uh and I think that as a result those businesses may take a hit from a future like for example collapse in value or even if it's not a collapse just like a minor hit in the projected future cash flow from AI those businesses are still going to be around and so I think it's a a a risk but to be completely honest if you're open AI or you're anthropic the hundreds of billions in free cash flow that you need in order to pay back the debt that
53:08you're taking on in order to accommodate these data center buildouts in order to get the next big training run. It's totally existential for them and so they need to become the single greatest free cash flowing businesses in the history of technology in order for them to just live and so it's a pretty massive bar.
53:27Do you think they should be moving into the chip layer as they are? I mean we've got the wonderfully named Jalapino uh and then we have Anthropic, you know, now reportedly working on their own chips. Do you think that is the right move?
53:37I think so. I think verticalization is clearly the strongest way to unscrew yourself from h taking on a massive amount of debt and burden. Uh in the future if they become multi- trillion dollar companies they will simply have to enter in this market and own more of that infrastructure layer. Uh so it seems like quite obvious that they want to play there. Uh I think the biggest question is how this changes their relationship in the midterm because if you are uh with with the their current
54:06vendors uh because ultimately you know their explicit goal is to replace their dependency on you. So I think that that's going to be an interesting challenge to navigate.
54:14I think it's a little bit like competition in VC which is like ah no one really has any loyalty anymore.
54:20I'm so sorry to say that Jensen's like I'm working on Nematron. I'm buying poolside. Sam knows that fully. It's coopetition with everyone.
54:29Yeah. Listen, this our job is to survive. Do you worry about where we are in the market today? I have like, you know, older, wiser friends being like, Harry, this is peak froth. And then I'm also like peak froth, but also cursed just sold for $60 billion after 4 years. That is cash that's coming back to hospitals, foundations.
54:50Mhm.
54:50That ain't froth or IR. That's cash back.
54:53Yeah. No, I mean I I think that what I am less concerned about is a like 2008 style financial crisis or massive bubble or asset crash. I think that that seems disconnected from the true reality of where this technology is and is going. we've already started to see outcomes in science and in sort of like real serious human like prosperity
55:22style changes now that it's very early but the technology pretty clearly to almost all experts in the fields of science and technology is on a trajectory towards accelerating human prosperity that is hard to deny from a outcome perspective. Now what's more interesting is who are the businesses that are going to be the sort of backbones of that transformation. We are probably in like the Yahoo era where we
55:51don't actually have or at least widely recognize the Googles of the world or like the sort of the thing that comes after. And so today when I look at open anthropic there are I think more analogies to Netscape and sort of these technology companies that were first but ultimately uh you know it's hard to navigate being first. I think Steve Jobs always had like a really great strategy at Apple of not being first but of being the best at almost everything they did.
56:17And that's something that we also like a lot you know being first or best but we heavily lean towards being best.
56:23Totally get that. I'm continuously changing my mind on outcome sizes. M you know we're an investor also in a lovable of the world and suddenly it's a 13.5 billion business at 600700 million of [ __ ] dude the the trajectory of company growth is just unparalleled do you think investors need to change their mindset on outcome expectations and company growth expectations
56:50yeah I think that it is hard because there's a current space where people are experimenting and they're buying a lot of technology in a fairly speculative way. And so it's hard to index on AI for every other possible industry. But you start to look at some of the other industries that are being influenced by this wave. And you know, even like CPG, right? It feels like every other day you hear about a massive brand that got
57:18bought for billions of dollars that started 2 or 3 years ago. So I think that maybe it is just true that the world gets faster. It grows bigger, better than ever before. Uh, and we're starting to see the early days. And I think that a lot of people say this is what the singularity will feel like. Things will just move faster. Things will grow bigger. There will be more.
57:39And we'll start to normalize it and build models and say, "Oh, yeah, that's just the way it is today." But it's I think it might just be us expanding as an economy.
57:51Before we move to like internals, which I do want to touch on cuz you got really interesting takes on hiring. Uh final one, we saw Air Table go for $2.5 billion give or take. Um listen, it's a fantastic outcome, incredible, but it's just a reduction from the 11 billion price before. Will we see a generation of SAS companies sell exit before we see this wave of cannibalization that could occur? Yeah, I mean I think that certainly we will
58:17and I heard this uh from uh a fellow founder who told me this and I totally resonate uh you know it basically contemporary SAS businesses are more like movie studios now where you have to hit a blockbuster and you have to keep hitting blockbusters in order to keep the attention of the world. And if you are an air table, you made that like one movie that was a hit and people loved
58:46it. So no doubt it's a great business, great outcome. But if you sort of rest on that, then yeah, bending spoons will come and eat you. And I think that the new world is much more about continuously getting bigger and growing larger. And so I won't be surprised to see a huge wave of M&A of these businesses because they're still good businesses fundamentally or you can make them good businesses. and they just aren't going to be like Stripe or these
59:14massive things that capture fundamental pieces of the economy.
