0:00
So, if you're like Elon and you've got 500,000 GB300s in Memphis, and you've got diesel generators that are really loud and and annoying to the locals and they're saying like, "Hey, we get it.
Search this video
0:08
You wrote us a check, but that's not enough.
0:10
Like, I'm starting to get a headache." Pay attention.
0:12
>> Humans are raging against the machine.
0:15
In this case, data centers cropping up in their backyards.
0:18
>> Listen, [music] these are human beings.
0:19
They will [ __ ] [ __ ] up for you if you do not come across as an empathetic human being and listen to them.
0:24
>> How should we scale AI responsibly and what should it even be used for?
0:28
>> Today's guest, Anjaneyulu has some ideas.
0:30
He's an investor, founder, and viral Stanford lecturer.
0:33
[music] He's also one of the loudest voices arguing that the AI industry needs more cooperation, more compute, and a lot less zero-sum thinking.
0:43
>> Look, to each their own, guys, but Ellis, I think you may want to hold back your Bay Area critique a little bit more.
0:48
[laughter] >> I think he got you there, Ellis. >> Fair enough. Fair enough.
0:53
>> I'm Alex Heath, here with my co-host Ellis Hamburger. >> Welcome to Access.
0:55
[music] >> Anjaneyulu, welcome.
1:01
We've gotten through our technical difficulties. It's good to see you.
1:04
You are fresh off of your class today, are you not? >> Yeah, it it was good.
1:07
It was Garry Tan and Diana who from Y Combinator, and they are amazing.
1:13
But, the kids uh the kids were aggressive.
1:17
These kids These Stanford kids are really aggressive these days, you know? >> What do you mean?
1:23
>> Um I'm I'm still processing, so don't quote me on this.
1:25
But, I generally I feel like you know, they don't really understand what an institution Y Combinator is in Silicon Valley.
1:32
Like, how much YC has done for Silicon Valley, for the institution of venture capital, for founders.
1:39
Like, culturally, you know, when I grow up I I introduced uh Garry and Diana today in the context of when I was an undergrad, you know, uh 12 13 years ago, as a sophomore I had the I was lucky enough to take Peter Thiel's first um sort of zero to one class.
1:52
It was called CS183B, How to Start a Startup.
1:57
And that, you know, one of the students was Blake Masters in the class, and Blake turned it in you know, wrote up the notes with Mike Solana's help, and they turned it into that book Zero to One.
2:06
And then the next year Sam was still a partner at YC at the time, and he taught CS183B, How to Start a Startup, sponsored by YC, and I took that one, too.
2:16
And like for me growing up in Silicon Valley, it it was so obvious that when YC open-sourced their SAFE, you know, the like simple agreement for future equity, like or rather in hindsight now it's so obvious to me that they it was extraordinary for founders because it standardized the concept of seed seed raises, and it it just made things so much more efficient, and the amount of time it gave back to founders like was extraordinary.
2:44
You know, years later when I was helping the Anthropic guys put together the seed round for for that company as an angel, we decided to use a SAFE.
2:52
Um you know, and Tom Brown, who was one of the co-founders of Anthropic, he was a co-lead author in GPT-3, and we we were friends for a long time. He was a YC alum.
3:01
He had done this group chat group dating app called Grouper.
3:05
Does that name ring a bell at all?
3:08
>> Ellis met his wife on Grouper.
3:09
>> [laughter] >> Ellis did? Oh, great.
3:10
Okay, so Ellis, you owe that >> No, I'm kidding. I'm kidding. >> No? Oh, no.
3:15
[laughter] I was going to start I was going to give Tom a quick call and say, "Hey, man, you made somebody you know, somebody's life happen."
3:22
>> That's one of the core features of this show is Alex and I making up stories about each other.
3:25
[laughter] >> Okay, I don't know you well enough to participate in that, but like at some point I'm sure we'll be >> You'll get it.
3:33
>> I'll be I'll be I'll be the inside club.
3:35
I I I kind of talked about YC as the Have you guys watched Dune? >> Oh, yeah. >> Of course. >> Yeah, yeah.
3:41
>> So, I told the kids like to think about YC as the Bene Gesserit, where like they plan in decades, you know, not in in particular quarters.
3:47
Like sometimes as a as a student you're just like in the quarterly mindset cuz you've got spring, fall, you know, and so on.
3:54
But YC plans is an ecosystem steward, right?
3:57
Like when when PG and Jessica started they almost created an alternative universe where sure the business model might be a venture capital firm, but from a experience perspective, it's more like a university, right?
4:09
And my wife and I started our startups at the same time. I was in 2017.
4:14
I I think I was I forgot how old I was.
4:16
But we we were we were in our early it was mid mid-20s and she did YC and I didn't.
4:22
And I sold my last company to Discord as we've talked about, Alex, and hers was an education company and she grew to millions of dollars of revenue.
4:29
It was much more successful as a business than mine.
4:30
And I always regretted not doing YC.
4:32
And so having Garry and Diana back today in the class to speak to the students gave me a chance to to sort of context set for the students who feel like, "Oh, YC's, you know, I think some student was like, 'Some of my friends don't want to do YC anymore because they think uh blah blah blah, like it's it's too big or whatever.'
4:51
I forget what the critique was.
4:52
Um but I I it was it was a bit of a surprise for me to to realize they don't know the extent to which Y Combinator as an institution is still sort of a first call for many of the entrepreneurs, certainly in my generation who are leading extraordinary new projects and companies and alums from YC who have gone on to start OpenAI and Anthropic.
5:13
Like as as if measured by the value created by YC alumni in the world, I don't think there's another institution like that.
5:19
And so I'm still digesting.
5:19
I I know Diana and after the um the main lecture, Garry and Diana did a sort of smaller office hour session with some of the students.
5:31
I could see that like there are some big misconceptions about YC that need to be corrected. >> Like what?
5:36
What are we I mean, I I I think YC is still considered prestigious, but you've got all these new accelerators.
5:42
I mean, Ellis has done work with Speedrun, your former employer Andreessen's has got their thing.
5:49
>> I I I just saw YC had like a $5 billion stake in OpenAI.
5:53
I think that just came out in the Elon Musk trial that they uh you know, that equity from helping incubate OpenAI is now worth $5 billion, which I think is YC's largest position.
6:03
Uh and so I I think YC is you know, unequivocally goated in terms of all the things you said, but I What it What is it about the the new generation?
6:11
They just don't get it or >> Do they think it's outdated?
6:15
>> I mean, so much about company formation and the advice that's valuable, even just in how the advice has shifted from maybe PMF world to distribution world.
6:23
I wonder uh how YC and the other accelerators have kept up or not in their mind as uh as the youngest innovators we have now.
6:33
>> You know, we do uh for every speaker, we do feedback forms.
6:35
And so once those come in, I I'm going to read and and see what what the students say.
