The CEO Must Be the Chief AI Officer

1. 0:00 – 1:13

   Why Every Problem Should Start With AI

   1. PFPedro Franceschi0:00

      You wake up, whatever problem you have in your life, why can't you solve it with AI? [laughs] And just, like, start there. I think the CEO needs to be the chief AI officer. [laughs] Like, it's not a engineering team thing. It's not, like, a product team thing. It's like you have to understand the bounds of the technology better than anyone. I think a good proxy for how to spend your time is what are things that only you can do, the models cannot do? You have to sort of refound the very concept of what the, what the, the company self-identity is.

   2. GTGarry Tan0:30

      [upbeat music] Welcome back to another episode of The Lightcone. Today, we're joined by Pedro Franceschi, co-founder and CEO of Brex. Pedro started Brex in the YC Winter '17 batch and built it into one of the most important fintech companies of the last decade. He's here today because Brex has gone deeper on AI than almost any enterprise company we know, and Pedro's own AI setup is so compelling that when he came to YC for lunch, it sent our entire team down a rabbit hole of building on their own. So Pedro, welcome to The Lightcone.

   3. PFPedro Franceschi1:06

      Thanks for having me. Excited to be here.

   4. SPSpeaker1:08

      Thanks for changing our lives.

   5. GTGarry Tan1:09

      Yeah. [laughs]

   6. PFPedro Franceschi1:10

      Oh, God. That, that, that, that, that lunch, I'm

2. 1:13 – 4:08

   How Pedro Became AI-Pilled

   1. PFPedro Franceschi1:13

      like, uh, I think the model company should be, should be sponsoring me for-

   2. GTGarry Tan1:16

      [laughs]

   3. SPSpeaker1:16

      [laughs]

   4. PFPedro Franceschi1:16

      ... the, the token consumption increase I generate. We, we, we, you know, supposedly generated on that lunch. That was the precursor to gBrain, I guess.

   5. GTGarry Tan1:23

      I was still working on G Stack. I was still a 2013 Web 2.0 engineer-

   6. PFPedro Franceschi1:28

      [laughs]

   7. GTGarry Tan1:28

      ... who time traveled instantly to j- the t- AI tools of January 2026, and I r- was, you know, probably half a million lines of rails code in-

   8. PFPedro Franceschi1:37

      Right

   9. GTGarry Tan1:37

      ... and I could create a G Stack because of that to like-

   10. PFPedro Franceschi1:40

       Yeah

   11. GTGarry Tan1:40

       ... help me make a software factory.

   12. PFPedro Franceschi1:42

       Yeah.

   13. GTGarry Tan1:42

       And then after I met you, I realized everything is about freeing the claw.

   14. SPSpeaker1:47

       Free the claw.

   15. PFPedro Franceschi1:48

       Exactly.

   16. SPSpeaker1:48

       I knew you were gonna say that.

   17. PFPedro Franceschi1:49

       Free the claw.

   18. GTGarry Tan1:49

       Yeah. And then-

   19. PFPedro Franceschi1:50

       And give it tokens.

   20. GTGarry Tan1:51

       Yeah.

   21. PFPedro Franceschi1:51

       Yeah.

   22. GTGarry Tan1:52

       Well, no, I mean-

   23. PFPedro Franceschi1:52

       Let it rip

   24. GTGarry Tan1:53

       ... the craziest thing was realizing, like, what I had gotten wrong, that I think actually most people in software are still getting it wrong, is, uh, you tr- they've been treating the LLM like this very precious thing-

   25. PFPedro Franceschi2:06

       Mm-hmm

   26. GTGarry Tan2:06

       ... that's very expensive.

   27. PFPedro Franceschi2:07

       Yeah.

   28. GTGarry Tan2:08

       And so as a result, you have to literally put the agent inside a Foxconn factory.

   29. PFPedro Franceschi2:13

       [laughs]

   30. GTGarry Tan2:14

       And it's like n- like, can you imagine? Like, I mean, that's what the half a million lines-
3. 4:08 – 5:21

   The Electricity Analogy

   1. PFPedro Franceschi4:08

      coding harnesses actually work. And, you know, cloud code existed for probably a year before, uh, but it wasn't that, that valuable yet. And, and I remember, you know, during the holiday break, I was, I was playing with it and, and it was, was pretty shocking, probably similar reaction that, that everybody here had. And I think the question becomes, you know, if you sort of... If you think about, you know, you're, you're, you're sort of standing, y- you know, looking at 200 years of history, and then you imagine you are... We're now in May. You're sort of five or six months after electricity was invented, and most people are still playing with candles and, you know, questioning, you know, what can you do with candles and fire-

   2. GTGarry Tan4:41

      Yeah

   3. PFPedro Franceschi4:41

      ... and you know, like-

   4. GTGarry Tan4:42

      Who needs light?

   5. PFPedro Franceschi4:43

      Yeah, exactly.

   6. GTGarry Tan4:44

      The sun's so bright. [laughs]

   7. PFPedro Franceschi4:44

      What about these lanterns, and what can you do with it? And, uh, and you know, the steam engine is, like, I don't know, maybe, like, 20 years away still, but you know, electricity already exists. That, to me, was the sort of the, the, the, the fundamental, uh, uh, light behind it. And, and I would say I think since then, OpenClaw was kind of a interesting sort of next step, which is, I think, when we realized that, uh, you know, the, the reality is good AI products are agentic loops with tools, uh, and we started doing this in our own product at Brex, but, but then on a personal side, I started spending a lot of time understanding, okay, what is at the frontier

4. 5:21 – 6:56

   Free the Claw

   1. PFPedro Franceschi5:21

      of, uh, using OpenClaw? And I think the insight was just, um, yeah, like markdowns can take you really far, just, like, configuring and automating a lot of the things in your life. It's kind of funny. I remembered I had this, this experience of, like, buying a movie ticket entirely in OpenClaw using, like, a Brex card that was provisioned through an API, and, uh, and then I showed it to my team, and they were like, "Oh, but, like, you can go online and, like, book it in 10 seconds." And I'm like, "That's not the point. You're missing-

   2. GTGarry Tan5:47

      Yeah

   3. PFPedro Franceschi5:47

      ... you're completely missing the point." Uh, but anyway, and then I, I went obviously very deep in this rabbit hole and, uh, started spending a lot of time thinking how to change the fabric of the company and the way we build the products and all kinds of fun things.

   4. SPSpeaker5:56

      Tell us about the, your personal OpenClaw journey, 'cause I-

   5. PFPedro Franceschi5:59

      Oh, God. [laughs]

   6. SPSpeaker5:59

      ... before you came for lunch, I had it, like, installed, but I was, like, way too scared to do anything with it.