59:17I think a very good parallel is also like gaming companies where it's like you have a banger of a game and you will have a hardcore user base that will sustain for 5 seven years and a life cycle that's great but you need another big big hit 100% and totally get you there. Listen, your hiring is candidly different. When we were chatting before, you said we expect 100% of our future hires to come through acquiring companies and bringing their founders and teams into factory.
59:44It's really I read this and I was like honestly I was like wow that's hard because often founders make bad employees. I would suck as an employee. How do you determine a team will thrive internally within factory or o that's an opinionated pretty arrogant egotistical founder. Y it's not a team play.
1:00:06I think that's a great question. To me, the most interesting thing about this is that the profile of an organization has changed so rapidly that acquiring an organization is no longer what it was 10 years ago. And what I mean by that is if you are someone who just created an open source project, then you are quitting your job and you're spending five, eight months just building that one thing. And that shows so much more conviction than
1:00:34you know you can track in an interview.
1:00:36And so it's much easier actually to spot these talented people who are sometimes companies of one who are just they're ready to execute. They want to be a part of a mission or they're already operating towards a mission. And basically what you're offering them is more resources to do that. And so I think that a lot of what we're looking for when we try to bring a team in is are you missional aligned? Are you someone who's going to operate independently and be able to take on a huge amount of responsibility? Um, and
1:01:05what's interesting is all of these people are also aware of this lack of technology mode. And so they're pretty willing and ready to basically either integrate everything they did in a couple of days or scrap what they've been working on in order to build something even bigger. And so for us, this profile is just such a match made in heaven for a team that already operates with a ton of former founders, a ton of people who were previously working at startups. Uh, can I be a dick?
1:01:32Mission aligned course. No one wants someone who's not mission aligned. And then also like willing to take on responsibility. It's not like groundbreaking. Do you know what I mean?
1:01:43No, I No, I totally get what you're saying. In my mind, the mission aligned for us means that you're literally working on the exact problem that we're working on and doing it very well. So, I think that when people come in and they say, well, I loved that we were working on payments. And in a way, payments is just like AI for software development, you're kind of like, all right, I'm I'm sure that you absolutely could work on this, and in fact, maybe we'll actually hire that person, right? So I'm not suggesting that they're if you worked on payments you can't join factory but the
1:02:12people that we're looking at are already deep in the weeds of building harnesses for software development where they've already built something that you know tens of thousands of people are using on a daily basis and maybe they even say that's better than the stuff we're getting from factory or they've thought so deeply about the problem of outcomes in AI and measuring that and they already have sat down with business leaders to say I want to solve how to translate these AI inputs into AI outcomes. And so mission align to me
1:02:41isn't like a property that you can suggest or say. It's actually extremely evident in the work of the founder. And so that has made it very easy to stress test. I think are you going to do well at a company cuz you're basically doing the same thing except backed by us.
1:02:57mission aligned goes contra what Chamath has got a lot of heat over the weekend for saying yeah uh which was like Silicon Valley has been too become too money centric and that's now a problem do you think he's right I think that that is something that it sounds like would say because frankly that's how he's made his money I like I think about what we're doing and how we got started when we created the concept of the software factory something that he you know loves to use that term as
1:03:25well uh you know we said to ourselves we are looking at this very hard fuzzy ambiguous problem we're every we're eating so much glass we're going up to people saying this is going to exist this technology here's how it works and people would say you know leave like go away like it has nothing to do with the pursuit of an outcome because at that point you're literally trying to almost be right about a technology and so it's like sort of very producty it's very
1:03:54engineeringheavy it's kind of contrary to what the world is saying or at least people are saying maybe at some point in the future they're kind of dusting you away. In my mind, that sort of mindset of I'm trying to build a thing. I see the way the future might look. I'm going to get super in the weeds. I'm going to have people tell me I'm wrong every single day for 3 years straight until they eventually agree with you. I think that that's actually filled uh in San Francisco. Like everywhere you go in Silicon Valley, there's tons of people
1:04:23who just care about technology, about doing something that might change the world. Uh, and I think that what you sort of get around that though are people who do want to profit on top of that and their only like way of participating because their background might not be in building uh is to try and like sort of build financial instruments around it. That's actually healthy and important part of the ecosystem, but it's not the only thing that happens in San Francisco.