6:42
But I I don't want to start conjecturing just quite yet. >> Fair.
6:46
Well, how does it feel to be running AI Coachella?
6:47
I saw that you were going to print t-shirts out after that went viral. Have you done that? >> We We did.
6:53
I Rosie, do we have one we can show on the >> Yeah. >> Okay, it We do.
6:57
It It It has CS153 on the pocket, but on the back it says um I took AI Coachella No, what is it?
7:03
I took CS153 and all I got was AI Coachella.
7:07
Um and so we have 500 of those t-shirts we're handing out this quarter.
7:12
And it's just a fun way to remind the kids that they should take life seriously.
7:15
Of course, they're taking you know, computer science at at Stanford in the middle of the AI revolution, but they shouldn't take it so seriously that they forget how to have fun.
7:25
You know, I I think back to my spring quarter as a senior and the one thing I didn't do enough of was I I I had already start working at Kleiner Perkins at the time.
7:35
Um I was working nights and weekends there and I had the chance to essentially be bag boy for John Doerr.
7:41
Uh my first board observer seat was on a board of a company that was based in Stockholm and they made me the board observer. John led the investment.
7:47
We we co-led it with Sequoia.
7:49
I I was trying so hard to like grow up that I forgot, you know, how to have fun for a little bit and looking back that's my biggest regret.
7:58
I I don't know if you guys feel the same way, but I feel like sometimes when you're in college you're trying your damndest to like get into the real world and you get to the real world and you're like wait, this is it? >> I don't know. I want to go back.
8:09
>> I had a lot of fun in college.
8:09
I don't know about Ellis, but uh I don't know.
8:11
I don't think I worked hard enough in college.
8:15
>> Might just be me then.
8:16
>> [laughter] >> I was the only English major in the business fraternity.
8:21
So I was attempting to be entrepreneurial and be interested in finance and investing.
8:27
Um but yeah, I guess I've found the perfect place for those skills this this point helping write manifestos and missions with founders that doesn't destroy their whole product strategy.
8:37
So maybe I did learn something in college.
8:40
>> When you say business fraternity Ellis, which one was that? >> Uh Phi Chi Theta.
8:45
>> Oh, you were in Phi Chi Theta. Okay, I was in AKPsi.
8:48
>> Oh yeah, the nemesis the nemesis.
8:50
>> Oh [ __ ] >> Was it the nemesis?
8:51
Okay, I didn't even know. Sorry. I didn't know. Okay.
8:54
I I I rushed it because I was trying to flirt with my now wife then non-girlfriend Vivian cuz it was the it was a co-ed fraternity.
9:01
Um >> Yeah, so was PCT which was incredible as well. Yeah.
9:07
>> And which campus were you on?
9:10
>> Uh University of Michigan central campus. >> Oh cool. Okay. Awesome. Nuclear central.
9:18
Uh there's a bunch country's best nuclear engineering program. >> Is that so? >> Michigan? Oh, yeah. Top top three for sure.
9:25
>> Is that something that stumbles across your desk usually Alex? Not mine.
9:29
>> [laughter] >> No, I can't say it does.
9:30
Uh that's an interesting factoid.
9:34
Uh yeah, no I I I was completely left out of all fraternity business.
9:37
So, all these words you guys are saying means nothing to me, but it sounds interesting.
9:43
>> You didn't miss out on much, dude.
9:44
[laughter] At least not not from from I like the thing about these in in in in lecture one of the class, which is now up officially on Stanford's YouTube channel, if you want to go check it out. It's CS153 lecture one.
9:57
I kind of talk about the AI scaling laws and I give them life scaling laws and I talk about how you might they might think that the most important speakers in the class people in the class are the speakers.
10:08
You know, we have Satya, Sam, Jensen, and so on.
10:10
But really it's the people in the seats next to them, right?
10:13
It's when they look to the left and right, it's the it's the students.
10:14
Like it's it's each other that they're going to build things together.
10:18
They're going to be the next generation of leaders and you took advantage of your time as a student to have fun with the people who mattered to you.
10:27
And so I was trying to remind them >> What's interesting though, isn't it?
10:29
I feel like so much of traditional higher education or even previously is kind of about teamwork and you read headlines today about the first solo billionaires and how little help you might need if you're adept at AI.
10:44
Do Do you feel like that's changing how the students are thinking about the companies that they're starting?
10:49
Or is it still the classic product founder and engineering founder?
10:55
>> So, you know, big picture I think about 80% of of the challenge of building a company is is the same.
11:02
And the 20% that's changed is still not legible to most people.
11:08
It's It's got new words and so on like, you know, that if you look at the the class, we've got speakers which who cover content like theory.
11:15
And then we have a a a project, a final project called the one-person frontier lab, where we basically say, "Hey, today with the right tools, you know, one person can accomplish what previously took many many groups of people."
11:30
So, as part of the final project, we've given them compute token credits and they can just they're they're welcome to see what they build as a single-person uh team, a one-person lab, and scale themselves.
11:43
But, we had Andrej Karpathy uh last week, and he said uh so I asked him, you know, Andrej, if you were if you were the head TA for the all these students, right, and their final projects, and they were asking you for what to build, what would you build?
11:55
And uh he said, "Yeah, you know, actually some of the most interesting problems to solve in industry right now, especially in frontier AI, are ones where you have coordination between multiple teams.
12:07
Where at that uh like at the moment, most labs are not cooperating. Right?
12:11
They're kind of playing the solo game.
12:13
So, if you think about a student in the class as kind of a microcosm of the entire industry, um we have not gotten yet productively to the stage where multiple labs are competing to try or sorry, cooperating to try and move the capabilities frontier forward.
12:25
But, could you imagine, like, for a second, just think I know it's hard to imagine, but here, just for a second, if if OpenAI and Anthropic decided they were going to train a model together, the world's best coding model together, can you imagine what the capabilities might be might produce?
12:45
>> said they would do something like that.
12:46
I'm old enough to remember when uh these companies were talking about cooperation, you know, around AGI and laying down arms and helping the other.
12:57
Uh it it used to be the spirit of that. Well, didn't it?
13:00
>> Yeah, and that used to be the spirit of Silicon Valley more generally, right?
13:03
Like, coopetition has been the reason why Silicon Valley has flourished.
13:06
Like, I I grew up in the tradition of venture capital where I I started at Kleiner when I was, you know, 20, as we discussed, and then I spent 4 and 1/2 years as an investor, and then I started my own AI company, and I got to raise from Benchmark and Index, and I had them on my board, and you know, um So, Kleiner and and First Round were part of my seed round.
13:26
Uh Index led my Series A, Benchmark led my series B, and I sold a company to Discord.
