Apr 1 '26, Wednesday
Apr 2 '26, Thursday
Apr 3 '26, Friday
Apr 4 '26, Saturday
19:03·2 SPEAKERS·2523 WORDS
ABro life crazy bro. Work my fun name.
BI mean you're at the highest level of abstraction possible, bro.
ALike that is also a part of the worry. Just I just wanted like perspective and like just discuss, discuss that also like in some sense like in terms of like organization and people and like that kind of leverage like within the context of this company and being able to hire people and could I say whatever leverage of like money and funding that we have.
BYeah.
ALike I can hire anybody, can do whatever, like have my own initiatives and stuff. But like ceiling of this company is like low. Like low in comparison to like like the people I studied with at Stanford. They're doing leverage this leverage inside trillion dollar companies.
BYeah.
AAnd like it's not like they are actually like it's like they just launch a training run which actually consumes like hundreds of millions of dollars. Right. Like, like it's not, it's not like it's not like real money. It's not like that. Like it's just that they have access to resources because skills that are applying to this tool. Like it's like a hardware like in a science lab. Right? Like yeah, obviously work with like million dollar tools or something, but they have like 100 million dollar tool and I have a million dollar tool. Like I'm working in a lab and I have full leverage in that lab.
BYeah.
ABut like I have the tools of like, like I don't have the same tools.
BYeah.
ASo. Yeah. So in some sense and then. Interesting research or company decisions or WhatsApp, they ask me and I discuss with them and it's like exchange. Like I'm like still like peripheral, like increasingly peripheral. Like how you handle like the SF office. Like it's a remote like office. Right?
BYeah.
AA lot of like they say water cooler discussions and they say in the room when the idea came about, you are like around and you like as an meeting schedule.
BYeah, yeah, yeah.
AMaybe it's just like this is my worldview and like I'm like I feel like grass is greener on the other side. Kind of. So. Not even like was like $5 billion company and now it has like 2 150x. Like if I got a million dollars early in open A, I would have $150 million.
BYeah.
ABut anyways like. Yeah, like labs, they are like six months ahead in terms of these agents and coding and like all of those things like developing the models also. So like I feel like ceiling is like, like here I feel I'm seeing like a ceiling. Like I'm not able to Fly. Like I'm adding shareholder, increasing shareholder value or whatever.
BYeah.
ABut like I'm not able to like spread my wings and like fly and stuff. Like. And I see my peers, they're all like clashing it, bro. And they have like. So I feel like, I feel a bit. I don't know if it's like crazy perspective or like.
BI mean I think they'll all. You'll always feel that way, right? Like there'll always be someone who's crushing it. You would, don't you think?
AYeah, like. Okay, but then like just like in the room with like five other people, five other researchers and we are all like good and we are like coding and we are exchanging information. Just that phase of like real time.
BYeah.
AWe haven't hired a single research. Like it's been like office has only been there for one year. This whole company for two years. We haven't hired a single researcher in sf single. Like single new research.
BSo what prevents that?
ALike we have hired like very good research engineers. Like 22 years old, 24 year old. Attract to people who can like code whatever. But they are not researchers. And what prevents that is just like these big labs, bro. Like if some like the comp is like somebody will give 5m or 10m or something and they have good colleagues. Like it's already like this problem. Yeah, so
BI see.
ASo it's like. Yeah, it's. Yeah. So Americo has a peer group and what I. What I enjoy the most is working with people in Germany. They also have other responsibilities. They have a bigger team there but you know like responsibilities like by SF group, company directions, strategy. But like even like company building and founder responsibility feels like peripheral increasingly because like the density in Germany is increasing and like the ball is rolling very quickly and then reset wise. Also I feel peripheral like a little bit because like I can, I can do a cute thing. I can do a cute like I. I won't be. I'm increasingly finding it difficult to work on the biggest model that BFL will train. And BFL anyways is like not training big models. Like, like BFL itself is. Has become like kind of tier two in terms of frontier. Because resources.
BYeah, but like I mean, yeah like scaling hyperscalers basically.
ABut then it's like. It's like if you think about share shareholder value, like I'm very confident in the team punching way about its weight. So we are only a 3B company. And then like the kind of models we are still training is insane for three big companies. So in that sense, it's like we can like increase value more. Like, I don't know, like it can be like a 10B company, 20B company maybe in like two, three years from now. And that is very good. But like for me it's like, yeah, maybe I can stay here for like to like vest more equity or I don't know, some from that post. But that also feels like not. Not great. But anyway.
BSo what's on the top of your mind? Like what are your possible alternatives?
ANo, this is in like it's. I'm just. I just want perspective whether like what I'm going. Like this is even making.
BNo, I mean yeah, but I feel like you can get that. I don't know how much. But can't you get that like feeling of like spitballing ideas with people outside of work? Like it doesn't work like that. Like you find a group of people
AI want to discuss this with. Yeah, they're all in class and then like it's just not the same. Like they. It's very secretive so they like don't say anything and they don't want to hear anything. We talk about general. We gossip about like we never get into like technical details. People I actually want to discuss with. And then like there are other people who work at other startups and stuff but they are themselves there in like, like around. And they like they don't have a clue themselves.
BYeah.
ASo there's a Mistral collaboration. It's like yeah, I. I don't know. Yeah.
BAnd what about like. I don't know, bro. Like I like from my perspective, you're already sitting at such a high leverage position to do basically anything you want.
ACrazy. Yeah, it's like high leverage to do
Bwhatever I want with so much capital. Like it's unimaginable capital to me at least right now to so like.
AAnd that seems like I hold like a big piece of a. Yeah.
BOf a big company, a frontier company. And you have freedom like you said, to do whatever you want in sf. You don't have to hire like maybe hire somewhere something or like hire someone. Right. Like who can discuss with you high level ideas. It doesn't have to be the best person.
ARight.
BOr like and research Abraham.
ALike it's like becoming like uphill battle. Like just a Germany may look better. Like we are. We are not properly geo normalizing our salaries. It's like oh, it's like they'll be like oh, 800k per year is so much money. Like it's so much Money in Germany, bro, nobody earns.
BYeah.