1:04:50I actually think it's the paradox of what your says, which I think the influx of money has led to a 1% realizing I've got bluntly a lot of money whatever I do. I can always go back to a big company, to a great company, and get paid a lot. So, I'm going to choose to work on something that's really interesting.
1:05:08Yeah.
1:05:09Do you see what I mean?
1:05:10Yeah. And I mean there's just no shortage of people who are just purely trying to realize a dream in San Francisco. It's it is amazing to see.
1:05:19And I think that it's actually quite harmful that the to to that culture that people really want to create a narrative that it's purely financial seeking because to in today's world the words that you say and how you portray something that becomes a part of the story that the intelligence systems that we're building ingest and they they're like world model LLMs and the tools that we're going to use to do work on a daily
1:05:47basis for the decade. You know, that technology is built on the stories that we tell. So, I'm always trying to share a little bit more of the optimistic side of how I perceive the world to be because I think that that actually helps make that world occur with a higher probability.
1:06:03We've talked about team additions.
1:06:05Cognition place a lot of emphasis on like the chess champion and like the math prodigy. Do you think we are overweighing the importance of traditional uh certification or do you think that is the right thing to focus on in a more verifiable engineeringheavy hiring process?
1:06:26Yeah, I think that it is conventional hiring wisdom to look at pedigree and look at achievements and say that's like the right way to pick people who are going to be smart. Um, I think though that in many ways the sort of like least agentic path that you could take is to only try and hit the goals that other people set in front of you. And that tends to look like you go to the right school, you do the right competitions, you like follow the rules well enough that you then get recognized for how
1:06:55well you follow rules or like operate within the system. Now actually there's plenty of smart people who are going to do that because that's also how you almost guarantee a great outcome for your life. So to be clear, you can still find many smart people who follow that path. But in our mind, the most important trait in an individual to hire for is how capable you are of operating outside the bounds of what today the system calls the rules. And that is something that is very hard to measure
1:07:24for. And so ultimately I think you need to look outside of that traditional pedigree and start to look for people who a that system might have overlooked. Uh so like I know that I'm a big proponent of this. You know I mean I went to an Ivy League school. I learned firsthand that that is barely a signal for competence. Uh there are plenty of idiots who went to Ivy League schools.
1:07:46And I think that the clearest signal for me that someone's done something great is that they have built something that they care about that they want to tell the world about. And I think that you can see that all over. And that's why I say it's not just companies that we think will sort of make up the people we quote unquote acquire, but it's, you know, one person shows who just built something in their spare time that demonstrates that they are going to go outside the boundaries of what traditional systems would reward. How do you think about placing a value on those
1:08:15one player small player teams? You know, we're seeing um poolside being bought and a lot of the employees moving over to Nvidia. Uh $12 billion rumored price.
1:08:25How the how the [ __ ] do we put a price tag on ads?
1:08:28No, it's a great question. I mean, and this is ultimately probably one of the biggest uh challenges of capitalism in general is this like desire to place a value on humans and talent, which is sort of challenging. It's like obviously basically some of the best outcomes in history have come out of effectively like one person making a gut check or the right call. People say this about like Jeff Bezos is he worth a hundred you know $200 billion or is it the company that built it? In my mind there is something magical that happens in the
1:08:57connections between the people. So it's not that any one node is worth, you know, a hundred million dollars, but when you put all of these nodes together, the graph they make, that can be worth tens of billions of dollars.
1:09:10And so I think that that's actually one of these interesting parts about a talent strategy is that you have to constantly be thinking about the graph you're building. And those nodes can come in and they're almost a little bit more uh you know there's traits and properties that you want them to have, but the best people are people that make that graph much stronger than it was before. And that I think can easily be worth 10, 20, 30 plus billion dollars, especially to companies who are operating in a model where their talent
1:09:38was built pre-AI. Any business that was decided on their graph before this insanely, you know, game-changing technology was created has to update their graph very rapidly. And by bringing people in, that can be the difference between a $2 trillion company being a $4 trillion company. So almost anything's worth it.