13:30
And then when I got to Andreessen Horowitz, you know, I I was led the series A into Mistral and co-founded Black Forest Labs and had the chance to get Ale Marina get going and I'm on seven different boards and and I and I just sort of fundamental part of this my my my psychology is you cooperate with people who you nominate who the world nominally thinks you compete with, but that's not true, right?
13:50
The way we like a here I'm sitting at Periodic Labs' office.
13:54
Um this is my current incubation.
13:57
I spend three days a week here, Tuesday, Wednesday, Thursday.
13:59
We're discovering new room temperature superconductors.
14:00
The co-CEOs are Liam Fedus who co-created ChatGPT and Wojciech Zaremba who was um who led the quantum uh materials team at DeepMind.
14:09
And sure, I led the seed round when I was at Andreessen Horowitz, but we also had Lightspeed and we had um I think we had like Felicis in there for a little bit of money and we had a bunch of other firms, right?
14:20
And and we and and that's how we like venture capital has had this cooperative spirit because when you're in an era of abundance and and these companies succeed, there's more than enough for everybody to go around.
14:32
Also, they just need different kinds of help.
14:34
And so it's not zero-sum.
14:36
And I think we're in a stage of the industry in frontier AI where zero-sum thinking is unfortunately quite dominant and we need to get to a place where positive-sum cooperation is is uh and positive-sum research is is the status quo is the consensus, not contrarianism.
14:54
If we get that, oh my god, like the capabilities jumps we could add for humanity to the frontier of science, engineering, biology, health care, we could genuinely be solving cancer already if we were consult like if we were pooling compute, labs were like co-training, you know, models and commercializing together, the way the biotech industry has has tried to do in the in the past.
15:16
Um you know, you could give the biotech industry a mixed scorecard.
15:18
I think we should take some of the best learnings from biotech and bring it to frontier AI.
15:22
But if that was happening, I I think we would have flying cars by now.
15:26
>> We've got self-driving. We're close.
15:28
>> But it's so slow, man.
15:29
It's taken us We're like, you know, Waymo was getting started when I was in college, and it's been 11 years, you know, like I don't think we have time to wait 11 years again for go from self-driving to flying cars.
15:43
>> Well >> Can we just skip ahead to the good parts?
15:45
>> Well, so help me square what you're saying, which I I agree with broadly.
15:46
I don't think it's possible, but I agree that would be nice if that happened with your class project, which I understand is the one person frontier AI lab.
15:55
That's what you're pushing students to do is is build a lab a solo lab.
15:56
Is the idea that they all then work together and cooperate?
16:00
Or like are they all doing it on their own and competing with each other?
16:04
>> We we did ask for So, it's not necessarily it's not one person frontier AI lab, it's frontier lab.
16:08
There's different kinds of labs.
16:10
I think frontier technology lab because you can apply AI to a different domain, right?
16:14
I think a frontier AI lab sort of sometimes means, "Oh, you are putting out Like what you do is you train machine learning models as Like you're training AI capabilities as a service.
16:27
I'm saying you can be a frontier technology lab that uses other people's capabilities.
16:30
Like if you're using Claude, and you're pushing the frontier the state of the art.
16:33
Frontier basically means state of the art in my view.
16:35
And you could be pushing the state of the art in cancer therapeutics.
16:38
You could be pushing the state of the art in material science.
16:41
And you could be pushing the state of the art in in creativity.
16:42
These are all different frontiers that are that can be pioneered that can be bootstrapped by the acceleration of capabilities in software engineering, right?
16:51
Like for from Claude, for example.
16:54
And so, what I'm what we're trying to do to encourage what what what we're trying to encourage the students to do in the class is find the mission they really care about.
17:01
Like what's the mission in life that you're most passionate about, and then align yourself on that on that frontier, and then use all the tools available to you today to move the state of the art in that domain. Does that make sense?
17:14
>> Are you finding a lot of diversity for what missions they want to achieve cuz, you know, I've I write missions almost every day and when the founders are all from a similar upbringing, similar culture, similar place, similar society, you often hear a lot of the same sets of things and I think what I hear more than anything is founders, you know, wanting to empower people to focus on what matters, which to me is just kind of a non-answer.
17:38
You know, it's just like, well, that's that's the go-to of the modern age in the same way that people say frontier instead of state of the art now.
17:48
Um do do you feel like you hear a diversity of ideas? >> Not enough. We need more.
17:53
We need less groupthink for sure.
17:56
We need less status seeking.
17:58
Like what I find is, you know, humans are status seeking monkeys.
18:03
>> Especially in San Francisco.
18:03
That's the rainforest for the status seeking monkeys, I think.
18:08
>> Well, well, I I don't know.
18:08
Where Where Where Where are you based?
18:11
>> Yeah, I don't think we have a lot of ground to stand on being in LA. >> Are you kidding?
18:15
>> [laughter] >> Oh, LA is way worse.
18:15
Oh my god, LA No, no, no.
18:19
At At least in in the Bay Area, the status that comes from the pursuit of difficult engineering problems and ideas and so on, right?
18:24
Everyone's trying to outdo each other in terms of What are the hardest problems you can tackle?
18:29
And so Elon's trying to outdo Demis and Demis is trying to outdo Sam and they're like, well, who who can produce the best capabilities for humanity?
18:36
Again, we're generalizing, but in LA, like what status What does status come from?
18:39
I don't even know at this point >> Status is how long you wait for the matcha community goods when it's raining.
18:46
>> [laughter] >> Look, to each their own, guys, but Ellis, I think you may want to hold back your Bay Area critique [laughter] a little bit more.
18:54
>> I think he got you there, Ellis. >> Fair enough.
18:55
[clears throat] Fair enough.
18:57
>> Andre, I was reading this Atlantic story, I don't know if you got to see it about Stanford.
18:59
I think it's based on a book excerpt and there were a couple good lines.
19:03
I didn't I didn't agree with a lot of it, but there were a couple good lines.
19:06
One was Stanford is an incubator with dorms and then another thing about pre-idea funding, um the first the first one, incubator with dorms, maybe your class is an example of this, like you're encouraging people to make companies in the class and you're bringing all these entrepreneurs in.
19:24
Um do you feel like higher ed, but Stanford in particular, has become overly commercial, it's not actually about the academics, it's not actually about what it used to be, or do you think this is uh more nuanced?
19:38
>> Well, I think America's incubator with dorms, right?
19:40
dorms, right? Like if you just look at I mean, literally, broadly speaking, if you just look at the history of technology, whether it's Bell Labs or Alphabet Holdings now or MIT or like DARPA, DOD, I mean, this is where some of the most foundational technologies in humanity were created, but like from a
19:58
like if you just zoom out, like the you just had a building where you had research scientists and engineers doing research, and then you had a building where they were where they used to live, and the closer you put those things, the more and more they look like, you know, an incubator with dorms. So, I don't I
20:10
So, I don't I don't think that's a very novel critique.
20:13
I also don't think this is this commercial proximity that Stanford has to Silicon Valley is new at all.