   7. GTGarry Tan6:04

      Yeah, we were scared.

   8. SPSpeaker6:05

      We were all scared.

   9. GTGarry Tan6:05

      To be frank. Yeah.

   10. PFPedro Franceschi6:06

       Yeah. Don't get me wrong. Like, we, we, we deal with financial services data. We spend a lot of time figuring out how to, how to be mindful of security and protection, but... And yet, uh, I think people are-A little bit more risk-averse than, than the technology probably requires them to be, given where the technology is. And, and when we, when we started using OpenClaw personally, I started doing it on, on a lot of my own personal setup. Basically, what I did in the f- the V1 was I'm gonna give it read access to everything and just create, like, OAuth tokens to my email, to Slack, and to everything, to just literally not write. And I was kinda shocked how far it got me. Um, and then the next question that we, we spent time on Brex was, okay, like, how do we actually get it to write into our systems? And everybody, including our security team, was, "Well, we cannot do that for all the reasons that we know." And then basically where I spent, I don't know, probably four weeks of my time was,

5. 6:56 – 10:57

   Making AI Safe for Enterprise

   1. PFPedro Franceschi6:56

      okay, let's solve the hardest problem, which is security. Uh, and we ended up realizing that the only way to actually do something about it was to do something in the network layer. Uh, and if you treat the agent like, you know, the agent has its own wills, desires, and, you know, they, they, they go to the Esalen Institute for, for, for, for agents, and, uh, you know, they have tools-

   2. SPSpeaker7:18

      Instead of Foxconn's factory. [laughs]

   3. PFPedro Franceschi7:18

      Instead of Foxconn's factory, they will try to do things at a network boundary that, that could not be the right ones. Um, and we decided to actually just focus on that. So a lot of folks were, you know... And we saw Nvidia and others on Nemo Claw, let's build these, like, open shell pr- forks that have controls over, you know, what tools the model calls. And the reality is, yeah, th- you can do all that, but you can also just make an HTTP request wrong. So we focused on that layer, and then we built this thing called Crab Trap, which we open sourced, uh, probably about two months ago, which is actually the way we use to, to, uh, secure agents at Brex in production. And the, the basic premise is you analyze, you, you HTTP proxy, uh, the entire network boundary of an agent, and the idea is, um, when a request goes through, that becomes auditable, and you basically can, uh, use another agent to analyze the traffic and create a policy to let traffic go through or not. And surprisingly or unsurprisingly, because these models are trained on, you know, hundreds of billions of web documents, HTTP traffic is actually, I would say, probably the way the models reason more so than anything else-

   4. SPSpeaker8:27

      Hmm

   5. PFPedro Franceschi8:27

      ... because they just literally learn on the web. So the, the ability of the model to watch, like, 1,000 requests and make sense of what's happening was way higher than we anticipated. So, so we actually built that, uh, put that in, in production at Brex, and after you record a traffic of an agent operating for a day, you can build a pretty good policy, uh, that, you know, sets things that should be automatically approved. And for things that y- the agent isn't really sure, uh, you can just use an LLM as a judge, and the LLM determines is this request something that should be approved or not based on the policy for what that agent should be doing. So, for example, we have, like, a recruiting agent at Brex called Jim. Uh, we have a policy for Jim, uh, and, you know, all the traffic goes through that same policy, and 98% of requests go through automatically, 2% use an LLM. So we sort of got that problem solved to a degree that we got comfortable experimenting much more aggressively and sort of freeing the claw on the enterprise, which was really hard inside, uh, Capital One. So I would, I would say if we found a way to experiment with these things, and, and granted, we don't do the most aggressive things with this stuff yet. We don't use it on, like, you know, customer data to the degree that we want one day to do, and, and there's, there's boundaries to how we do it. Uh, I don't see any reason why a YC company shouldn't be at the bleeding edge of this stuff.

   6. SPSpeaker9:38

      Yeah, I mean, I think your intuition around, like, the, the proxy at the network level ended up being quite prescient. Like, I think a lot of the stuff that I'm seeing kind of around the OpenClaw ecosystem at the moment, at least, or just agent ecosystem, is essentially doing that. Like, we're seeing that, like, with credentials, credential brokering. Like-

   7. PFPedro Franceschi9:54

      Mm-hmm

   8. SPSpeaker9:54

      ... Agent Vault is doing a lot of that. I think you had mentioned the first version of Crab Trap included, like, uh-

   9. PFPedro Franceschi9:59

      Yeah

   10. SPSpeaker9:59

       ... credentials vault. Why did you decide not to include that?

   11. PFPedro Franceschi10:02

       I think it was just let's just do one thing really well. Um, and, and, and, you know, at the end of the day, I think there's gonna be a lot of solutions that do that. You, you could do credential brokering in other tools already, but the LLM as a judge was, for us, the determining capability to say, "Do you trust this in production or not?" And our security team at Brex, very rigorous and very good at what they do, for a long time were, "Well, you know, not really." Uh, to getting them to a, "Yes, we actually believe this is enough," uh, was a big unlock for us. And look, I, I always say this, like, we're not in the business of building HTTP proxies. We are in the business of being at the bleeding edge of what it can do with AI, and to get to the bleeding edge required us to build this proxy. That's why we did it. Hopefully, someone's gonna build, a YC company, hopefully, is gonna build a better version, and we're just gonna go use it. But at the end of the day, um, that's the journey that took us to just sort of being at, at the bleeding edge in a way.

   12. SPSpeaker10:52

       And how much was, was you, like, sort of pushing this forward and, like, how much resistance

6. 10:57 – 13:09

   Why Most Companies Are Behind

   1. SPSpeaker10:57

      did you get internally, and just how did you, like-

   2. PFPedro Franceschi10:59

      I mean, we got-

   3. SPSpeaker10:59

      ... get everyone AI-Pilled?