1:09:57Penultimate one before we do a quick fire, what do you see other founders make in terms of mistakes on hiring that makes you go, "Oh no, you know, Sarah or Simon, I wish I hadn't done that. I think the biggest thing is performative work culture. Like people who look for people who are saying, "I'm going to grind 24/7 and I'm going to be unstoppable and I'm going to work myself to a bone." I think a lot of founders see that and they sort of are catching a signaling that this person is going to
1:10:27be really productive. But what we have found is that this sort of performative work culture, this like 996 sort of attitude is almost always correlated with making up for some other, you know, detractor or trait that, you know, basically means that this person might not be a great hire. And I think that that actually applies at the company level as well. All the companies that sort of for the most part say, you know, we work people to the bone Saturdays and Sundays and we've, you know, have to
1:10:55have people, you know, in all the time. I think that at different points in your company's life, you will work on weekends. You will work, you know, 15 hours in a row. Like that's going to happen. I think trying to make that your culture points out that your business doesn't make a ton of sense without it.
1:11:13I get in a lot of trouble in the UK and in Europe for being kind of the 996 guy.
1:11:18And I think that people take 996 from me too literally. I definitely do not mean 9:00 a.m. to 9:00 p.m. 6 days a week. I mean a culture of if I ping you on Sunday morning saying, "Hey, a big client's got a problem. We need to jump on a call with the buyer there, you jump on on Sunday morning, y it probably won't happen, but it's not, I'm sorry, it's my weekend and I will resume on Monday."
1:11:40100%. I mean, and that's just the reality of of building a startup. I I mean, you know, I am constantly talking to people on weekends and we're doing stuff that indicates that we the business operates outside of Monday through Friday. I think though that there's clearly a difference and that's why I say performative work culture.
1:11:58Anytime you create an incentive to check to show people that you're working rather than to actually do work, you're basically incentivizing the wrong thing.
1:12:08And so I think really focusing the business on outcomes, like it's interesting you just mentioned, you know, you jump on a call with a client in order to achieve something. Well, that's pretty easy to see that you're not talking about the act itself. You're talking about the outcome you want to achieve. And that difference is pretty massive when you're hiring. And I just see a lot of people think that they're making the the right call by only selecting for people who have that trait. And I think you miss out on a lot of great talent who knows that that's kind of [ __ ] I think the other thing is senior
1:12:37engineering talent especially with families say is instantly put off by the performative often young hustle culture. And actually I've learned that the leverage that you get from like especially when it comes like infrastructure engineering or architectural engineering y is very real.
1:12:53Yeah. I I mean right now basically being able to point a set of agents in the right direction and knowing from the beginning what direction to go on is worth not only more because it gets the job done faster but it now translates into real dollars. Right? Like if you spend a 100 times more tokens trying to get an outcome because you just don't know as much. It doesn't matter that you worked harder. It just basically means that you missed the ball the first time.
1:13:19I I have Brandon on the show from Mccor and he said that they spend more on tokens than they do on engineering headcount. My dear friend Jason Lamin from SAS who we do a weekly show with on news with Rory um said that we'll give $100,000 of tokens to our best engineers. Y where do you sit on that today and how do you think that changes?
1:13:38We actually don't even think about allocating tokens or credits towards people like that. Like I think that that's actually in fact a very weird way to think about it and I think gets at a measure that ultimately people are looking at inputs. What we think about is how many tokens or basically how much spend that we allocate towards projects and outcomes. So I'll give you a great example. There's an evaluation that we have been hill climbing against in order to try and see if we can build a system that beats it. It's incredibly
1:14:07difficult. It's called program bench.
1:14:08And one of the things that we've done is effectively allocated almost seven figures of like credits in a given in one day uh on this benchmark and that was currently being done by like one person. So I guess you could say that we allocated seven figures of credits to that person but in reality what we're trying to do is we're trying to see does our research pan out on this project and so of course we're willing to spend that much in order to see if that outcome comes true. Uh similarly for a lot of
1:14:38our uh engineers like what we do is we try to say basically scope out these projects and once you know the scale and scope of the project all you have to do is basically share the bid of what you think it's going to cost and then we go and we send it off and we in fact have products uh called agent effectiveness that let you allocate and look at how many credits were spent on a given project in order to measure are you achieving the outcomes that you'd like in your given the spend that you're
1:15:08putting towards those areas. And I think that in this new world, the like relationship of onetoone mapping like agents to humans or like saying like an agent has a name is sort of a weird way to think about it when really you have an agent system and you allocate capital towards projects. So I I see that number for some businesses approaching eight and nine figures easily. What really good idea did you say no to that was very hard to?