20:17
In fact, my whole class the whole spiel I gave you guys at the beginning about how this is just a long lineage in like Peter Thiel's class and Sam Altman's class, and there were a bunch of others, and now you got my class.
20:27
This has been the case for a long, long, long time, and I think it's the it's the magic engine that keeps the dynamism of Silicon Valley going.
20:35
What is different now, I think, is that there's a uh the scale of capital required to continue deploying capabilities or like making meaningful progress, you know, at the frontier, and the fact that a lot of the scaling requires the marshaling of physical resources, like compute infrastructure,
20:53
that and and that we're hitting bottlenecks of that in America, where there just isn't enough energy to produce all the compute we want, requires a level of um alignment with global capital allocators that is quite new for for a lot of Silicon Valley folks. And uh you know, I think whether
21:06
folks. And uh you know, I think whether it's the Middle East or um it's India or you know, I was on a call this morning with um the Ministry of IT of a well-known country like a country in the Middle East that has extraordinary amounts of energy, and they're all you know,
21:27
it's what's fascinating is the leaders running like coming into who are starting to get like those senior-most positions in the in governments around the world actually have grown up with technology and disruption theory as very much part of their psyche. Many of them
21:40
Many of them actually came to America as college students or under a post-grads like many of them came here for business school and then went back to their countries and have become sort of policy makers and so on.
21:52
And so they there's a level of understanding of Silicon Valley culture that's quite native. Right?
21:58
And like the the the the gentleman who runs the Ministry I was um on the call with this morning actually used to work at a technology company in San Jose, and then when he went back to his country he was the country GM for that that company, and then was tapped by the government to come in and and lead the Ministry of information technology.
22:16
And when I chat with him, you know, is I was on TVPN yesterday.
22:21
I was chatting about the game the GameStop eBay acquisition with the with the TVPN guys, and he sends it he he that tweet went viral, and I woke up this morning and he had found the tweet and put it in our group like our WhatsApp.
22:37
And so the fact that you've got this essentially the CTO of of a country halfway across the world is is watching me comment on Silicon Valley uh sorry, technology acquisitions and is sending it back to our group chat within 20 12 hours. Like that's new.
22:53
>> I want to make sure I'm understanding the geopolitical implications of GameStop buying eBay.
22:59
>> Of GameStop buying eBay for the reasons I outlined, which is that agent agentic commerce needs physical verification.
23:07
And so, you know, what they were trying to understand is okay, if they want to enable agents to do physical commerce in their countries, does that mean they also need to have integration between a physical verification layer where somebody can go into a store and say, like a physical location that says this pen is real, it's been verified, and now it's going online for sales. >> Right. Talk about that.
23:30
>> The the single biggest problem Oh, wait.
23:32
We talked about this two three four years ago, Alex, when you and I did the Hangout for Discord's like demo day when I was platform, right?
23:41
>> We were we had just come off of I was still, you know, I was the head of platform at the time.
23:45
And we'd come off a 24-month run at Discord where the it had become this uh like this incredible petri dish for e-commerce where people who are passionate about some something You know, Discord was a bought a chat a chat app for gamers that became a community uh chat app, right?
24:03
There were all all sorts of communities who were like into keyboards and sneakers and so on hanging out on Discord during COVID.
24:12
And you start we started observing that many of these communities started adding a particular kind of channel called BTS.
24:18
Does that mean anything to you guys? BTS? >> Behind the scenes?
24:23
>> [laughter] >> Is that a band in Korea or something?
24:26
>> It it is, but it's also buy buy trade sell.
24:30
>> What are the names of these group chats?
24:32
And what apps are we talking? Are we Signal? Are we WhatsApp?
24:33
What's Give us the tea here, Anj. Come on.
24:37
You're with the masters of the universe in disappearing group chats.
24:39
They they've got to have a good name at the very least.
24:42
>> Illuminati with a bunch of different emoji stars.
24:45
>> No, I I can't give you the names. I'm sorry. >> divulge the names.
24:49
>> I've already I've already said too much.
24:50
You know, the rule number one of Fight Club is you don't talk about Fight Club. >> Yeah. Yeah.
24:54
Before we move on from from Stanford and all of this, um you've had a lot of impressive uh leaders come into the class already and you've got more booked.
25:02
Um Jensen's already come in, right?
25:04
So, you've had you've had some the big ones, the biggest one.
25:08
Um any standout moments uh so far that people can go watch everything on YouTube.
25:13
It's hours and hours of content, but uh maybe not everything.
25:16
There's maybe moments that aren't on YouTube, but yeah, has there been a standout moment uh besides like students not caring about YC? >> No, no.
25:23
Okay, I want to re- Please, you do not put words in my mouth. I did not say that. >> Okay, okay.
25:28
You you did not say that.
25:29
>> The kids were ecstatic about YC, by the way.
25:31
I mean, it was non-stop questions.
25:33
There were kids from like who are not in the class who were dying to get in. We had to turn away.
25:36
We had like bouncers at the doors.
25:38
So, just to be clear, very popular lecture.
25:40
You always have a few kids in the class who who have like who do like um they they like read Twitter and then they like they they're very combative in their questions and I think that's not a healthy posture.
25:53
Like I I I we'd like it's a systems class and so I encourage first first principles thinking, which is like analyze the system that you're trying to ask a question about and then ask that question.
26:02
Don't say, "Oh, I read on Twitter that like blah blah blah did blah.
26:07
What do you think about that?"
26:08
It's like, "Well, that's not an interesting question, right?"
26:09
Um anyway, to answer your question, what were the best moments notable moments in the class?
26:15
You know, we we we promised the speakers that they their teams will review all the the lectures before they're published online, but they all will be published online.
26:22
So, I'm going to hold off on this one just cuz I don't know what'll make it through the >> Okay. >> the ringer. >> Okay. >> Yeah.
26:29
But there But there's some moments that I think will go down in history because it's a very honest exchange, right?
26:35
Between students and leaders.
26:35
Like one of the re- like what self-selects people who come back or who come to campus.
26:41
This is the fourth year we're teaching the class and what I've noticed is when we invite folks, it selects for a type of leader who feels responsible for the future because they don't They have so much other stuff going on.
26:52
They don't need to invest the time to come back and speak to these students.
26:55
They can get get on your podcast and scale their message.
26:57
They have full comms teams.
26:59
But when they come back and they sit in the class and they stare at these students in the face and they're looking at like the you know, literally the their the the future, right?
27:06
Like these are kids who are going to grow up and and run the world, there's a level of humility and honesty and transparency that is rare in today's environment of comms training and and media training and and can you know, being cancelled and and so on.
27:23
It's it's a very powerful setting I find.
27:25
It's very raw, it's honest and um I hope I hope the powers that be at both at Stanford and the comms teams who will review all the lectures will will let some of that vulnerability shine through.