AThen, then we are like paying like, like something like 250, 300k base and like 500, 600k equity. Like this is like was the AI salaries three years ago. And this at least for like people who can like, yeah, like who can like do. Who can change the needle, move the needle kind of thing though.
BYeah.
ASo recently we lost a new grad, bro, we lost a PhD new grad. She just graduated from CMU and we gave her like something like 700, 800k and she had like so every of every offer was more than that. And then we were like, oh, we have to break our internal consistency significantly because like there are people in Germany who are more capable than she is who earn less. So like it doesn't make. They are like it doesn't make sense. And I was like bro, geo normalization then though, SFs are constantly feeling uphill battle. Not like for me, for my career trajectory. Like my learning has like feels like stagnated. Like it's like next two years. Yeah, like. Or like I don't know if it's like just to like, I don't know. Yeah, Like, like if I do the math on how much I make and like and if I think of it as salary vesting over two years, it's
Byeah,
Alike I. I won't get this ever in life. Like, yeah, like if, if I'm starting a company in person, there'll be like other co founders and everybody will be like, there'll be pressure. Right. Like there is literally no. There's nobody to tell. Like I am the like.
BYeah, yeah, makes sense. I mean ghosting. I mean, I don't know, bro.
AI feel like bro. Yeah, by two years, my job, like job.
BI don't know, bro. That's why I'm like torn about stuff. Like maybe coasting is okay because like I don't know. Like I said like I feel like labor value zero jaari like bhat increasingly concentrate actually valuable and like they will keep getting paid more and more and more and like you see the concentration happening right in front of you.
ASo. Yeah. Does that mean like I feel like that will be true and there'll be like NBA player kind of salaries given to like a very few people.
BYeah.
AI don't want to like coast and like deskill myself. Right. Like I do want to be in the room where like the best AGI models are being trained and I can be like, I can contribute to that.
BYeah.
ASo it's like a company and Then bro, German boys, they have increasingly become more passionate for the company building and the power and like like the feeling of like this like power. They increasingly value that more than doing frontier.
BThe technology.
AThe technology. It's like like like I'm seeing. Oh we are not able to do like a very state of the art video model. I was like maybe merging with another company. Like oh yeah, acquisition. Yeah. I was kind of thinking that will give us resources. We are very smart. Then we can continue to be at the technology forefront. But they increasingly, they're okay with doing something that is not forefront but still runs like be run in the driver's seat of a three billion dollar company.
BYeah, like Mistral.
ABasically like Mistral and. And bro, this like power is like a little bit intoxicating. Once you feel like you're the CEO or like co founder of like 80 people will listen to whatever you say and like stuff. Right? Yeah, I feel like those are like not the right motivations.
BWhat is like what is the situation? How many co founders do you have like how much of a say do you have in what happens in the company? Like, like obviously like they say acquisition. Karwani, that won't be just something you can call. Right.
ASix. Six people. Six. We are six people. Regularly discuss Karthiya and founding team is bigger than that. But like I have very little like I can persuade and I can like really like guys. Like I can make strong arguments and I can make them see why we should do it. But like again I'm one in six. Like it's like if there is like two or three people is like we are in Germany. We are the German company. Let's go. If they are in that mindset then like Sumit's persuasion which is not going to work.
BYeah, I see.
ASo and then yeah I think yeah coasting but like company zero
Bprobability distribution. Two years.
APeople are not dumb. But you can never say like it might be out of our control. Like we can say that oh we'll sell when we really need to sell but it might not work that way. But we have like many smart people. Like, But like
Byeah, I mean what you're describing. But yeah, like don't you give me your perspective on this. Like thought in the next five years there will be a smaller and smaller group of people who will be leading this technology and they will be getting more and more salary and job automate to make pyramid. And you're automating lower levels with AI or tools that you yourself are building and capturing that value for yourself. Right in the limit. Like why doesn't that gets shrink like gets shrunken to 100 people and then maybe five years later to 10 people?
ABecause government will change the structure of how this technology is developed and diffused and stuff like that. But Kafi Hoga and then like there is no reason you cannot be in that hundred people. You can be.
BYeah, I mean but I'm just. I just think if you can be in the 100 people, like if it comes down to like 100 people, there's nothing stopping it to come down to 10 people also. Right?
AYeah. Then I just like in the, in the wave. You just ride it for a little bit longer, bro.
BYeah.
A100 to 10. I would have made like bank so but I don't want to cut off so early. Like I have been in the industry for. I have only three years of work experience in like, you know, like I have not made liquid like yeah, I've not made liquid money.
BBut what will you do with the capital? Like what is your ultimate like desire to do with that money?
AOf course, but like it's not solely capital. The solely capital. Then I can actually like coasting here and then persuading for acquisition or just like making the right move is like probably capital to effort ratio. My best like going anywhere else. Effort is going to be 10 times more and capital might be equal in best case. But. Yeah,
BSure.
AYeah, I'll have maybe more updates.
BYeah, yeah, sure.
ABye. Bye.
75:53·2 SPEAKERS·8216 WORDS
ABro bhatia bro. Crazy bro. Complicated. Bro.
BStory of everyone's life.
ALike expected value of the trajectory is very much like misdrop monetarily on paper you can have like decent amount of money but the the ceiling is capped. It won't be like open air anthropic or like like even like the researcher quality and research thoughts and crazy bro.
BSpeaking of idea like myth. So they've. Is it because of scaling or some new architecture?
AYeah, it's like I think apitak. It's like maybe like two, three times bigger size models and it's like it's like the. The best. Even Opus 4.6 was trained with Hopper GPUs. Like big massive clusters of like Hopper GPUs.
BNow Blackwells are coming online.
ABlackwells have been online since like like we had Blackwells last year this time. So this is the first output like big run output from Blackwells. So it's like it came online and then large enough cluster trained.
BI mean clusters fully live. New 100k cluster Blackwell or Vera Rubin
Ais already in production and it's crazy.
BYeah.
ASo next two years. Like next concrete. The basic coding is gone. Bro.
BBro. And I feel like flywheel create. Now these models are good enough to bootstrap themselves like. Yeah, Yeah. Scaling. Insane. Hello.