   4. PFPedro Franceschi11:00

      I think there was a lot of excitement about it, but the way I describe AI adoption inside most companies is I think there's, like, sort of three tiers. There's tier number one, which is your token maxers, like your engineers that are pushing a bunch of code and, and, and typically living inside coding harnesses. And, and those are sort of well known. We know who those are. Then you have the, the sort of average engineer, uh, that is building a few things but, you know, not, not sort of token maxer to the same degree and probably, I don't know, a tenth of the productivity. And then you have, like, the entire rest of the company, and the entire rest of the company typically is interacting with AI in what I call, like, Google Search mode way, which is a chatbot with a few MCPs, um, or a G Suite equivalent. Like, yeah, you have a few tools from Google, but at the end of the day, it's really just, like, a, a search. And I think the, where our thesis was, if you think about the value that AI creates for, like, a, a token maxer, for example, a lot of the value comes from the harness, and the thesis was how to actually build an equivalent harness for other teams that are non-technical. And our whole sort of thinking behind it was, like, that's a lot of what OpenClaw, you know, created, which is this ability that you can self-bootstrap a lot of the capabilities of the agent by the way you-Edit your skills and markdowns and, and, and sort of set up the, the environment around the agent. Uh, and how far can we get this ability to, for the agent to self-bootstrap a capability without anyone actually going in and coding it by hand? So the analogy we use internally for, I would say, the sort of the, the, the company-wide adoption of AI is we, we don't believe in the, yes, like give people a few MCPs and let them go, because I think what people really want is, uh, in my opinion, is really a way of saying, "Okay, this is actually a virtual employee almost that has, you know, it's on Slack, it has an email, I can actually invite it to a meeting, it can join a meeting, take notes," and, and you're trying to replicate that as much as possible. So how do you build the infrastructure to support that kind of use case? And I think the harnesses will, will, will look a little different and probably more like OpenClaw than a coding model.

   5. GTGarry Tan13:04

      Jared and I just did this this week for the first time where we installed Aqua Voice,

7. 13:09 – 14:22

   AI Teammates, Not Chatbots

   1. GTGarry Tan13:09

      and then you open Telegram with the Claw, or actually we have it in Slack now.

   2. PFPedro Franceschi13:13

      Slack.

   3. GTGarry Tan13:13

      And then basically it was, like, me and Jared and, like, three engineers-

   4. PFPedro Franceschi13:18

      Mm-hmm

   5. GTGarry Tan13:18

      ... and our, someone from the events team, and we're trying to put together, um, how do we put together 60 dinners with-

   6. PFPedro Franceschi13:26

      Hmm

   7. GTGarry Tan13:26

      ... uh, 20 people each of attendees from Startup School with, uh-

   8. PFPedro Franceschi13:30

      Nice

   9. GTGarry Tan13:30

      ... with 21, uh, partners and visiting partners at YC.

   10. PFPedro Franceschi13:34

       Sounds like a great problem.

   11. GTGarry Tan13:35

       And then we just basically started talking about it, and then I picked that up, and then I pressed enter, and then, you know, our claw just started doing it. None of us opened Claw Code.

   12. PFPedro Franceschi13:44

       Mm-hmm.

   13. GTGarry Tan13:44

       Like, it just sort of built a bunch of markdown. It did the analysis and, uh-

   14. PFPedro Franceschi13:48

       Yeah, people, people forget that Claw Code isn't magic. It's just, it's just literally a harness around the same-

   15. GTGarry Tan13:54

       Yeah

   16. PFPedro Franceschi13:54

       ... models we, you can use on an API, right? So, so I think that's the, the unlock of... And by the way, there's a few things that Claw Code's doing that I think are really cool.

   17. GTGarry Tan14:01

       Oh, they're amazing.

   18. PFPedro Franceschi14:01

       And yeah, it's, uh, it's just, it's just a harness.

   19. GTGarry Tan14:04

       Yeah.

   20. PFPedro Franceschi14:04

       Uh, and, and you can-

   21. GTGarry Tan14:05

       And Claw can use Claw Code.

   22. PFPedro Franceschi14:06

       Exactly. Claw can use it.

   23. GTGarry Tan14:07

       And Codex, right?

   24. PFPedro Franceschi14:08

       Yeah.

   25. GTGarry Tan14:08

       It really prefers to use Codex these days.

   26. PFPedro Franceschi14:10

       Exactly. It really does. [laughs]

   27. GTGarry Tan14:11

       [laughs]

   28. SPSpeaker14:11

       For everything, actually.

   29. GTGarry Tan14:12

       I don't know why. [laughs]

   30. PFPedro Franceschi14:14

       Exactly. Um-
8. 14:22 – 18:24

   The Case for Tokenmaxxing

   1. SPSpeaker14:22

      really taken off? The thing that we've found it very curious working with a lot of startups early on is a lot of founders are very shy about burning tokens. I think you really get to experience this when you really go all the way.

   2. PFPedro Franceschi14:35

      Garry mentioned this point, which is tokens are expensive, and, and I think there are... You know, I'm in a fortunate position to, to be able to spend on tokens, but I, I would say, I keep trying to picture myself. Imagine if I was, like, 14 or 12 when I started coding for real, and I had the technology we have now. I would be token maxing in the cheapest way possible, and there are people doing that. You know, you look at the-

   3. GTGarry Tan14:57

      Yeah

   4. PFPedro Franceschi14:57

      ... Chinese models, for example, like, they're, they're pretty decent.

   5. GTGarry Tan14:59

      There's a huge, um, [lip smack] a hobbyist community where they, you know, build a gaming rig.

   6. PFPedro Franceschi15:04

      Mm-hmm.

   7. GTGarry Tan15:05

      But then they try to build, like, local LLM.

   8. PFPedro Franceschi15:08

      Yeah. 100%.

   9. GTGarry Tan15:08

      And then that actually is, like, totally reasonable way to do it, so.

   10. PFPedro Franceschi15:11

       100%. 100%. I have a friend that, that, that has the exact same setup. He has his, like, little GPU farm in his house and, uh, and first time I went there, I was like, "Wow, heating's on here. It's really warm."

   11. GTGarry Tan15:22

       It's really hot in here.

   12. PFPedro Franceschi15:22

       And he's like, "No, no, it's my GPUs." [laughs]

   13. GTGarry Tan15:25

       [laughs]

   14. SPSpeaker15:25

       [laughs]

   15. PFPedro Franceschi15:25

       And I was like, "Great," like, uh, you know, power efficiency all, all the way through. It's funny because at, at Brex, and, and we should talk about managing token costs and spend management for tokens, which is a topic we're spending a bunch of cycles on now. I think the, the, the cost part is one. But, but even, even if you take the cost part aside, you know, the first symptom is a lot more people should be complaining about the max plan limits. And, you know, I, you, you see how, what's the percentage of Twitter that probably complains about it, like 0.1%? So, so I think, I think people are probably still early. To me, there's this, like, the AI pill test, in my opinion, is whatever problem shows up in your life, do you default to AI first or not? It's like, of course, mechanically you can do it, but there's a point that it becomes like second nature, and then your whole, like, brain gets rewired, and you cannot think in a different way. And, and, and there's the whole topic about AI dependency, human machine interaction. Yeah, there, there's all these things that we can, we can talk about and put in the corner. It still sort of surprises me how many people you go talk to about a problem, and I'm like, "It's so cheap to intimately understand the bounds of this problem now. Like, why haven't you done that yet and come in with, like, a much more digested view on the problem?" And I think the second thing is, like, I, I think, I think if you have the luxury of building a company now, the fabric of the company from day one can be built in such a different way that I think I- if I were to start a company today, I would say, "Okay, the premise is why can't it be just me?" Like, and then you start from there, and your token consumption is probably gonna be a lot higher than if you said, "Well, I'm gonna have, like, three people or five people or seven people." And, and I think the, the fundamental constraint isn't as much, in my opinion, uh, like, uh, uh, oh, like, like AI as a cost savings or I'm gonna be more efficient. I think the unlock is, like, the fabric of the company just looks very different when the boundaries become type systems, interfaces, agents talking to each other versus people. Uh, and, and it, it... I, I think people are still didn't fully grasp by, okay, what does it mean to build code with new agents? Like, and, and the new technologies we have, I think that's, like, well understood. But how to live in a world where intelligence is on a tap, and your default answer is, "Let me actually solve this problem with AI first," even if it feels suboptimal, and then from there saying, "Okay, how do I actually make it optimal?" Because I think for the majority of problems, there is a way to solve it with AI that is probably better, and your job is to figure that out, even if it's gonna take you more time, because that will compound.