27:37
>> We can make up a moment, they say.
27:39
Back in my day >> Oh, so many of those.
27:41
We used to only get access to we only used to get access to new primitives with the new iOS update every fall.
27:48
Cuz that was that was startups forever, right?
27:51
It's like every year Apple would relieve or release a new SDK of some kind for a new sensor on the phone or a new API and then a hundred startups would spin up around it and yeah, I mean I'm curious like how how do you advise these students or your guests advise these students to navigate that?
28:09
I like the mission approach, but it's like tactically, I mean I feel like half the startups I work with are pivoting during our conversations in some way just cuz the the foundation is changing so quickly.
28:21
>> So I I think pivoting is a feature not a bug.
28:22
You know, typically what I've learned is whether it's Anthropic from the early days or periodic, like it depending on what you mean by pivoting and there's no canonical definition of pivoting, like if you're not iterating with massive updates based on new information coming in and so hill climbing so to speak, you know, towards your mission, you're just not making progress.
28:40
And so um my biggest regret of course in life as a founder when I started Ubiquity 6, which is the last company I I failed at it miserably, was I wasn't updating fast enough because I felt like I had committed something uh to my investors and and my team that I had to hold constant and consistent artificially.
29:01
And now in hindsight I'm like, that was so stupid.
29:03
Like I wasted so much of my life um trying to hold constraints that were not not really there.
29:10
Um and so I think I I I I I think the thing you should hold pretty constant was your mission, right?
29:14
And I I if you if you're if you say "We're starting a superconductor discovery lab to move the frontier of superconductors."
29:20
Okay, then let's do that.
29:24
And then hold everything else flexible. Right?
29:26
Um I'm not of the school of thought like famously Mark, you know, Andreessen who I've learned a lot from because I was a partner at A16Z on the GP on the infrastructure fund and I went there primarily to learn from Mark and Ben.
29:36
You know, I think my I I I I have read that quote Mark has, right, about product market fit, which is when you don't have it when you have it you know it and if you don't have it change everything about your company to to get there.
29:45
And I I I I think you can change everything except two things, which is your mission and the people who you go on the the journey with.
29:57
It is very very taxing uh especially on the culture of these of your company to keep changing the mission in a way that's not legible to your team for why you're doing it. Right?
30:10
And that feels like trash.
30:10
It feels like purposeless.
30:13
It feels like you're chasing somebody else.
30:15
It it it's just the worst.
30:17
You know, what's really exciting is when you pick a really big meaty problem and you're like, this is a hard problem.
30:23
I could dedicate my whole life to this mission.
30:24
In fact, I would I would get I would pay somebody to let me work on this.
30:29
That's how much I love the idea of like if if on my grave, right?
30:31
Like on my tombstone it said these words like, "He figured this out."
30:39
That's that's the quality bar I think we need to have for missions, of startups.
30:43
And then you hold everything else, you know, sort of flexible.
30:47
Um and then you should I think you want to pick a team that that is as excited about that mission and is super flexible, because you learn so much in a world that's evolving so fast, that like actually I think your team finds it very liberating when they're like, "Yeah, we thought we were going to do that, but actually there's a better path to our mission, so let's do that."
31:00
You know, I I when I was a founder, I I used to do these weekly all-hands.
31:03
Um And these are the they are these, you know, it's a it's a ritual that I is is canonical in Silicon Valley, right?
31:10
You get your team together, you describe how the week went, where you're going, and so on. Sort of very anchoring.
31:17
Um and over the years I did that and on the boards I've been on, I've realized that a pretty good mental model for how to think about a startup is you've got like a destination in mind and then you're going on a road trip with your team and nobody has any clue on the exact path on how to get there.
31:33
And if you're the founder, you're in the driver's seat and you're feeling really stressed and responsible, cuz you got to get people there eventually.
31:39
Sometimes you have to take a detour to to get there, but if that detour is not legible to people, like why you're taking that detour or that it's just a pit stop along the way, cuz somebody's got to use the bathroom, and we're getting back on the freeway to the destination, [ __ ] goes really hits the fan.
31:54
And I find if you think about yourself as a as a leadership crew that way, which is one, pick a really fun destination that everyone wants to get to.
32:03
Two, fill the car with fun people who love the song.
32:05
Like I I love um like I'm not a very when I got to America, like I was born in India, I grew up in Singapore and then I came here for college at Stanford.
32:13
And Vivian, who's my wife, she loves country music.
32:15
She's like a country junkie.
32:16
And I'd never listen to country music growing up. I was a Linkin Park guy.
32:20
And so we would get into the we would go on these road trips and I'd start blasting Linkin Park and then she'd be like, "Well, can we please put on some Darius Rucker?"
32:28
And I was I was like, "Darius Rucker? What's that?"
32:30
And so we put on her like we had we had a little playlist called country chicken fried or something like that.
32:37
And uh you know, originally I was like, "What What is this music?"
32:39
And then but but I I had to listen to it cuz it she was I was trying to date her and I love her, right?
32:44
And so like become now I'm a country junkie. I love my Wagon Wheel.
32:48
And it's because I love being in the car with her and being able to enjoy that moment of listening to the music with her.
32:56
It gives her so much joy. It gives me joy.
32:59
And startups are no different, right?
33:01
Fill the car with people you love and sometimes they want to listen to some random music. You know what?
33:04
You're going to learn something about them, right?
33:08
And >> I think they're all listening to the Social Network soundtrack, but continue.
33:13
>> Yeah, it it's a bit too gloomy for me, man.
33:14
I used to listen to it in college while coding cuz I thought I was like Zuck with my hoodie on and like >> Yeah. >> do do do do.
33:19
And then at some point I'm like, "This is so lame."
33:21
>> Wagon Wheel is the new uh new coding anthem. You heard it here first.
33:25
>> Wagon Wheel Wagon Wheel great great pump up song for uh um for all hands.
33:30
The other one, you know which is great, is um the Chef soundtrack.
33:33
Have you guys heard Chef as watched Chef the the the the movie on Netflix? No? >> Oh, yeah. Yeah, yeah, yeah. >> Amazing soundtrack.
33:41
>> Yeah, it's a good cut.
33:41
>> Yeah, he's got one where he's like, "I like it like that."
33:42
It's he's got that whole remix. And I love that guy.
33:47
>> It's it's such a good cut.
33:47
And so that the you know, that's so much better than the Social Network music.
33:50
Um >> [laughter] >> the the Spotify epic soundtracks playlist is all is is great for coding though, I would say. That's my that's my jam.
34:00
In the class, you know, for the for 153 we play like walkout walk on songs and um we had a Scott Nolan, who's the founder of General Matter.
34:08
He used to be a GP at Founders Fund.
34:10
And he came by and we were in the car like driving to Stanford and I was like, "So, what's your song, Scott?"