ABut bro, it almost feels like I'm talking to you in person. Right? But how coding crazy. Like I'm just thinking how should I think? Like for me the Dylan I felt, I feel like I was at the right place at the right time. Like the stable diffusion guys, they had already created stable diffusion when I did research. But I proved my value that they gave me this counter equivalent role. And like I'm basically writing the value. Like the value I'm getting for my liquid capital is basically stable diffusions value. So I was right at the right time. And then like then I obviously I did the work right. Like I got the opportunity and I did the work for it. Yeah, but like yeah. And it's also like they like me. I am relaxed. Like I said, I've never been this relaxed at work. And like working very like self driven that like forced. But it's. Yeah, it's like do I want to do four years of like not like 20 years of this maybe? Definitely not 20 years. Now it's a question do I want to do four years to west like the insanity I have. But then that point you're already kind of in it for like the wrong reasons.
BYeah, yeah. I mean like capital like you know, in a world where like, like, I think you believe that too, right? Capital. Like, I mean, there's arguments key.
AEven after starting a company, growing it, I will not be able to afford to buy a house. Isn't that insane?
BThat's insane. Yeah.
ASo. And then maybe like governments and rich people will get involved. Like, even though the technology is marvelous, there will be like bureaucratic challenges and like sovereign challenges and stuff like that. So the diffusion of technology will slow down because of that. And then maybe having access to sexy AI is what will. Like, like you can like show more output, right? Like, yeah. Looking at anthropic, like your output. What will your output be for six months? We can work with like, newer models. Like, you can be an incompetent person and have like a lot of competence.
BYeah, yeah, yeah. I mean, I think at the end of the world, like, compute is all that matters. Like, your money will basically be invested in compute, right?
ALike, you will, like, basically, yeah.
BLike, compute will be the next currency. Basically.
AIt's like utility, right? Like energy.
BYeah.
AAnd robotics will be a big thing and like, people will. I think there'll be a lot of work on hardware and energy and like materials and space even. Yeah, Fields. Fields will look different.
BYeah.
AJust like software engineering was the hot dominant thing to work on in the last 20 years. Yeah, it will probably change.
BYeah. I mean, I think 18G. Say like, like you said robotics. Yeah, Chips, robotics and space. Yeah. I didn't send, you know, like, Bostrom writes about this, only he calls it the cosmic endowment. Like how. How we're still like such a small percentage of all humans that will ever exist and how like, small changes of what we can do in the world will affect so many trillions and trillions of humans that we need to make the right choices right now.
AYeah, bro. Somebody told me, like, one of my colleagues is like, bro, if like clusters are in space, it's literally like gods that harness energy from the sun and we pray to them and they give us our prayers. Like, I want it like you from sand. You'll create something. You send it to space. It's the harsher energy from the sun. And then you'll ask, make me my JavaScript web app. But yeah, like I said, I don't know. Like, it does seem like for me, like, feels like almost near perfect. I don't want to it up. It's like, it's like almost like near perfect. It's like leverage.
BI mean, like, if you do believe future power of like capital and you will be able to use it effectively, Then wait, no for two years. Like then according to you, two years may things aren't changing. Like there's no ultimatum. Then once you've accumulated diversification a little
Abit even in capital, right? Like it's like bought insane salary. Like I said like 5 to 10 million. Like if that, like if his AI companies even like 3x or 4x.
BYeah, they will. I think they will like anthropic definitely.
ASo it's, it's, it's like riding on and also, it's also about like if I make that switch that also kind of guarantees me like that I'm in this upper like. Yeah, I'm just worried what is the gap guidance. And then I am not able to like today like I'm still like valued as AI researcher because they know like my history. Yes, they are doing a lot of cool shit. But gap is only like two years. But he's a frontier researcher. But four years from now they'll be like, oh, he's a founder, manager, something, something like like in like short term gain. Like if I'm, if I actually I'm on the, the critical trajectory of AI, then go long term games, you know, like I'm valuable for like I'll be in the group of like I don't know, it's like short term discomfort for like career. I'm, I'm. We are so young, bro. Like next 40 years. Career.
BYeah.
AOr slow
BI guess. Yeah, no, that makes sense. I mean I, I don't know. It also depends on how much like we're talking about like the depreciation of your value, perceived value set versus the like.
AAnd also like one more thing I, I want to add is like when I was at Stanford or something like I felt like I was in person. I made so many connections that I could like, like I talked to them, they are all interesting. And then it also came in handy when we were starting the company like getting customers and getting some advice and stuff. But increasingly I feel like I am like isolated from the. Even though I was like the best city to do this, I'm like I talked to some Lando startup founders and like some people who don't understand. I'm like isolated from like the actual frontier. Yeah, it's like. But I was with them. I was with these guys. They were all like we were all together. They're all doing crazy.
BIf you can get into anthropic, may
AI say anthropic, there are 8M an hypothetical and then like in NBFL it's like obviously Much, much more. Right. Because we started the company, I own like a piece but like but it's fully liquid. Could go to zero expected value.
BYeah, I mean distribution.
ABut anthropica to grow 300 tier and then they are saying that in by 2028 or 2029 there'll be 2T so it's like 7X.
BYeah, yeah, Yeah. I mean I don't know bro. Like value model companies. I don't know. Like we discussed last time, IP dissociates right?
AAnd then like these model companies will be the future banks. Like the, the, the the 6,7 trillion like now it is trillion dollar companies. In the future of course it'll be 10 trillion dollar companies and there'll be like 6, 710 trillion dollar companies and there'll be like a big gap with the rest of the crowd and it will be anthropic. OpenAI, SpaceX, CDM. Maybe Nvidia. Like yeah, it'll be these companies right. So value and obviously there'll be like lot of other players who help with the diffusion of technology. Maybe with the robots, maybe with.
BNo, but I don't understand why Value Oga like if you're saying AI model research is getting automated then they are
Acreating the gods they have. They have positioned themselves to have the massive compute required to dream gods.
BBut they are not the ones that will continue to hold more compute than the rest of the world. Right?
ABut why would they not? They are like Sam Altman has committed to buying trillion dollar compute for 2028.
BBut I mean like at a point where people realize these gods are we can create these gods and they are, they can recursively self improve and manage themselves. Nvidia will stop selling GPUs right? Like why will they give the value away bro?