   16. SPSpeaker18:04

       YC Startup School is back. We're hand selecting the most promising builders in the world and flying them out to San Francisco for July 25th and 26th to discuss the cutting edge of tech and startups. Apply now for your spot. When you started Brex, I mean, like, it's well known, like, your, like, MVP, like, had no

9. 18:24 – 20:54

   The Company of One

   1. SPSpeaker18:24

      web UI or IA, which is, like, all terminal, like super scrappy.

   2. PFPedro Franceschi18:27

      Yeah. Today it would have because-

   3. SPSpeaker18:28

      Yeah, that's what I'm curious-

   4. PFPedro Franceschi18:29

      No one needs HTML and CSS anymore.

   5. SPSpeaker18:31

      Like, was it actually still the right approach, just have a really simple MVP and test that anyone worked?

   6. PFPedro Franceschi18:35

      100%.

   7. SPSpeaker18:35

      Or would

   8. SPSpeaker18:36

      Would you have like a way more fully featured-

   9. PFPedro Franceschi18:37

      So, so I, I, I have this controversial view, which maybe you all will disagree, uh, which is like, I actually think if I look into a pattern of companies that succeed, I think there's a really interesting pattern, which is minimal surface area. A- and the problem is with AI, I think you see... Like, look at Stripe, for example. Stripe, early days, was like literally an API. Brex in the early days, no UI, just like literally a terminal. Um, you look at Airbnb, it's like the web- the website was a form, and the form was just like literally where you inputted what you needed, and then someone somehow went there and figured out how to actually make the booking happen. Like DoorDash in the early days, similar, right? Like it was just like literally... So, so the, the, the surface area was so small with the customer, and so much of the, the band- the sort of the intelligence and the bandwidth of the founders were spent nailing this one single interaction pattern. And I think the risk with AI is that the agency behind choice goes away. So, so you have this, you have this, this, this, this lack of discipline on what matters to solve, and I think people tend to believe that I can just experiment a lot of things, and that's absolutely true, but, but that doesn't preclude you from actually choosing what matters. I, I, I always tell people, like, I think if you don't... If you can't minimize your surface area and, and solve the problem with a very clear set of boundaries, you haven't found the right problem to solve. And I think that's... And you can of course find how to compress the problem into a smaller surface area using AI, and that's really valuable. But I don't think you should use it as an excuse to not do that, which I think is, "Well, I can just build so many other things." But you know, I, I always tell this to people, like intelligence is compression, so when someone comes to pitch me an idea in, in the company, I'm like, "It has to fit in a napkin." Like, "Great ideas fit in a napkin. What's your, what's your napkin?" And then someone comes with this, and I'm like, "I don't know where you buy napkins, but the ones in my house are not this size."

   10. SPSpeaker20:28

       [laughs] How about the step before it then even? I... Like actually, a lot of the pivot advice I give founders during the batch comes from, um, you talking about how you found the Brex idea. And if I... Like the approximate view I remember is that you thought about it as like two-week cycles, and like you're either in like exploration or exploitation mode-

   11. PFPedro Franceschi20:44

       Mm-hmm

   12. SPSpeaker20:44

       ... and you're like trying a bunch of things, but then you wanna like hone down. Like would you still use that pattern now or would you-

   13. PFPedro Franceschi20:49

       Oh, 100%. I think one of the most... one of the hardest things of building a company

10. 20:54 – 28:06

    The One Thing AI Can't Replace

    1. PFPedro Franceschi20:54

       is talking to customers and, and, and, and not, not just having the conversation, but how to extract the sort of unspoken signal from these conversations. And, and I think to me, the, the can AI solve this lens, like, like whatever problem shows up in your life, can AI go solve that? And you think about like building a successful company, like why can't you prompt your way into that? And, and the reason is very simple. It's because there's signal that the models weren't trained on. And the signal is when you go talk to a person and they tell you about a problem they have, th- they're not gonna tell... They're, they're not gonna give you the answer shaped into a prompt that you can put into an LLM and that LLM is gonna go and output the product that's gonna win and be a billion-dollar company. They're gonna tell you a very sort of local optimum answer based on their worldviews and their constraints and the way they see things. And, and I think a lot of the job is... The, the job now is to have the wisdom to choose what you want. And because before the wisdom was not just to choose, it was to choose and know how to execute it. The execution is out, right? The execution's gone, and the models are gonna do that better. The wisdom to choose is still, I think, the, the missing bottleneck. And to me, that all comes from which signals are not in the models.

    2. SPSpeaker22:02

       So say like pre-AI you had personal bandwidth to explore like three ideas in parallel.

    3. PFPedro Franceschi22:07

       Mm-hmm.

    4. SPSpeaker22:07

       You're saying like now in AI world, you'd still do three in parallel, or would you like 30 and let the models try and-

    5. PFPedro Franceschi22:14

       The way I would probably approach it is, is l- like let's, let's pick a broader universe of things as your sort of a early initial exploration. But, but to me the lens is, okay, why can't AI solve it, and like which signal is not in the model? And I think the signal is typically the customer. And, and then, and then when you go talk to the customer, I think I wouldn't paralyze that probably. I would be, "Okay, let's try to get in the head space of this person." And, and, and I think there's like... It's so easy, and, and we, we did a lot of exploration with like synthetic customers and building customer world models and things like that and, and those are really valuable once you know a lot about the customer.

    6. SPSpeaker22:48

       Hmm.