34:16
And he starts playing the sound the the opening theme for um for Dune. Right? Was for Dune?
34:22
It was for [laughter] Dune. Yeah. And it was so good.
34:25
And we ended up playing that.
34:25
Um Look, I I think everybody in Silicon Valley has a different soundtrack playing in their head.
34:31
You know, when I was with uh Ben Horowitz, like I was having he was he was one of our guest speakers 3 weeks ago and I said, "Ben, what's your song?"
34:38
And he he asked me to line up one from Nas. Right?
34:42
And it's so emblematic of what >> He loves Nas. >> Yeah.
34:45
>> That's who he sees himself as, right? In his head. He's like Nas. >> Yeah.
34:48
>> Um and and and Scott Nolan feels like he's one of the uh you know, the the protagonist from Dune.
34:54
Like like and that's that's the beauty of Silicon Valley.
34:56
Like it pulls together people from all over the world who have their own lore.
35:00
And this And And when you're going on this this road road trip on the startup, um well, the first thing I love being able to go on multiple road trips with the founders I work with.
35:09
But each startup is just a road trip and you have to figure out what what's the music that's playing in that car and who are the people in there that love each other and who can't stand each other cuz if they can't stand each other, you you got to get them out of the car.
35:22
And then two, what's the destination they all love?
35:26
And And the sooner you like debug that, the rest of the startup stuff, the pivot stuff, it all takes care of itself, man.
35:30
>> So, speaking of road trips, you've got your own you're on and with the limited time we've got left, I do want to ask you about Amp.
35:36
You You broke the seal on the first VC on the pod.
35:38
We've had a no VC rule because we've been really trying to just have founders on, but you are a founder, you've been a founder, we've talked all about that.
35:47
>> Alex is attempting to gain entry into your group chat. >> Yes. Um >> Oh, I see.
35:52
That's >> to butter you up.
35:53
Um but actually no, I'm trying to understand Amp because there's been some reporting on it.
35:57
You haven't said a ton about it.
35:59
It's this new compute startup you're building.
36:01
There's rumors about raising billions and billions of dollars.
36:05
I'm not sure what's been confirmed, what hasn't.
36:06
What can you tell us about it? What's the state of it?
36:11
>> Um Amp is a public benefit corporation, which means everything we do is actually governed by a public benefit charter.
36:16
Um there's two reasons why I did that.
36:19
The first is I didn't want to get sued for giving away billions of dollars of compute at at cost uh to teams that I think are doing interesting frontier research.
36:29
One of them is Periodic Labs.
36:30
Uh just because that wasn't legible to shareholders in the short term as something that's that's actually the best for the company long term.
36:39
Um and then the second is that I do think if you believe that frontier AI progress is one of the most uh important uh arcs of our lifetime and that if you can unblock progress in domains like you know, health care and material science and engineering like here, that the the primary bottleneck on that progress is is compute infrastructure.
37:01
And if the public doesn't feel like that infrastructure is being it's it's inter- computer infrastructure is a very scarce physical resource, right?
37:07
It's It requires land, power, shell, water.
37:09
It it it it it is It takes a very physical place in communities, especially in America.
37:17
And my view is that if it's not legible to the public that the these data centers and the infrastructure required to unblock this kind of frontier technology progress is serving their benefit, then it's not going to work out.
37:28
They're going to say shut it down.
37:30
We don't want this around here.
37:30
And to be honest with you, we might have to do that because America is a democracy and if the communities don't want the infrastructure in their neighborhoods, we'll have to move it elsewhere.
37:39
And I'm prepared for that, but I >> It's happening.
37:42
I mean, the backlash is strong. >> It's strong.
37:44
I don't think all of it is actually unfair, either.
37:46
Um I think on the spectrum of like what should we do with compute infrastructure, there's two at least ends of the spectrum, right?
37:55
There's the accelerationist movement, which is like just what you're talking about.
37:58
This is the like it should be a no-brainer.
38:01
Let's just keep build build build build build.
38:03
And then on the other side, you have folks who are literally like shut it down, like stop AI.
38:07
And obviously, neither is is is the optimal place.
38:09
The optimal place is let's scale infrastructure op- optimally while making sure we listen to the concerns that are substant- of of local communities.
38:20
So, if you're like Elon and you've got 500,000 GB300s in Memphis running at like non-peak utilization and you've got diesel generators that are really loud and and annoying to the locals and they're saying like, "Hey, we get it.
38:31
You wrote us a check, but that's not enough.
38:33
Like, I'm starting to get a headache." Pay attention.
38:35
Listen, these are human beings.
38:37
They will [ __ ] [ __ ] up for you if you do not come across as an empathetic human being and listen to them.
38:42
And uh I I can't speak for Elon.
38:42
I don't know if he is listening or not, but I can tell you communities around America are not happy.
38:48
And uh so, when I set up Amp as a public benefit corporation, we said, "We everything we do has to be in the public benefit."
38:54
And if you don't if if our teams who are training frontier models to do discover new materials and new cancer cures, um you know, need compute and they they come to you and when we come to you and say, "Please give us some part of your your your land to help us bring up that compute."
39:10
If the answer is not hell yes, then it's a no.
39:14
And so, I think of Amp as a new kind of sort of public benefit institution. We have two businesses.
39:18
We have an infrastructure business that secures compute for frontier research team, scientists, engineers who we think are doing interesting novel work.
39:24
And then we have a capital business that raises the money required to help these teams secure like access the compute they need and do research.
39:31
Our venture capital business is called Amp Foundry.
39:34
We have had billions of dollars in commits and we haven't announced anything publicly yet because of a few regulatory filings we got to make.
39:41
Um but we will re- reveal that publicly soon.
39:43
And then on the infrastructure side, uh by the end of this year, we have yeah, several hundred megawatts of compute capacity coming online, which is already started.
39:51
Um upstairs is the Amp team and we were debugging a cluster earlier today and we do have thousands of chips that are running in production and models are being trained.
40:01
And the closest analogy I can think about us is PJM Interconnection, which is the you know, for the in the in the Northeast, depending on where you grew up.
40:07
Uh it's it's a utility company, right?
40:09
It's a it's what's called an independent system operator of the grid where it doesn't own its own data um in the in the electric grid world, you have a network of a few different entities.
40:19
You have power generators, you have factories, and you have transmission lines, and then you have a utility company that coordinates capacity along all of these folks, you know, pools capacity, drives up utilization, make sure everyone can provision for their base loads and not their peak so that there's no blackouts.
40:35
That doesn't exist for compute.
40:37
And so, we're an independent system operator of the grid. We pool data centers.
40:41
They you know, our grid is data centers, which by the way run on electric grid.
40:44
So, you can think of us as a grid on grid on a grid.
40:46
But, we have you know, cloud providers, AI teams like Periodic, and we have the transmission lines or optic fiber cables.