ANvidia will not be able to like why will they give it away? I mean like because Anthropic will not make their own chips. Like I don't know at what point for Anthropic it's better to make their own chips. Anthropic will still like supply chain but
BI feel like value will accumulate to the chip manufacturers. It will not reside at the model companies.
AI mean.
BHuh. Bottlenecks exist at different like levels of the supply chain.
AWhat value will accrue? Of course if there is a $10 trillion Nvidia will still be like a $7 trillion company. Like I'm not saying like but Anthropic will be more valued than Nvidia. It's hard to say. Maybe Nvidia is More valued than ANT public. Because Nvidia sells to both OpenAI and Ant Public.
BNo, I mean I'm trying to understand like what makes a company valuable is that it's only it is able to do what it's doing. Right? Like there is no competition. That's why if, if there is competition, it becomes a commodity. If it's perfect competition, it's a commodity. Right. And we are saying anthropic holds value because it's able to do something that other people with compute aren't able to do. But we are also saying.
AYeah, it's like recursively self improved.
BYeah, yeah, I also agree like, because if you're talking about literal rsi, so even a three month gap is enough to like always be in the lead. Right? Like anthropic will like, but I don't know if we are in that. Like we, Anthropic will not be dumb.
ALike anthropic will not give away a model that will help them like will decrease their internal development value.
BNo, but like okay, think about this. IP diffuse. You have 6 months Puranawala algorithm and, but you have your compute efficiency and everything included. You are 10x anthropic compute. Let's say China does this or like some other company.
AYeah, but like no, no, nobody else in us can. Anthropic. GDM, SpaceX and OpenAI have already planned it out for the next four years.
BBut anthropic compute compared to, compared to OpenAI?
AYes.
BAnd compared to like Google also.
AI mean they are, it's still insane. The exponential is probably a little bit lower because they are thinking that there'll be more friction from the society and like governments and all of that for the diffusion. Like that's their bet they're taking.
BYeah, but
Athat's how these big labs have realized that it's literally like accumulating capital. Compute is like very critical and like, and they saw this like sooner than we did like of course, like, because then like Sam was still making these insane deals.
BYeah, Yeah.
AIt's like if you have like AI itself picking up experiments.
BYeah, yeah, I'm not, Yeah, I also feel like it makes sense but I'm just like will it accrue in the hands of people who just who have like a three months ahead algorithm or will it accrue? I guess. Yeah, you're, you're also saying compute.
ABut like will anthropic be 10t or not? And I, I, I would take the bet. The anthropic will be 10t maybe in like four years, five years. But today, if you are deciding to join Anthropic, there is still like a 10, 10x upside 100%, bro.
BLike the world doesn't know about. Like, okay, they know about AI, but no one's heard of Anthropic for.
ALike insane, bro.
BYeah, I mean like, yeah, I mean, no brainer. Like it doesn't matter,
ABut then bro.
BAnd right.
ALike intense. Like, Like all the hard work of getting it off the ground is done. Now is the time to like capitalize, not do it.
BBut then what will happen to your equity? You'll be able to leave with your equity, right?
ANo. Yeah.
BWhere is like, what are you. So what are you losing out on by leaving?
AI'll have like whatever stake in my companies. I'll have 50% of that.
BOh, okay. Okay. Oh, it's a four year west, Bro. 50%. Heb is fine, bro. 50%. I think.
AI don't know if other founders decide to fuck me over or something. Like, I think there are ways to engineer the cap table, in which case I'll get like less payout if I'm not there at the time of exit. But like, it's very tough. Maybe they won't do it. Like, they'll be like, sunil deserves this, so it's fine. Or maybe they're like, they can buck me over or they can also. My expected outcome is they will run this company for like two, three years and it'll be like a mistar. Like kind of like zoom. And then like, like maybe 5B next year this time. Maybe it'll be 8B two years from now.
BYeah, Like, I think you should consider
Amoving and in person, bro. Like in person, network work and like,
Blike effects are invaluable, I think compared to just capital. Like, you'll be also building so much social capital.
AYeah, social capital is amazing.
BYeah. I mean working with top people where. Who are building the God. You'll be like the priests, bro. The. The people who are channeling the God to the common. Same. Same. Yeah, yeah. It was unimaginable two years ago also, bro. How dependent we've become.
AI was working on fourth generation, bro. 2018 Maya. Why I was working in fourth generation at Stanford, which at the time was Frontier, because Frontier Labs were not like Frontier, right? Like at Stanford, people will be like, like I used to tell G, bro, like, what is this boy who used for? Like, eventually I was like, I imagine code forces, problem description. They're like, that's not happening for like 20 years.
BInsane. Yeah.
ALike, how out of that, like, cutting edge folks can be. I was also like, I didn't know what would happen or whatever, but I was like, that is like a problem statement, right?
BYeah. I'm like, fully convinced, like, the universe wants to learn itself. Like, it's just scaling will get you anywhere. Inherent learning, like, any kind of like, repetitive algorithm will want to learn itself in a way. And
Awe are so curious about, like, learning.
BNo more like. I don't know what I'm trying to say, actually, but I feel like the universe is just full of representations and apnea modalities, specific representations of the universe. But you can learn these modalities. You can learn other modalities of the world as well. And the universe is just filled with these patterns. And you can just scale any algorithm and it will learn it. It'll be inefficient, but it will just learn those patterns of like, predicting the world.
AAm I insane or like, am I not? Like, I don't know.
BNo. Malava. I think it makes sense, bro. Karo bhai, frontier. I mean, how feasible do you think it is to get a job at one of the.
AI don't know, something like that. And like crazy interesting models. Are reviewing, agents are coding, And then like, definitely multiply.
BYeah, I think it's worth. Yeah, definitely worth it. Like, I can also see it 10 xing aram say as it diffuses into the world.
AAnd that is less important. I mean, it's like enough money that I will. I have to not worry about having less money than my peers. What do you want to de risk? Yeah, and I can buy a house in sfk, bro.
BYou'll be competing with all the other anthropic employees, bro. Yeah, memes. It's like open AI employees. Like 15 million anthropic employees. Like 20 million. And then, and then the SF homeowners are like, boomers are like, oh, glad our hard work is paying off. All about the value that we've created, Bro. This Anjni guy is so well connected, bro. Stanford is holding a lecture and
Athey
Bcall her the coach. They're calling bro. What the. And like, Stanford students are so privileged. Bro. What the.