    7. PFPedro Franceschi22:48

       But when you don't know enough yet, I think there's this like very basic thing which is even at Brex, for example, like one of the hardest things for us as a company was we initially sold to founders. We are founders. We knew about ourselves. We knew about our problems. And then as the company got bigger, we were selling to finance teams, and finance teams are different. So, so building that mental model of like what's the value system. Like of course you can eventually make the model represent that and, and, and have that worldview, but, but there's, there's an intangible that I think is, is where a lot of the alpha still comes from. And I think to me is like the, the... I think a good proxy for how to spend your time is what are the things that only you can do? And even in the company of one, what are things that only you can do the models cannot do? And that to me is like one of them.

    8. SPSpeaker23:30

       I think that's so on point. I think a lot of founders like you that successfully navigate a pivot have this loop.

    9. PFPedro Franceschi23:38

       Mm-hmm.

    10. SPSpeaker23:38

        Basically, there's this, uh, there's this book, uh, Others in Mind, from psychology-

    11. PFPedro Franceschi23:41

        Mm-hmm

    12. SPSpeaker23:42

        ... that has to do with people that have very good emotional connection with people are able to simulate what the other person-

    13. PFPedro Franceschi23:49

        Mm-hmm

    14. SPSpeaker23:49

        ... is thinking and what the-

    15. PFPedro Franceschi23:50

        Yeah, theory of mind

    16. SPSpeaker23:50

        ... other in others. Theory of mind.

    17. PFPedro Franceschi23:51

        E- exactly.

    18. SPSpeaker23:52

        And I think the founders that get that and have the empathy to figure out what the customer is not verbalizing-

    19. PFPedro Franceschi23:58

        Mm-hmm

    20. SPSpeaker23:58

        ... is the, what is, uh, make the... I think Garry says this, "Make the implicit explicit."

    21. PFPedro Franceschi24:03

        100%.

    22. SPSpeaker24:04

        Of what are all those desires.

    23. PFPedro Franceschi24:06

        100%.

    24. SPSpeaker24:06

        And they're very subtle signs a lot of time because they're murmurs.

    25. PFPedro Franceschi24:09

        Mm-hmm.

    26. SPSpeaker24:10

        As founders go through them and figure out the insights-

    27. PFPedro Franceschi24:12

        Mm-hmm

    28. SPSpeaker24:12

        ... like, "Oh, is this really a thing?" But how do you know when to poke for it?

    29. PFPedro Franceschi24:15

        Exactly.

    30. SPSpeaker24:16

        And, and the problem with, uh, relying on models, and right now, which is I'm still very optimistic that there's still a lot of job through founders-
11. 28:06 – 32:58

    Building Customer World Models

    1. GTGarry Tan28:06

       as long as there are limits on, uh, RAM-

    2. PFPedro Franceschi28:10

       Mm-hmm

    3. GTGarry Tan28:11

       ... actually, like, there will be.

    4. PFPedro Franceschi28:12

       Mm-hmm.

    5. GTGarry Tan28:12

       So I don't know. I mean-

    6. PFPedro Franceschi28:14

       I think so

    7. GTGarry Tan28:14

       ... that is kind of an interesting one, right? Like-

    8. PFPedro Franceschi28:15

       I think so

    9. GTGarry Tan28:16

       ... literally you can't have a, a model that has enough parameters that could, like, have everything that you could possibly need in distribution.

    10. PFPedro Franceschi28:23

        Mm-hmm.

    11. GTGarry Tan28:23

        Like, there aren't enough atoms in the universe, right?

    12. PFPedro Franceschi28:25

        Mm-hmm.

    13. GTGarry Tan28:25

        It's like a modeling problem.

    14. PFPedro Franceschi28:26

        I, I think we forget that the world models in which the models are tr- Like, there is something that the designers of the models influence the way the model actually behaves in the end. So, so, you know, one of the things that we spend a lot of time thinking is, like, how to make LLMs work for people that look very different from us, uh, how to make LLMs work for, like, the average finance person in the US that if you're talking about an answer and, you know, the model defaults to, like, AI CapEx as the finan- a- as a default category for like, like, for example, that's a really funny example. Like, I was playing with AI for accounting categorization, and then, like, the first example of, like, an example of an expe-- It's just like ex- writing pros, and an example is, like, AI CapEx. And I'm like, "Oh, why is it AI CapEx, the first example it comes up with?" Because the people that are building the models fucking only think about AI CapEx.

    15. GTGarry Tan29:16

        [laughs]

    16. PFPedro Franceschi29:16

        Right? So, so there, there are things like that that I think is, like, uh, kind of interesting to think about, that, that the, the mental models of the models I think are out of the box are, are more biased than we may give them credit for.

    17. GTGarry Tan29:28

        I mean, speaking of AI CapEx, like earlier you were saying that-

    18. PFPedro Franceschi29:30

        [laughs]

    19. GTGarry Tan29:31

        [laughs] You know, we're, we're so early still.

    20. PFPedro Franceschi29:34

        Very early.

    21. GTGarry Tan29:34

        I don't know. It's, the funniest thing about AI to me is, uh, how often I find myself thinking, um, crypto maxims.

    22. PFPedro Franceschi29:40

        Yes.

    23. GTGarry Tan29:40

        This is the worst the models will ever be.

    24. PFPedro Franceschi29:43

        Yes.

    25. GTGarry Tan29:43

        My favorite now is, uh, telling people who hate AI coding, like, "Have fun coding at 1X speed." [laughs]

    26. PFPedro Franceschi29:49

        Exactly. Exactly. Exactly.

    27. GTGarry Tan29:50

        Nice.

    28. PFPedro Franceschi29:51

        I was telling a friend about, you know, how, how, how to, how to be, you know, long inference, that basically the thesis that there's gonna be a lot more inference than people think. And people were expecting a lot of inference if you just look at public markets and, you know, semi supply chain, all that.

    29. GTGarry Tan30:04

        People are saying, like, 10,000X.

    30. PFPedro Franceschi30:06

        Yeah. But, but the, the underwriting, which is kind of funny, is like, I think there's one image, 2,500 dots. Each dot is 3.2 million people on the planet, and basically, you know, 84% of the world never used AI. 16% have used at least once a free chatbot. Then 0.3%, which is, I guess, six or seven squares, uh, pay 20 bucks a month for AI, and one box out of the 2,500 actually use agents in, in whatever capacity. So that's the, that's the argument to be long inference. Uh, and, uh, I think it's, I think it's just, it's just starting out. And I think a funny thing on, on this is I think theYou know, it will be the biggest expense in a company, like, like easily, right? And, and yes, there's a lot of margin in tokens right now, but people always wanna be at the bleeding edge. But even if token costs decrease by 10X, you're gonna have 10X more usage, so it will be a still large cost. Um, and we're spending a lot of time thinking how to help companies actually manage token spend. On Brex, we, we ended up building our internal version of this. We, we call it Magpie, where the idea is you can effectively, you know, every dollar of token spend in the company, you can attribute to a product we have to customers, an internal tool that we use to serve, uh, or an internal employee, uh, and understand model usage, et cetera. And, and we're now figuring out how to build the analytics on what are we trying to do with the tokens, um, to start to get a sense of ROI. Uh, but anyway, it's a s- fascinating topic that I think has a lot of, like, early, early, early work compared to what it will be one day.
12. 32:58 – 39:02