40:55
And we coordinate capacity across those.
40:58
We do long-term commitments across different clouds.
40:59
We pool capacity across clouds.
41:01
Um we allow our teams that our teams commit to procuring the bulk of their capacity from the grid over 4 years.
41:08
But, like a flexible meal plan, they can spend whenever they want with sufficient heads-up notice to us.
41:13
It's kind of like a dorm meal plan, to be honest.
41:14
And we don't own our data centers.
41:17
We don't own our own transmission lines.
41:18
We are We're We're a neutral third party.
41:20
Now, we do secure a lot of that capacity and give it to our portfolio companies at cost.
41:26
And if we have any excess left over, we sell it to the ecosystem at a modest profit. Does that make sense? >> Airbnb for GPUs?
41:32
Am I being too reductionist here? >> No. I love that.
41:38
>> [laughter] >> You can keep it. >> that. Okay.
41:42
>> Well, it does change the equation, doesn't it, when you talk about what the data center is for.
41:45
I think it says one thing to someone if a data center is being built in the neighborhood.
41:49
They said, "Hey, what's this going to be used for?"
41:53
And they say, "Well, there's this newsletter guy named Alex Heath who needs it to run his granola for like his bajillion conversations he has all the time."
42:02
Meanwhile, if it's for something that they can actually feel that's solving a problem in their lives, I think that does change the equation.
42:07
And I guess it just seems like a lot of that stuff maybe just hasn't come to pass quite yet in terms of the outcomes of AI.
42:15
I mean, what do you feel like are the brightest spots you've seen uh with with uh productizing some of this AI stuff outside of outside of tech?
42:25
>> Like you want when you go pick up you know, a juice or whatever now you have a label that says here's what where this stuff came from.
42:31
I think we have an inverse label on data centers where people say will say, I want to know what what is the label on the data center say for who is using this and for what?
42:37
And then I get to vote on whether I want that in my community or not.
42:43
And uh if if if the label says this is a data center that's ultimately being used for creativity because that's what is powering like it it's being used by some application somewhere that makes video editing easier or if it's being used for to discover new room temperature superconductors, which is what we're doing here.
43:02
Um that is the level at which communities will want clarity before they provide strong enough vocal enough support for data centers.
43:11
I hope we can self-enforce that level of transparency before it's too late because once the government gets involved and starts regulating things it always gets implemented like in some watered-down game of telephone way of what the technologists who actually understand how this stuff is being done can implement right ahead of time.
43:32
>> I'm uh visualizing a trebuchet laying siege to the Instagram Reels data center.
43:38
That's what's in my mind right now.
43:41
Surely someone's made an AI video of that.
43:43
>> The irony is I'll probably have to buy that trebuchet on Instagram cuz their shopping is amazing. >> [laughter] >> Yes.
43:50
>> The ads on Instagram are so cool. It's incredible, right? Like I love pens. I'm a huge pen junkie.
43:55
Like I spent I spent like $300 on pens on Instagram last week.
44:01
It's extraordinary and I feel like I learned something cuz I'm a nerd.
44:03
I mean, this is the eBay opportunity, right?
44:06
If Ryan can actually figure out how to turn eBay into a shop into a destination where you don't feel like you're, you know, in the back streets of like some like some bizarre where like some guy jumps out of the shadows and says like, "Hey, hey, take my pen."
44:18
But and it's like a fake.
44:19
But instead it feels like this delightful I don't know. What what is it? Abbot Kinney? Right?
44:25
If you can make eBay feel like If you can make eBay feel like Abbot Kinney, I mean, that's a trillion-dollar company, don't you think?
44:32
>> I mean, it's an interesting idea.
44:32
I definitely haven't thought about that.
44:35
I've been behind on the GameStop saga, but yeah.
44:37
I mean, yeah, that's very cool.
44:40
I'm just still thinking about the flops.
44:41
Are the flops being used? Are the flops flipping? Etcetera.
44:43
So, that's where my mind is.
44:47
>> definitely want the flops to flow like megawatts in this country.
44:49
We want flops to be a stable, accessible resource.
44:51
You shouldn't have to think about which chip set it is, which cluster it is.
44:55
In a sense, I would like to go back to the future where, you know, when I was in grad school at I wasn't smart enough to finish grad school.
45:03
I dropped out before I could even start my PhD in bioinformatics at Stanford Med.
45:05
You know, I did undergrad and then started my master's and then got sidetracked to Kleiner Perkins.
45:09
But it was a simpler time where I could just SSH into a single cluster.
45:13
I didn't even know where it was on campus and just hit run and it would train The model would train.
45:18
And I could focus on what I knew best, which was machine learning.
45:21
And today, the amount of time these independent startup teams have to spend on infrastructure, on compute, on like converting GB300s into H100s into Blackwalls versus Vera Rubins versus AMD versus GPUs, like which chip set?
45:35
I mean, it's just so stupid. It makes no sense.
45:40
And so, we do need to get to a world to to maximize the smartest minds working on frontier capabilities and not like compute.
45:47
It We We do need to We need to have fungibility be solved.
45:52
We need these flops to flow like megawatts.
45:53
Um and then another block is we need American communities to say that's a good thing.
45:58
Like, you know, the the grid and the electric grid undergirds everything we do.
46:02
We don't think about it, but it's like mission-critical infrastructure so that when you like hit the light switch, it just works.
46:09
You know, I grew up in India in a rural school called Rishi Valley, which was started by a philosopher by the name of Jiddu Krishnamurti, like 350 acres of like bird preserve.
46:18
And we had blackouts all the time.
46:18
And that's because there was no grid.
46:20
And so we had to have backup generators and so on.
46:23
Instead, America figured out when we were electrifying the country in the early 1920s, you know, a bunch of smart people were like, "Let's Let's take the learnings of microgrids, standardize AC/DC hookup grids, and force this in as as as like as a national standard."
46:39
And we're in the pre-standardization era of compute right now, and it's resulting in a lot of pain.
46:45
>> No more cloud blackouts.
46:45
That's going to be your mark on >> I think that would be great.
46:49
I think Claude would be much much more liberal with me and my time with him if he didn't have to stress out about his token budget or his his [laughter] GPU budget. For sure.
47:01
>> Anj, we appreciate your time.
47:01
Uh this has been a great conversation.
47:03
We'll We can let you get back to uh whatever the AI is inventing behind you. >> All right. Thanks, guys. Talk to you soon. See you. >> Stay tuned.
47:11
After the break, we're going to recap this convo with Anj and talk about what else is going on.
47:18
>> [music] >> All right, Ellis.
47:19
Well, I think Anj is the modern definition of a Renaissance man.
47:23
That That guy is [laughter] doing a lot.
47:27
>> Prince of Prince of the new guard of Silicon Valley.
47:30
>> It seems like the baton has been properly passed to him from pretty much everyone. >> [laughter] >> Yeah.