ABut he understands frontier. So, like, I. I think he'd understand like, like, if I tell him, like, like, if I tell him, bro. Like, this is like, not a joke. Like, this is.
BYeah, I mean, clearly, huh. He'll get like, you're coming out of a place of, like, curiosity mostly. Right? And like, wanting to work with the best.
AYeah. And also he knows that I'm Remote. Like, it's not like. Like, I am actually doing something. I'm very critical to BFL or something, right?
BYeah,
Ayeah. And this crazy, connected bro, he makes the. Random SM folks. And there was, like, one guy, he's like, oh, blah, blah, blah. And then, like, this guy, he's like, oh, BG is like Mr. In Sesame. We keep chatting in some content and stuff. And I speak to this, like, some BG guy, Brandon. I speak to him for, like, 30 minutes, 45 minutes. He's, like, really picking my brain. What do you think about AI? This and that. Very, very nice. Convers.
BYeah.
AExchange numbers. Anyway, then I go back and I Google that guy. He was Mo Watson's ex, bro.
BInteresting.
AFounder of.
BWasn't the founder of Ocular. Okay.
AAnd then he had invited Brendan to just talk to everybody and, like, hang out and stuff. And Hamara lawyer was like, oh, Brendan had bought, like, a sample sesame glass. He wears it. It's very cool. Hamara lawyer was like, oh, this is so good. This is so refreshing. That Oculus. Oculus was always so bad. He doesn't know this guy, right? Like, he's saying Oculus has this problem.
BAmazing. But what is the Oculus founder doing right now?
ALike, Brendan, right? Like ssmx. AI, do you know sesame?
BYeah, I know sesame. And wait, I thought you were talking about two different people. One guy who had the sesame specs and Brendan.
AOkay. Yeah, Brendan had. Brendan was wearing sesame specs. And the lawyer.
BOh, lawyer was bitching about looking at
Athe class, and he was like, oh, this is very nice.
BOkay, okay, okay.
AEverything that, Bro.
BYeah, bro.
ALike, some fear factors is like. Like, I'll be asked to work insanely, like, ass and hard, and I've forgotten how to do that or something.
BI don't.
AKnow. That is one fear. And then there is so much luck that I might get fired in six months. Like,
Bbut maybe problem, like, 50% equity. Like, it's not like, bro, like, 50%. You're still in this company, right? I think it's. It's a decent price to pay to, like, have a. Or upward trajectory.
ABut, bro, like, my. My slope. Like, when the company started, my slope was this. And then now the type of.
BYeah.
AAI scope is this. And, like, my scope as this. Yeah, I can, like, switch.
BYeah. Pay 50 premium and switch. I think call option, basically. Yeah. Yeah. I think just buy, bro. Like, pay the Premium. Buy the IoT.
AIf you also think about it, if you think about your, like, yearly as, like, money that you're paying to buy stocks and like I have paid two years to buy like BFL stocks.
BYeah.
ALike I mean I'm really not talking about emotional and two years I have like paid to buy BFL stuff. So next two years do I want to buy more BFL stock or do I want to buy antic.
BYeah exactly. Makes sense. I mean that thinking makes sense. Yeah.
ASo like it's my what. What do I buy with my time, right?
BYeah.
AAnd then it's also better for me to say like what would give me more capital, more like more options to buy things 20 years from now. Like I have more options to buy like Frontier Lab with my capability than like bfl.
BSo what will you do Abi? To like like you'll start pinging your colleagues Confidential.
ABut I will like chat with my Stanford Bros. Bro. That's the plan.
BSo are they all working in the same team or Anthropic. Anthropic is the only company that doesn't respond. I even talked to like they had a meeting for like quants and stuff. Like I talked to that one guy but basically he was like haha. KAR apply. Kar officially apply. Kara you Like I'm. I can't like he basically indicated I can't refer you. I will apply in anthropic again to be.
AYeah anthropic. And do you want introduction to other startups? Like I don't know like
Bbro, I just joined this job.
AI don't know what?
BYeah, I mean clearly I don't think they're hiring right now and obviously I think yeah it's like a small knit thing. They will only probably get like closer people for now. I think that's what it sounded like.
AYeah. So bro. But that equity would be crazy. Bro.
BBro, I didn't know this guy is so well connected.
AWhen he has 30k twitter followers now. He left 3 million easy.
BYeah. Guys insanely well connected and both must investments. Yeah I saw, I saw his roster.
AInsane open router. Black forest lab like last two years.
BYeah 2021 anthropic I think
A500 million five.
BOh yeah.
A500 billion
B300 minus.
AI have only like we are so used to thinking of money as like linear. Like yeah you know, multiply. Investment perogue please. Maybe, Maybe yeah.
BInvestment. Yeah. Like private investment.
ACrazy competitive area. Private like if somebody's starting a company.
BYeah.
APeople. Yeah. There is no room.
BYeah yeah. Rishabh. 21x wow. Yeah but now I'm like very actively any kind of private investment. But yeah I think private investments are.
ABut how much of that will actually be liquid. Especially if a market downturn happens.
BYeah, true, true. But like, like at small prices it's fine. Like small percentages like really good people is very safe. Yeah, exactly. But yeah, good bro. Yeah. Yeah. I mean I think that, I think you should make sure 50% like the founders won't you over completely and downside. But I don't even think.
AYeah from going from startup to this is like won't be insane. I think people are used to discomfort and doing big city things.
BYeah. And I think it's an obvious choice. Like do you want to buy more BFL stock versus anthropic stock is an obvious choice.
AAnd like future May, what would you like slow versus intercept. Like there is a intercept that you have. Like you have, you have the bags for the intercept but like no slope.
BYeah.
AAlways optimize for the slope. Like slope can be crazy.
BYeah,
Ayeah.
BWhat is Maggie up to bro? Maggie catch a rest slope intercept.
APeriodic labs is going to be 7B bro. I think they already. I don't know if they've closed the round.
BBut are they doing anything? Like what are they working on right now?
ADo you know like robots that is automated like science lab. Like robots that are doing science experiments so they'll like try to be frontier in like physical science. Like actual science. Yeah, science stuff. I don't know.