    Rebuilding Brex Around AI

    1. PFPedro Franceschi32:58

       years ago where I sat down with, you know, a lot of the, the engineering product leaders in the company, and we had this question, which was, "If we started Brex again in 2024," the answer would be even more different now, "what would we do differently?" And turns out, like, everything. [laughs] And we started going down this route, and it's like, it's kind of maddening because you're like, "Okay, we, we have this, like, completely old way of, like, even thinking about the fabric of the company and, and the way we build a product and the way we build our processes internally." The first best answer is, "Yes, we wish we had started now." Second-best answer is like, "Let's go do something about it and change the way we do things," right? And I think a lot of our approach in terms of, like, adopting AI has also been, you know, how do you, how do you pause and say, "Okay, like, there is a discontinuity in the-- not just in how we solve the problem, but on what the definition of the problem actually even is," and sort of take a step back and rethink it. Um, and, and you know, like, there's, there's, like, millions of examples of that. But, you know, one example which is kind of funny is, you know, we're redesigning our KYC process. Like, whenever we onboard a customer, we have to do all these checks to KYC the customer. And KYC historically is something that you can automate, like, 80% of it. 20% is manual. Uh, and of course, the original impetus for anyone is let's build an agent that does it. Yes, we can go do that, but what we decided to do is actually say, "Let's redesign the entire process end to end." And then what we redesigned is the entire onboarding process. And when you redesign the entire onboarding process, what you realize is there's a very important thing that happens in the beginning of the funnel, which is deal qualification. Like, is this customer even remotely qualified to be a Brex customer? But when you have KYC for free, you can qu- you can KYC a lead versus a customer. So you start to have risk orientation up in your funnel.

    2. SPSpeaker34:41

       Hmm.

    3. PFPedro Franceschi34:42

       And that changes who you even target because you know who's gonna qualify, and the same thing's true for credit to some degree. So now the bounds of the problem have changed, and, and you can go in and say... And, and I think a lot of, a lot, including a lot of our competitors, had this approach of saying, "Oh, I have this entire old process. Let me go and, like, latch on AI on top of it, or latch on AI on top of our product." And I think the, the biggest discontinuities in a positive way that we've had were when we said, "Hey, let's keep this old way here. Keep, put it in a corner," and, like, how would we design it if we started the company today from scratch? And then just doing that. It takes a little bit of founder energy to do that, but I think it's the, it's the only thing we've seen working to, to really sort of inflect.

    4. SPSpeaker35:20

       I think that reminds me a lot about this is sort of, uh, way back, I don't know, you ever try to compile Arch distributions of Linux?

    5. PFPedro Franceschi35:28

       Mm-hmm.

    6. SPSpeaker35:29

       The culture within s- power users of Arch Linux versus Ubuntu is very different.

    7. PFPedro Franceschi35:34

       Mm-hmm. Very different.

    8. SPSpeaker35:36

       I think the Ubuntu people kind of feel more like people that try ChatGPT. Stuff kind of just works out of the box. There's some stuff that you can get r- up and running.

    9. PFPedro Franceschi35:44

       Mm-hmm.

    10. SPSpeaker35:44

        There's still not a lot of people that use Linux, by the way.

    11. PFPedro Franceschi35:46

        Mm-hmm.

    12. SPSpeaker35:46

        Which I think it feels where AI is. But with Arch, you're, like, super hardcore.

    13. PFPedro Franceschi35:51

        Mm-hmm.

    14. SPSpeaker35:51

        And I think that's what OpenClaw and Hermes feel like.

    15. PFPedro Franceschi35:54

        Mm-hmm.

    16. SPSpeaker35:54

        Because you have to really customize it to your own, own use case, maintain your, your skills-

    17. PFPedro Franceschi36:01

        Yeah

    18. SPSpeaker36:01

        ... have all the markdowns, and if you get it working, you can build something awesome. One of the mo-most impressive thing I've seen people build with Arch is actually, I don't know if you know, uh, Valve, the Steam engine.

    19. PFPedro Franceschi36:12

        Mm-hmm.

    20. SPSpeaker36:12

        The operating system that runs, that makes it feel like a Nintendo Switch-

    21. PFPedro Franceschi36:16

        Mm-hmm

    22. SPSpeaker36:17

        ... is actually built on top of Arch.

    23. PFPedro Franceschi36:19

        Oh, interesting.

    24. SPSpeaker36:19

        They customize all the drivers, over-the-air updates. It works with all consoles. It work with all sorts of hardware out of the box, but they super-duper customized it, and I think this is kind of what's happening. If you get your OpenClaw to work really well for you, you could kind of build your own custom Nintendo Switch for whatever you need to do.

    25. PFPedro Franceschi36:40

        Yeah. I, I always have this thing that I tell peop- which is, which is funny, which is think about your time two years ago. Like-I feel like you're working a lot more now than two years ago, right? And probably the same for everybody here. So then the argument is, "Oh, but what, what about the productivity? Where's the productivity?"

    26. SPSpeaker36:55

        [laughs]

    27. PFPedro Franceschi36:56

        Right? And, uh, and I was talking to the CFO of a very large public company, uh, uh, this week, and she was telling me that we see all this token consumption and, and, and, you know, we're trying to measure, like, like, product velocity, and we're seeing, like, more lines of code pushed. So, so yes, maybe that's the way to measure the ROI, but, but, but is it really there? Because people are spending so much on tokens. And, a-and I think the... I, I think, I think this analysis, like, yes, of course, I think having a sense on ROI on tokens is important, but I think it misses the point that you're standing in the timeline of history, and it's six months after electricity was invented. Like, thinking about, like, i-i-imagine someone saying in, in, like, I don't know, 18- the 1800s like, "Oh, my electricity bill is so high now. Like, gosh, let's use a little less. Let's keep... Let's push the steam engine to come, like, maybe 20 years l-later because the cost savings are..." Like, like, yes, of course, like, I'm, I'm... Like, don't bankrupt your company on tokens.