1:15:32I think self-service is by far for us the hardest thing to continuously say no to. Uh it is actually quite painful as a builder of products. I want more people to use our product and there's these certain types of adjustments that we could make and I think many of them would unfortunately come at odds with one making our business more successful or two uh you know the experience for the enterprise and that's something that I don't see as being permanent. Like I
1:16:01do think we're going to hit a certain scale where we're allowed to pursue many things at once, but it constantly sort of nags at me that I can't just like hit a button and then 10 million people are using the product cuz when we look in the market at comparable solutions that have, you know, millions of users, basically the biggest difference is economics and that's it. And so we know from a product perspective, we have a lot that's there. It's just that we have to make a hard decision to not uh subsidize.
1:16:28If you own the outcomes of those consumers and you get that data back, could you not make an argument that the improvements that that data would provide to the core product would outweigh the cost to serve those free self-s serve users. So that's actually how we use self-service today is basically we have a product that uh you know obviously you can download it off the internet and you can try it. We we have a to be clear a pretty steady stream of you know tens of thousands of people that use the product every day
1:16:57from that segment. Uh and I think that in the for those self-service users they are giving us feedback and they're helping shape I would say more of like the experience from a user at the individual level. But I think at this point a lot of that um can be achieved with you know less than I'd say 250,000 people. And so once you start to hit critical mass, uh you get that feedback loop. You have everything you need. Uh you don't need like 5 10 million people
1:17:27to get that those bug fixes in. Uh however, there is something really special about seeing the community build like sort of media and content and storytelling around your product. And I think that that's a part of the experience that we have to work really hard to create a comparable for.
1:17:45Yeah. a grassroots community that kind of grassroots brand is harder if you don't have that 100%.
1:17:51Um we're going to do a quick fire round.
1:17:53Let's do it.
1:17:54And in the UK we have a game uh and I'm probably getting in trouble for this and it's called um shag marry kill.
1:18:00Brutal, but you get the theory which is short like you know buy for the short term, buy for the long term and sell hard.
1:18:08Yeah.
1:18:08Um Meta, Microsoft and Nvidia. Oh, this is a fun one. You know, I think that I would have to say marry Microsoft, shag Nvidia, and kill Meta. And I'll tell you why. Microsoft to me represents a software company that has become an everything company that really does sit in the lifeblood of almost every Fortune
1:18:38500. There is not a single business that doesn't have Microsoft something and I think that that means that that company is going to be here for a very long time. I say this also as a former Microsoft employee for a year and a half. Um Nvidia is the current kingmaker of technology. They get to decide who is currently even sitting at the table. And so I have no doubt that that's going to continue to grow massively. Uh and how
1:19:07durable that is. I think that it's just a matter of if they accumulate power at the right rate. If they do it at the current rate and then at some point people might ask questions about are we going to let one company control the whole supply chain? Uh but today I think it probably the answer is what other choice do we have? So they're going to keep growing. Um, and Meta, you know, I say this with with love to the to the company. I think that ultimately from a technology perspective, they're really actually quite accurate often. Like I
1:19:37think the Zuck's push into VR was like technologically correct. That felt like the right move. I think the push now into open models technologically is correct. I think that they have one cash cow, which is their ads business. And I think that that means that they're going to have to work really hard to figure out if there's anything other than that that can sustain the business. So it just makes it the weakest of the three.
1:20:00Will Nvidia be a 10 trillion business in 3 years?
1:20:03I think that there's a real serious chance that if we let SpaceX be worth 2 or 3 trillion, then Nvidia probably is worth 10. Uh and so I think that the answer is likely yes.
1:20:17It depends on the buoyancy of multiples.
1:20:19Do you you said about the entrenchment of Microsoft in businesses? Would you be a buyer or a sell on Salesforce given that?
1:20:26Oh, I'm a buy on Salesforce.
1:20:28I believe that the businesses that are going to be most durable are the ones that have a workflow and a system of record that they've defined that everyone agrees is consensus. So you look at like the Salesforce of the world, you look at uh today at least at Lassian. And I think that the biggest thing is when people say, "I hate that software." And everyone buys it, that's probably a pretty good business because they're not buying the software. They're buying what's underneath of it. And that to me is actually much more durable than
1:20:57the technology itself.