BEven anthropic bought some kind of this kind of startup only you know Abby for 400 million.
AYeah. Coefficient labs. Yeah, yeah bro. I say accelerate investing like in the case of acquisition. I don't know if you know like I. I don't know like I wouldn't know. Yeah, it's like. Pulled back and like gets paid out and then like fresh like retention process and all of that start. So yeah if like BFL gets acquired like 6 months after in case I decide to leave and 6 months after I leave then like I would have, I would do the next if there was easy money on the table of the next one.
BBut heck yeah trajectory May next six months.
AHey bro, wait. Like the boys, like the German boys, they want to very much like be independent and like even if they feel like if it grows like 3b this year to 5b next year to 8 weeks the year after that, they're very happy.
BYeah, Yeah, Yeah, yeah. That makes sense. That makes your decision easier, right?
AIt's making sense, right? I'm not thinking out like grass is green.
BNo, I don't think so. Like especially because if it was like completely binary BFL Bhai Bhai. Everything. I think 50% is a decent stake already. You have like accumulated enough like yeah bro, just yeah, I think that it makes sense and you should like anyway get a feeler for like
Aand then also like this like one thing is like when we started the company in writing there was there like co founder, co founder. Co founder. J. Logan. I'm co founder. Co founder. And as a company grew and the narratives shifted and stuff like I am not basically public person. Like nobody. Everybody thinks oh Robin is the founder and like as a party. Sumit is a co founder of bfl. I said use as a narrative. It's not in the public narrative. It's there in the internal company narrative and I have the power internally and all of that. So I'm basically grinding. I anyways don't hold as much. Like it's not equal splits. Like I hold very high but like still less than them and I'm not in the narrative as well. So it's like yeah, like the sub opportunities sacrifice situation. Like if it is literally you're the founder, you're in the board, I'm not in the board. It becomes like more yeah serious. So like that is also another thing where I feel like I want to start my own company. Like
Bso it'll be like relatively easier to exit also. No, it won't be a big like public facing deal. Right.
AHopefully it will not be a public facing thing but like internally bought by SF Office.
BWho will be the next in charge? Oh
Ano.
BIn charge.
AI don't know if I'm not there.
BSo they'll send one of the. One of the German people come for some time I guess or no,
AI don't know. I don't see them doing it. But like let's see. I don't know, they might have to hire somebody. I don't know how to hire like head of North America or something. I don't know. People do it. Right? Right.
BYeah. What happens to unvested equity? It gets redistributed.
ANo redistributed name goes back to employee offset. Like employee pool. I guess like that like that can be like. Or maybe like I mean if the founders really want they can split it among the founders to vest more as retention for.
BBut is it not predetermined? Like who decides like founder.
AOh sorry employees. It goes back to employee option pool. I thought founder equity will also go back to employee option pool because the pool is depreciating. Right? Like actually hiring will be exceptionally good for the option pool aspect because you like lot of it will come back. But it's crazy, bro. Like I own like so many shares and now when I see offers for new hires, I feel like this bro is sacrificing so much and like working for four years and they are getting such a small percentage to what I own.
BYeah. Founder Equity.
A0.1%, 0.1% or like buy multiple like double digit like capacity. This is like.
BCute. Like. Yeah. How old is she now?
AFive years.
BDamn. Which personality change?
APersonality. I feel like she's like after living with us, getting less and less disciplined. Like all the good when she came, came home from like she had, she had so many good habits and behaviors. It's not like rapidly decreasing. But over the years there was like one good thing that she had like in the previous years that she doesn't
Bhave like
Akitchen man. Spending any time in the kitchen. I think wherever she was like there was like a strict, like she knew it. Like kitchen is a place where I don't go like, I don't climb on the kitchen counter. I don't like go anywhere near the kitchen, smell anything. Now it's like has decreased. Like he goes to the kitchen cabinets, opens it.
BBut how did you initially enforce kitchen? Did you have barricade or something?
ANo, no, we didn't have anything. She was like one year. One year she like was in the category, right? Like she got her one year old. Like when she was one year old. Yeah, one year. You already had all this. She was programmed, she was trained. We didn't have to teach her anything.
BWho trained her?
AThe cat people like the cat.
BOh, okay, okay. And they train specifically kitchen.
AIt's like they have like 15, like 20, 15, 20 cats with them at any point. And they like strongly enforce this rules in the house, bro. Like they see other cats, they just don't do what other cats like, they just enforce. Yeah. So that. And then she really understands like human faces and hands and like humans. She doesn't like claw or bite even if she uncomfortable. She would like never claw or bite or anything. But recently. Like she'll come ask for pets. She'll rub her head, she'll rub her body with you and then you like scatter. Status. Scatter. Then she gets stimulated and then she can get over stimulated. Loses her head.
BYeah, bro.
AShe's like, she's been scratching me. Like she's like, bro, She's like trying to take my hand to her more and like that is a new behavior where like she was never so aggressively asking for something where she would use that Power. Oh yeah, yeah, yeah.
BThe concept of like animal super stimulus was new to me. Like, was like animals don't have this like they're not in regular evolution over like the last few thousand years. They have not had this like human interaction. Right? Like or like any kind of human touching them or grooming them or something. And that is like something that their body is not adapted to. And it's like it can be a super stimulus. Like something that's completely out of distribution for them.
AYeah, it is. It could be out of distribution, but
Bit's not the same thing. Right. Like they can't get all the body parts. They can't apply the same pressure and stuff. So that's what I've heard. Like some things, especially for some animals, it literally acts like super stimulus. They're not used to it and they go crazy because of that. But yeah, what you're describing is different.
ALike gradual super stimulus, bro. Yeah, over stimulation is a thing for sure. She like becomes her eyes. Like usually it's like some bit right. Like she can control her bit. Right. And then like when she's super stimulated, it's like fully white, like fully black. Like.
BDamn.
AAnd when she's most relaxed, it's like very thin and like. Yeah, bro, it's like it's. It's like pull back. Yeah, bro. I just wanted like a temperature check from like external like perspective, like you know me. And also like can evaluated the distance.
BYeah, I get some more opinions obviously, but make. And yeah, I mean also get a vibe check at like these frontier labs. That will also help make your decision easier. Yeah, probably easier.