    28. SPSpeaker37:52

        It's actually a perfect analogy-

    29. PFPedro Franceschi37:53

        [laughs]

    30. SPSpeaker37:53

        ... because I don't know if you know this, but when electricity was first invented, it didn't work very well, and the ROI was actually bad. And so if shortly after the invention of electricity, some of a-accountants had done this analysis, they would've been like, "Oh, this electricity thing is, like... Is it never gonna be a thing? The ROI sucks."
13. 39:02 – 43:50

    The CEO Must Be the Chief AI Officer

    1. SPSpeaker39:02

       who wish that their companies could be, like, as AI-pilled as possible, and you run this, like, big company now with all of these employees, and that's only the Brex side. There's also the, like, how to run... So I'm curious what you've done to, like, bring the rest of the company along with you on this journey, and if you have advice for other people-

    2. PFPedro Franceschi39:20

       Well, I think there's-

    3. SPSpeaker39:21

       ... or other CEOs

    4. PFPedro Franceschi39:21

       ... there's a lot to do. I think the CEO needs to be the chief AI officer. [laughs] Like, it's not a engineering team thing. It's not, like, a product team thing. It's, like, you have to understand the bounds of the technology better than anyone. I would argue that unless you, unless you really experience the, the, the, the limits of the technology every day, I think it's really hard to even understand what, what it can possibly do.

    5. GTGarry Tan39:42

       Oh, you know why? It's because nobody can say no to the CEO except the board, and the board won't be in the weeds per se. [laughs]

    6. PFPedro Franceschi39:49

       That is 100% true. When you go think about, like, you know, the whole, the whole example of KYC that we were saying, like, the KYC team would never think of using the KYC technology to score a lead. The only people that can think about the organization of the system itself is if you have the context of the whole. And, and to me, like, the, the, the single most important question that any CEO needs to answer is forget about the competitive landscape. I-i-imagine you could get the state of the technology today and transport to the, the moment you started your company. The opportunity was still the same, but, but just the possibilities of the way to build a company are totally different. How would you do it? And then diff this versus what you have, and then first suffer in silence for a little bit because you will. I, I mean, I-

    7. GTGarry Tan40:35

       [laughs]

    8. PFPedro Franceschi40:35

       ... I do every day. But then the second thing is, okay, what do you do about it? And how would you do it if you were starting from scratch? You'll be the one figuring out, okay, how do we design our onboarding process or how we design our, our growth engine and our customer acquisition and the way we talk to users and the way we synthesize that data, and all of that would be, would be redesigned from scratch. So, so I think it's like, it's almost like a... You have to sort of refound the very concept of what the, what the, the company self-identity is and the way the functions and, and, and, and people's sense of success gets structured. AI is a, is, is an umbrella that I think has, like, three things in the way we talk about it internally. Uh, there's product AI, the product we actually ship to customers. There's operational AI, which is things that directly affect our ability to serve customers at scale, like think of customer success, risk, onboarding operations, et cetera. And then there's corporate AI, which is how people work internally. The three agendas matter, and they matter in different ways depending on the timing of the company. Um, and, and I think people will sometimes sort of pigeonhole themselves in one of the three. But in reality, I think you have to take a step back and be like, you know, the same, same thing we were talking about earlier. Like, why can't you solve everything with AI? Like, at a limit, that's the question, and, and then sort of start from there and sort of problem solve around that question. It's a turnaround almost. I think you have to assume that if you're a big, large company that's not AI native, you're doing a turnaround, uh, to some degree.

    9. GTGarry Tan41:54

       I guess we've been making fun of Foxconn factories for some time.

    10. PFPedro Franceschi41:57

        Mm-hmm.

    11. GTGarry Tan41:57

        But on the other hand, like, if you look at them, they're like this paragon of, like, very extreme efficiency.

    12. PFPedro Franceschi42:02

        Yeah.

    13. GTGarry Tan42:02

        But they also, uh, are designed to be that, to, like, create, like, one thing perfectly-

    14. PFPedro Franceschi42:08

        Yeah

    15. GTGarry Tan42:08

        ... back to back to back, and so you have to build a factory like that to do that.

    16. PFPedro Franceschi42:11

        And most companies are designed that way, right?

    17. GTGarry Tan42:12

        Yeah.

    18. PFPedro Franceschi42:12

        I think, like, processes are designed not to change.

    19. GTGarry Tan42:15

        Yeah.

    20. PFPedro Franceschi42:15

        There is a certain amount of broken glass required. The question is how... Like, I think it's 10X easier for the CEO to break glass than an executive.

    21. GTGarry Tan42:24

        Oh, by far.

    22. PFPedro Franceschi42:25

        And 10X easier for an executive than an employee. So, you know, a lot of times, like, someone comes to me and says, "I'm trying to do this with AI, but someone is saying no because we haven't tested this in this use case or in that thing." And I'm like, "Okay, what are you trying to do? Like, do you understand the risks? Do you understand the guardrails?" "Yes." "Okay." It takes me literally 10 seconds to solve that problem, and it would take someone 10 hours to go in-

    23. GTGarry Tan42:48

        Yeah

    24. PFPedro Franceschi42:48

        ... into the meetings and escalate and understand, okay, can we-

    25. GTGarry Tan42:51

        Or maybe never

    26. PFPedro Franceschi42:51

        ... can we do this with AI? Or maybe never.

    27. GTGarry Tan42:53

        Literally never.

    28. PFPedro Franceschi42:53

        And I think the conclusion is probably never because most people would sayYou know what? I'm just gonna, like, build this soft- this product in the old way because, like, why wouldn't we?

    29. GTGarry Tan43:01

        Yeah.

    30. PFPedro Franceschi43:01

        It just works. We know it's here and-
14. 43:50 – 51:43

    Building Company AGI

    1. SPSpeaker43:50

       do you buy into sort of like the Jack Dorsey view of every company's essentially trying to, like, build its own little company AGI and-

    2. PFPedro Franceschi43:57

       I, I do, but, but maybe in a slightly different way. I do think domain specificity matters, so, like, I, I don't believe in the, like, oh, I'm gonna have, like, a single company model that has, like, every piece of data, like, in a single, like, with no judgment or lens into anything. So a- and, and the way I think about it more is, like, is more the sort of the, the, the, the virtual employee analogy, so to speak, which is like, how do I build an agent or a virtual employee that is exceptional at understanding everything that matters about this customer? That is a well-defined problem with clear boundaries, with, like, clear APIs of who people, who need, who depends on the data, who interacts with the data. That, that is self-contained. Then there's another, another agent that can be, okay, given all the customers that we have and the problems they have, how do I manage my product roadmap? That can be a separate agent, but that, that builds on top of this customer roadmap.

    3. SPSpeaker44:46

       Like a virtual exec team, basically.