1:20:59I'm an investor in linear. Linear are absolutely crushing though. And they what's interesting is actually, and it's a really interesting one, is Atassian are crushing and so are linear. And it just goes to again the the market size being so much bigger than anyone comprehends. And I think we think too much in zero sum. I take from you and so you lose.
1:21:16Yep.
1:21:17We're both just crushing actually.
1:21:18I I think that's really true. And one thing that's important is Linear and Atlassian are selling the same thing. It's the same workflow, agile, and a system of record that represents agile.
1:21:28I will say though that one of the things that is pretty clear to me is that the way we build software is fundamentally changing. And I think that agile might be one of the things that gets hit with this new way of developing. And that makes me think that there's a huge opening for what the next system of record and the next workflow looks like.
1:21:49And I do think it's going to have to be much more radical. And it will be very challenging for both linear and Atlassian to transform into that new way of building.
1:21:58Codeax cursor cognition and claw code rank one through four in terms of threat level you feel from them.
1:22:06Oh threat level is interesting. I think uh in my mind the current ranking would be because I actually feel that I have to condition this by saying I don't feel a impending threat from the rise of these players because I think that they're actually all going in a different direction from what we're building and that's really important and I'll touch on that in a second. But in terms of I'd say relevance to the conversations when I'm talking to
1:22:33enterprise buyers number one is uh claude code I think that it's just brought up in every single conversation and so we always have to sort of share with people why we see ourselves as largely complimentary to the anthropic platform and suite. Um the second is now uh codeex has increasingly been referenced in deals and conversations where people are saying look like we were on cloud code and now we're switching to codeex. That's actually the greatest news for us because it shows
1:23:02how basically unsticky this is and it gives uncertainty. However, they're coming up way more frequently now and I think it's because their work platform is better than anthropics. So it's not codecs for coding but rather codecs for work that's coming up more frequently which is very fascinating. Then I would say cognition because they're basically the only other modelindependent vendor in the enterprise. And so I'd say that one of the big like sort of consistent feedback points we hear from them is
1:23:30that uh you know they're building this cloud offering. It's very sort of futuristic on the idea of imitating a software engineer as a human. And so I think that that is something that makes people ask is that different or the same as your strategy. Uh and then cursor has is sort of present actually in a lot of these businesses. I don't think that anyone really perceives cursor to be their primary enterprise software development strategy as much as an IDE which is I think still a great business because they're still going to get a lot
1:24:00of usage. But that sort of is how I see them. Uh and this is most informed by this idea of almost all four of these businesses are sort of coming to enterprises and saying we're going to build you a eventually human level AI replacement for labor and then we're going to sort of Indiana Jones swap this for the people in your business. And I think that that is just so different from what we're going into and sort of sharing with them, which is that you're
1:24:27not going to replace human with AI as much as you're going to build a new system for developing software. And humans are going to build that new system alongside AI. And that new system is going to look very unfamiliar. And so it's not so much a onetoone labor mapping as it is a entirely new development methodology. And that's something they're really only hearing from us right now. And I think it sort of contextualizes the many tools in the space compared to factory.
1:24:57Will Chamath be successful with his I can't remember the 1809 or whatever it is.
1:25:02Yeah, the the I I think that uh my answer would be it depends on how real the software is. I haven't seen any examples of it sort of like working in an enterprise environment. Uh I think that if they're very focused on building software that works and delivers outcomes, um I believe that he has just as good a chance as anyone and is very well connected. Uh but ultimately I do think that part of this is about like
1:25:30building with the enterprise. And so I think that that's probably going to be the biggest question is like can they get enterprise traction in the markets that matter with people who take them seriously as a like full-time uh software development opportunity. single biggest advice on selling to large enterprises in today's world.
1:25:49I think the biggest thing that I've learned about selling to enterprises is to stop treating it like persuasion where you're trying to convince them that you're right and instead treat it as a discovery opportunity to learn about what's currently the biggest problem they care about. And that approach difference is I think unique to new markets. So if we're selling something where the market's established, it's finite, zero sum, and everyone knows that it's a commodity like databases where there's a million
1:26:18options, I do think persuasion is the strategy. You're trying to sort of convince them, you know, all else being equal, you buy from your friends. And I think that in this market, it's much more about trying to understand just how big of an opportunity it is and learning with the customer. And I think that people actually put a huge amount of value in enterprise and especially in software on people who they perceive to be trying to problem solve with them.