AProbably easier. Let's see. Yeah,
BI wish I had more people to discuss AI stuff with, bro. Not interested, bro. Like all the like, all my finance friends that I made in mfe, I do hang out with them sometimes. They're. They're not getting it. They're. They're like bubble every scam. I don't think so, bro. Like, I don't know. It's either my friend. I don't think it's mispriced. Like Torah. Like memory stocks, for example, are I still feel are underpriced. Key they're, they're priced for like being cyclical in nature, right? So right now they're trading at very low price per earnings. Earnings have been spectacular. But they're still priced very low compared to that because people expect demand will drop because now supply will come in and we won't need it. But Vera Ruben has what, 192 GB of RAM per Chip right. Memory requirement is not going down, bro. It's only going to scale. Yeah but industry.
ALike startup. I know you guys are very set in like New York and stuff like as a startup.
BBut I think like, I think like you said but if random startup won't make sense right?
ALike if like it's like something that like say openaika senior or executive like the core automation. Like something like that like where, where you know this like the leadership is good and like the worst case they'll get acquired bro.
BBecause like yeah. If, you know, like if you would recommend certain startup. I think I'm open to relocation at this point. Like I, I, I would quit for the right startup. Honestly. She, yeah she's also remote. She's also looking for a new job. But yeah skill set. I don't think she'll, she also doesn't like she just wants to. Oh yeah. I think yeah. Relocation where she's fine. Yeah. He'll be okay with it. Yeah. You think that you would like think would be a good move to move. Like what about. Have you heard of Listen Labs.
APortfolio? Portfolio.
BYeah.
AFair Open router is racing at 1.5 gm early 2024.
BI was like are you looking for more engineers on your team? They're like no. They didn't reply.
AThey're very small. I know that. Should I. I know people at open router. That's the way. So I can refer you now
Blike open order though has been on my radar also for a while. Yeah, yeah.
ABut it's, it's fine. It's like you will get the skills
Bto go do something some other place. Yeah.
ASenior full stack engineer. Remote. Engineering.
BYeah. Yeah. Senior full stack engineer looks good. Let me see. Let me open my DMs.
AMy bro has been a big fan of open routers.
BYeah. September 2024. Not early 2024.
A2024. Yeah. Two years ago like September 2024.
BSo in June 2024 man, I was like are you guys working on a front end? They didn't have a front end. They were just like a routing company. Now they have a playground and stuff.
AYeah. Yeah. So let me, let me, let me ask.
BInteresting
Afigure. Other cutting edge option would be working at a robotics company. But now is the time to get it.
BI agree. Like I think software pivot mark. I need to develop either robotics chips. Yes. Space and these, these three are the things maybe chips is also pretty saturated.
ASo it might be a good opportunity to get him.
BDuring our times for finalist. He works well. I know him. So he referred me and could, could interview but I got scared of the lifestyle they're leading, bro.
AYeah, it's like chaos. Like it's like easy to get in because of that. Right?
BYeah, that's why that's not worth it. Yeah, that is not worth it.
AI'm not joking.
BYeah.
ABut the salary, you know, right. Salary will be like.
BIt'd be a good entry point also in regardless of if I stick around or not.
ARight. It's a good life. It's like baseline credibility. But the next day I say yeah, I don't know like maybe you can go to OpenAI. Anthropic will probably like not or immediately.
BYeah but you, you keep radar interesting startup. I think that'll be, that'll be nice. But yeah,
ACrazy application of AI to
BI think like, like quant may apply. I think it's better to direct like directly work on AI research or something. Yeah. Quant industry feels like. Like middleman cut out. It's just capital creating more capital or like
Ado you think that will survive? Like it's like some AI AGI company will like figure like capital market efficient. Banana would also be subsumed in like AGI company 100%.
BYeah, 100%. Like unless they come up with like. Unless they are like three months ahead of whatever company in that specific niche. Right. Like they might get hired or something like that. Yeah, I think like some kind of collaboration will happen. But there was like rumors, I don't know, you might know if it's true or not. But Anthropic was using trading signals as mid frequent like mid horizon, mid time horizon. RL signals. Like I thought people are already doing this at big labs.
ADoes seem like something who might be interested in finance. Like it's like they're going after high leverage. They're doing few things but like, like for example engineering. Right. Like one area but like high like people will pay top dollars for it.
BRight.
ASo finance seems like finance and like medical and these things seem like like like other industries where.
BYeah, like I mean, yeah, like capital allocation is the right problem. You have all the information better than like hard, hard, hard numerical data. But also like whatever sentiment, news, what people are thinking, demand patterns, consumer patterns.
AYeah bro, you have such good thought.
BI do feel bro, I have a system also. I've recorded all my thoughts. They're all categorized. I have a website where like all of this is not like mapped out. I'm mostly I think interested in like the philosophy of things more than like the. I think that this is actually Just
Atalking to somebody, you know, like. Yeah, just like talk to somebody and like convince them that you have good thoughts.
BYeah, I think I just need to like the thing is I don't think I have original thoughts. I just think I've read a lot of many people's thoughts and some of them have resonated and stuck around with me. So I also feel like original can contribute. I'm just like, I've just read a bunch, that's all.
AInteresting startup might be a role. Like obviously if I were you I would apply to that. Like literally like OpenAI anthropic, XAI JDM Navy meta. Like that's it. Like and then I don't think there is any serious contenders at the top level. And you I would exhaustively apply to all the interesting roles, all the recruiters, all the hiring managers.
BBut the thing is President, I don't know if there's precedent for this but dude, like Google, DeepMind etc have like weird rules bro. Like AI policy. Like and Anthropic had some demolitions expert type of thing also like Joe Explosion, explosives or demolitions. There was a literal page about this. Wait, let me find it. Explosions expert anthropic.
AOh, it's maybe for like testing the model.
BYeah, it was a safety thing. Yeah. BBC AI firm March 2026 AI firm anthropic seems seeks weapon expert to stop users from misuse.
ACrazy. And then the other other next skill I would go is either like top robotics companies like physical intelligence Sunday and then or go to like executives from these like top labs who left and who are starting. Starting very early when there is a chance you can fetch that they are good so you learn a lot and they might also be acquired. So yeah, you get entry.