    4. PFPedro Franceschi44:48

       Uh, e- exactly. Functional and domain knowledge still matter, right? The- these things are not gonna go away, and I think the, the way knowledge is structured, I think, still is still true, right? Th- that doesn't necessarily change that much. And, and you, you should separate the agent that it... and, and, and the systems that are actually emitting code from the system that is talking to customers and the system that is reasoning about the conversations with customers and translating into product roadmap. These should be three separate things. We're kind of like the Tesla for AI. We're like, I don't believe in anything that doesn't have real usage. So it's like, yeah, I built this great model, and I'm like, okay, how many people are using it? Is it actually displacing the need to hire a person inside a company? Is it actually displacing the need to, you know, spend a- literally hours, like, how many hours is this thing saving? And, and I think a lot of times people say, "Well, you know, it's a, it's a cool model," and I'm like, "Yeah, but, like, that's not, that's not gonna cut it," right? Once you have that orientation, I think customer roadmap, okay, like, your... Or for example, our client sales team now runs on our customer roadmap.

    5. GTGarry Tan45:43

       Cool.

    6. PFPedro Franceschi45:43

       So I know it works. [laughs]

    7. SPSpeaker45:45

       Mm-hmm.

    8. PFPedro Franceschi45:45

       I'm, I'm actually having lunch with, uh, a customer tomorrow, and I don't know the state of that account as well as I probably should. Customer roadmap answered a question for me, and I now have a report with inc- including things that the team didn't know about that came through support tickets and, you know, an executive that was traveling, had an issue at an airport with their card. They, so all these things-

    9. GTGarry Tan46:04

       Total information awareness.

    10. PFPedro Franceschi46:05

        Total information awareness, right. That is a well-defined problem. That is working. I can trust this building block as part of my company model as a whole. A- and you can have evals on it, like, we know, like, you know, y- you... I, I think a very... We should talk about evals. There's a, a bunch of learnings on this, uh, and, and how to build evals into the fabric of a company. But, but anyway, I, I think it's more of, like, you have to decompose the problem a little bit.

    11. GTGarry Tan46:26

        Yeah. My favorite thing about evals is, uh, just running cross-modal evals against each other.

    12. PFPedro Franceschi46:31

        So one of the things that we're doing that I, I, I... I- it is related, but I think is really fun, which is how do you have every single human interaction in the company becoming an eval when you have an agent? So for example, we have the onboarding, onboarding agents d- doing something, and then you have a team that actually goes in and looks at KYC exceptions that the model can't figure out. How to make that a breaking change? [laughs] And, and okay, like, this, this manual interaction will become an eval case. You know, we have an expense agent in, in Brex. Whenever someone has a conversation with the agents that is... that flags, uh, an issue or, or a bug or, or, or, or something that feels like the conversation didn't go as smoothly, that creates a bug. That bug triggers an agent that's gonna go in and modify the code base and the prompts and everything to make that eval pass. And if that doesn't break, then an engineer's gonna go in and figure out how to make that pass. Because the goal at the end, I think, is to make the whole thing a self-reinv- a self-learning system, right? And I think the... A lot of what I see with, with, with companies is they spend a lot of time getting an agent working, but never thinking how to make the agent improve every day, and I think that's, like, always the biggest unlock of these models.

    13. GTGarry Tan47:44

        You need a dream cycle.

    14. PFPedro Franceschi47:45

        You need a dream cycle.

    15. GTGarry Tan47:46

        The dream cycle sees-

    16. PFPedro Franceschi47:46

        Exactly

    17. GTGarry Tan47:46

        ... everything every night.

    18. PFPedro Franceschi47:48

        Exactly.

    19. GTGarry Tan47:48

        And then it's like, oh, what's going on there? I need to put this over here.

    20. PFPedro Franceschi47:51

        Exactly.

    21. GTGarry Tan47:52

        What actually happened? Is there a pattern? How do I-

    22. PFPedro Franceschi47:54

        Mm-hmm

    23. GTGarry Tan47:54

        ... root cause this?

    24. PFPedro Franceschi47:55

        So how to, how to, how to bake the dream cycle into the product and into the agent and into the things you ship.

    25. GTGarry Tan48:00

        My favorite thing right now is I'm building, like, three or four agents for my friends. [laughs]

    26. PFPedro Franceschi48:05

        Oh, interesting.

    27. GTGarry Tan48:05

        And, uh, part... Some of it is, like, this is, like, user research for me for Gbrain-

    28. PFPedro Franceschi48:09

        Mm-hmm

    29. GTGarry Tan48:09

        ... 'cause it's like I have one. It's working really well. I have 350,000 markdown pages-

    30. PFPedro Franceschi48:14

        Mm-hmm
15. 51:43 – 54:06

    Why We're Still So Early

    1. PFPedro Franceschi51:43

       a 200-year timeline of human history. There's a point in time where electricity was invented. It sucked in the beginning, you're six months after that point. What do you do differently, knowing everything that would be true about electricity, uh, knowing that data centers are now gonna consume electricity and, and, and even AI, right? Well, you would do a lot of things differently, I think. So I, I think that's one of just, just marveling at the possibility of the exact moment in time we're now. Um, I think the second is, like, have a Post-it on your computer, which is you wake up, whatever problem you have in your life, why can't you solve it with AI? [laughs] And just, like, start there. And, and 80%, yeah, you can use a chatbot, but the 20% that you can't, figure out why and go build something that makes you solve that problem, less so because of the, because of the immediate usefulness that solving that thing at scale will have, but because it gives you a texture and a feel for the possibilities of the technology, which are really hard if you're not playing with it every day. And, and maybe the third thing is, like, I, I think it's, like, just measure your token consumption and how much you're just pushing the, the, the limits of the company and starting at the premise of, like, okay, why can't it just be one person? Like, why can't it just be me that to b- that builds the whole thing? And you're gonna probably face a wall of the, uh, the elements of, you know, what, what models can and cannot do. But at a limit, I think the question is, uh, you know, how do you spend your time on the things that only you can do as a founder? And these things to me are, number one, which problems are worth solving, and two, uh, and, and sort of the choice thing we talked about. And the second thing is, okay, given that the, given these choices, what are the limitations of an LLM, uh, that they still cannot do, and I have to go in and do th- those, those things myself? But almost, uh, you know, to some degree, you're working for the LLM to some point. Uh, and, and, and if you're in a bigger company, you're in a turnaround to put the LLM as almost the founder and the CEO, and you're, you're, you're, you're almost architecting the entire company around that idea. But I think early on, uh, so much of it is, you know, choosing what matters, talking to customers, injecting the signal that the models don't have, and just, you know, rebuilding it the way you would do it in 2026 with electricity being six months old.

    2. GTGarry Tan53:50

       Thanks, Pedro. This was awesome.

    3. PFPedro Franceschi53:52

       Yeah.

    4. GTGarry Tan53:52

       This was great.

    5. PFPedro Franceschi53:52

       Thanks for having me. Appreciate it.

    6. SPSpeaker53:53

       Thanks for coming. [upbeat music]

Episode duration: 54:06