1:26:45And so if you're on the same team and you're both trying to problem solve, then you're going to land on a real problem that that business has not yet solved. And that means almost 10 times out of 10 you're going to bring value if you can figure out a solution to their problem. So I think that that's like something that people underrate. You're not trying to trick them or like, you know, persuade them. You're just trying to help solve a problem for them.
1:27:07Totally gay. Age old enterprise sales doesn't change that much.
1:27:11No, definitely not.
1:27:12Um, final one. I like to ask the question which is what seems ludicrous or strange today. They will be incredibly common place in 5 years time. And I can give examples of, you know, finding your husband or wife online. Bizarre. Putting your credit card details into your phone. Of course, I'm not doing that. that's so dangerous.
1:27:33Can go on and on and on. What today do we think is crazy that would just be obvious in 5 years time? I think that the biggest thing that we're going to be surprised by is the fact that we let a sort of like priestly class of maybe 2 million people decide the fate of all software for all of humanity. And in 3 to 5 years, it'll be actually like unthinkable that you couldn't just generate the thing that solved your
1:28:01problem with software on the fly in the moment for nearly any problem that you have in front of you that can be solved by information manipulation. And so like today we sort of see a little bit of that with you know lovable and bolts and these tools that let you sort of build personal applications. But I think that this will look quite interesting when you think about what problems not like maybe an individual consumer has but really like a general society we have like you'll be walking on a vacation in
1:28:30Bise and the boat operator that gets you from point A to point B will have a software and interface fully custom to them that looks better than like your you know your your HRIT software back at home. And I think like this sort of total dispersement and distribution of amazing software to the entire world is going to make everything just feel way more futuristic and that I think is going to happen very very quickly uh on the order of like 3 to 5 years from now.
1:28:58You know I I love doing what I do cuz I I genuinely just get to pursue my own curiosity in a very natural way. So thank you so much for entertain entertaining my curiosity and you've been an amazing guest.
1:29:09Thanks for having me. This was a fantastic conversation.
Insights
Research extraction
20VC with Harry Stebbings
https://www.youtube.com/channel/UCf0PBRjhf0rF8fWBIxTuoWA
Direct answer: The transcript contains no fully specified book, article, paper, podcast, or video recommendation. The one authentic content-resource endorsement is Eno Reyes’s positive reference to a Satya Nadella/Microsoft “great piece”; its title and article URL are not provided.
Satya Nadella/Microsoft article or piece — soft but authentic endorsement
- Title: Not stated; Eno calls it “a great piece.” Creator: Satya Nadella (the transcript renders the name as “Satia”), associated with Microsoft. Link: No article link is present.
- Recommender: Eno Reyes, introduced as Factory’s CTO and co-founder.
- Exact takeaway: Eno says, “I think that all of that opining is very spot-on.” The referent is the idea that enterprises should own their intelligence—especially the data, learnings, and workflows that produce their business outcomes.
- Context distinguishing endorsement from mere mention: Eno is discussing the risk that model providers will pursue the industries they supply and gain leverage over businesses that outsource their intelligence. His praise of Nadella’s piece as “great” and “very spot-on” is a clear positive endorsement, though not an explicit “read this” instruction.
Arena.ai — borderline implicit tool recommendation, not a content recommendation
- Resource/link:
arena.ai; no creator is identified. Recommender: the host/interviewer, speaking in the first person. Takeaway: He says he uses Arena as a discovery mechanism for new models and is “loving the output.” - Classification: This is favorable firsthand-use testimony, but not an explicit recommendation to read, watch, or listen to anything. The surrounding exchange frames it as a way to discover new model releases and as part of the information/news stream around model routers.
- Resource/link:
Explicit non-media recommendation for completeness: Eno recommends that American companies analyze current Chinese frontier models for the tasks they care about, but says U.S. national-security users should not use them and that he “highly recommend[s] against using a Chinese model” for American defense or preparation. This is model-use advice, not a book/article/paper/podcast/video recommendation.
20VC with Harry Stebbings
https://www.youtube.com/channel/UCf0PBRjhf0rF8fWBIxTuoWA
- Article by Satya Nadella (title and URL not provided): Factory co-founder Eno Reyes called Nadella’s piece on businesses owning their intelligence “great” and said its argument was “very spot-on.” He connected it to the risk that model providers could learn a company’s workflows, compete with its business, or gain leverage over it when intelligence is outsourced.