BBut what are these physical lab people doing like. Like will it be like different from normal software engineering and like ML research? Right. As in the robot labs.
ASoftware of AI intelligence has solved. Right. Like the robotic labs are basically figuring out how to implement or like get a feedback loop going for actually improving. Like, like robot like different, different triggers. Look, some people are targeting household, some people are targeting still like enterprise factory automation. Like literally just like perfect. Like high, high randomness settings may actually say execute karna hardware problems but like RL problems evolve.
BYeah, I see. But like it's more Physical intelligence is
Athe gold standard I think for like a robotic only company.
BI'm not holding this.
AThey are like gold standard for like doing, doing this.
BOh okay.
AYou won't feel the AGI you know, like, like different set of Problems but you won't feel like for you to feel the AGI. I think it have to be in an API phase.
BYeah. Torah chalaga Pesa subsistence level. Curious conversations. Are you only going to like speak to anthropic people or OpenAI also or other labs.
AMight not even be a culture fit like in terms of like whatever industrial corporate setting, crazy intensity. I think
Bit doesn't. It doesn't even feel like Musk is a like really grimy person, bro. Pun not intended. He. I don't know at least anthropic at least says humanity and stuff like that and like their actions show it.
AThis guy is like a very good friend of mine. You probably saw this went really viral.
BNo, I did not see. This
Ais instinct. The senior bro, huh?
BNice. Is what is Gulu working on? Video models.
AYeah. This guy, bro, this guy is directly reporting to Elon. He said he has like three, four meetings with Elon every day. Different psychic context. But like. Yeah.
BSo.
AHe also didn't think about like. Yeah, like I don't know if you thought
Bplease.
AThat like I won't do like pure money move. Like not motivated by that.
BBut I do feel like Elon might be able to out build or out like scale these other people. And if it comes down to like if he distills automatic AI research till a point, like after a point, then all you need is like scaling your compute, bro. Like
Awe build very interesting long lasting entities but the humans that build it will be like churning in and out and like like you will be like a inconsequential person like that.
BYeah, yeah, I agree.
AYeah, yeah. And then like. Bro, the world is changing so quickly that like. And taking no risks is a risk.
BYeah dude. Yeah, I. I feel like both. Yeah. Sometimes I feel like like Ray Kurzweil and like Schmidt Huber. Omega point. All of those might have some merit to them. Like what is Omega point, bro?
AI don't know.
BOh shmidh Uber's obeya point is like he just tracked like when the fundamental discoveries happened in the universe like whatever agriculture and like usepele fire. And then he plotted obviously is hand wavy. It's not a scientific theory. The times are halving or whatever. And the Omega point based on this extrapolation where it happens to be somewhere in 2045. 2040. 2045. And Kurzweil's projections are also the same. Like he arrives at that conclusion by computing time density per dollar and per energy and he arrives to like this similar conclusions by 2045, we'll have no control over what's happening. And like, yeah, like information will be representing itself beyond like human comprehension at that point.
ACrack something and like we really like fucking like ride it to the top of that technology. Like Internet was the thing, computer is the thing. And like that creates a massive impact to the society. And then like human progress is just like many such sigmoids. And then the rate at which the sigmoids happen and they're like, like that's how the exponential looks. Like many sigmoids.
BYeah.
ASo they say like, we are this AI sigmoid is like a crazy like sigmoid. And then like, maybe there's. Yeah,
Byeah. I, I felt the same way when I talked to my finance people. I was like. Incompetent person. And I can see, and I can see how this is going to like take my job and be like, it's already supercharged me. But I can also see how can it can like replace me very soon. How can it be so much of a disconnect in the world? Like, it's so hard to judge actually.
AGermany, They are running so behind in the thought process. They literally. Bro, we are raising capital and running a company that works in this thing. Yeah, like, I feel it's crazy that the difference in like being in sf, like I feel like you are talking like somebody who's in sf. Like everybody I talk to outside, even the. Including founders of like Frontier whatever company. Yeah, they are so behind in like,
Bbro, I should put all my money in compute, bro. Like, what am I doing?
AI. I bought like, I bought like 400k of Nvidia. And like, I'm like, bro, I'm not.
BNvidia is underpriced right now anyway. Yeah, good, good call. Like, yeah, like, Yeah, yeah, computer is good. But I think I should buy leaps. Don't buy stock, bro. Just buy leaps on Nvidia. Buy in the money leaps. It's a long two year option call option. It's a way to leverage.
ATalks directly, I think. Right.
BResearch. Just look into it.
ALike, this is exactly what I didn't want to do.
BThat's why I buy only V. Fair enough. Yeah.
ABut then I'm like, no, my conviction for this company is so much higher than the.
BExactly. If it is, then I think leverage it even more. Yeah. I mean, effectively time leverage. So yeah, it doesn't matter honestly for you.
ASo I didn't make any, like, my time didn't. Like, then it may be paid off when we started the company, but I don't now four years from now I don't want to think that had a good time. Zero ninja but like bro I cannot believe I'm thinking like 10 million but enough
Bno bro you need to remember what you were like five years ago bro. Like you will never be happy essay mentality say I think like no, you didn't just make 10 million. You also bought the opportunity to be go into anthropic with such.
ABut like in the other case I would have made the 10 million plus I'll be yeah opportunity cost like regret will be like yeah regret will be less bro that I saw the world like I saw the world from that
Blike story would be crazy regret minimization problem life
Abut regret interview basically high stakes decision with crazy good feel. Now I feel like cog in the wheel and GFL ripped to like 30
Bmillion or something bro Unlikely to cog in the wheel for some time like monetarily it will still be beneficial I think to shift right.
AI think I have a big void in my heart that like I won't be satisfied with all things. So I like I will if I go there I'll do like but yeah
Bbro is that that's the right attitude to have now then perfect
Alike I quit and start another company like okay like.
BYeah bro.
AIntense vibe. Very helpful.
BYeah yeah.
ACool. Take care bros.
BYes yes. Keep me updated. Yeah yeah, will do.
ACiao.
BSee you.
ABye bye. There.
Apr 5 '26, Sunday
Apr 6 '26, Monday
Apr 7 '26, Tuesday
Open this page on desktop.