47:36
>> Uh I will say I I would like to sit in on the Coachella of it of AI. Is that what it is? >> Yeah. Mhm. >> I wanted to ask.
47:44
We didn't have time, but it's like I wonder what those Stanford kids are even doing these days in the class.
47:49
Do they have their Plaud out?
47:51
Do they have their Claude out?
47:51
Do they have their Omi headband out?
47:53
Do they have friend AI pendant out, uh or some other pin to record all the insights from all the hottest founders, or are they just grown-up iPad kids who are just watching YouTube Kids for the whole lecture?
48:10
>> Yeah, I wish we had would have had time to ask him uh what his rules for AI are in the classroom, if he has any.
48:15
If he just lets people fully embrace it, and I assume the answer is yes, but uh would've been good to ask. >> It's tough, isn't it?
48:25
It's one of those incontrovertible dilemmas that in tech we all want to stay in the moment, but we also want to record everything. >> Yeah.
48:35
>> For our later context. >> Yeah.
48:37
Yeah, I you know, Stanford is a pretty magical place.
48:39
Like, I think it gets its knocks, you know, for being the bubble, definitely, you know, uh being the center of that, but uh I've I've walked around it a few times, and gone in and out of halls, and I you know, it would've been cool to have gone to school there, you know, it it is a very special place.
48:58
>> Has Claude ever told you to stop working? >> Yeah.
49:01
>> When does that even happen? When does that happen?
49:03
>> Um you know, you just you do too much in one day, you get too many co-work threads going, uh you're you're pumping it for answers. >> a tokens thing?
49:15
>> I don't think it's tokens uh based actually.
49:17
Um I think it is just a vibes thing.
49:19
I think it's a an alignment thing.
49:22
[laughter] >> It says uh it says, "Alex, maybe maybe time to call it quits."
49:27
I mean, there is something to that though, not that it's programmed this way, but I often find that, you know, toward the end of the day, you try and keep writing, and you're like, "I'm just going to read this tomorrow morning," and then half of it's [ __ ] and you have to redo it again.
49:41
It's like, it's just kind of pointless, I think, at some hour of the night to keep working. >> Yeah. Yeah. You haven't gotten that?
49:48
It hasn't told you to stop? >> No, I don't think so.
49:53
Um though I've moved a lot of my my Claude usage into Opus inside Notion and into poke.
49:59
And so I feel like I'm using the Claude app a little less, but the model still the same. >> Mhm. >> Is that stupid?
50:08
>> Moving your Opus into poke is something you could only say on this show and me not just laugh.
50:14
>> [snorts] >> Uh >> I also enjoyed the road trip imagery on this episode of the show as long as the road trip goes better than the one in Surat.
50:25
Have you seen that movie yet? >> No.
50:28
>> Pretty crazy art house film.
50:28
I think it was the runner-up to the Palm d'Or behind uh I think it was it was it was just an accident or uh something like that.
50:38
But anyway, yeah, the uh the chill road trip in Surat goes terribly wrong and uh without giving too many spoilers, almost everybody explodes by the end of the movie.
50:50
>> [laughter] [snorts] >> So that's kind of how I was picturing a bunch of like coders in a van on their way to Coachella and uh >> [laughter] >> what what terrible things might might befall them.
51:03
>> Well, we can generate that with Via or something uh or Midjourney.
51:04
Yeah, I think we should maybe get a wagon wheel type song for the show theme this week.
51:12
Something that's got a little bit of twang to it.
51:13
Um I feel like it would be appropriate. >> Yeah.
51:18
One other thing I I liked the tension around if pivots are good or bad from this conversation.
51:23
I mean, the way that I was thinking about it largely was like almost I don't want to say desperation pivots, but like every time you do a pivot, it kind of sets you back to square one to some extent.
51:38
I guess maybe there's a difference between a reboot and a pivot.
51:39
But it just seems very stressful to be a founder right now trying to make something out of nothing and then you find out that now the model just does it automatically, [laughter] and you kind of have to either go deeper along this path without knowing if the model is also going to go deeper into your vertical, or if you just want to choose new topic entirely.
51:58
And each time that happens, you know, you kind of have set your product and your story and your outreach and whatever sales you're doing a lot of times back to square one.
52:06
And it does seem like a tough time to be doing this.
52:12
I mean, it's just faster than anybody realistically keep up with, you know?
52:19
It's like back in the old days, yeah, you're in stealth for a few years, you come out, and then you sure pivot toward the best version of yourself that's resonating, but >> You've talked a lot about that a lot in the recent episodes about how the founders you work with are pivoting all the time like mid mid work stream with you, and that does seem to be the case.
52:39
I mean, when Claude does, you know, Claude design and tries to kill Figma in like a day, right?
52:44
I mean, that kind of stuff feels like it's happening every week.
52:47
It's probably impossible to stay on top of it.
52:51
>> Especially when the new the new expectation, as we've talked about, is that these services are going to work with each other, and any new data you generate inside of one product is also no longer the mode that it once was. For better or for worse. >> Yeah.
53:05
>> I think historically you could think of that there was some stickiness to the usage or the data you'd collected, but now everything is just kind of getting pipeified.
53:13
And making it easier than ever to switch, which I guess is good ostensibly for consumers who are getting a very rapid evolution in their products, but yeah, I wonder if it's chase chase the odds of success for startups or if all the VCs are just going to tell the other VCs to acquire everybody's startups and keep the wheel going.
53:34
>> Well, Andreessen Horowitz doing a lot with startups, and I'm glad we got him to talk about Amp.
53:37
That's a thing that my sources, you know, especially ones that work at the other cloud companies, the AI neo clouds, all these all these this burgeoning ecosystem.
53:45
Uh very curious what he's been up to with that cuz he hasn't said a lot about it and he's raising a ton of money for it.
53:52
And uh glad we got a little bit out of him on that.
53:55
And I think his perspective on data centers and they're needing to be transparency about what they're used for, almost like a nutrition label. I I really like that.
54:02
I think that could go a long way. I doubt we'll get there.
54:06
But if he could push that a little bit even just with uh big but relatively small thing he's doing um with amp, I think that's that's for the better.
54:14
>> I must say, maybe it's just a nerd thing, but I'm very triggered by the word amp.
54:18
Thinking back to Google amp and how it just kind of destroyed and manipulated and transfigured the mobile web.
54:28
Uh and I mean, it's only gotten worse, I feel like.
54:30
You go on pretty much any site these days and >> about amp, yeah.
54:36
>> How could you forget about amp?
54:37
>> Uh I just I don't know why.
54:38
That's It feels like an eternity ago, but you're right.
54:41
I mean, it did have a profound impact on the web and um I guess it still exists, I don't know.
54:45
I don't really read websites anymore.
54:48
I'm all about those newsletters now.
54:50
>> have a question for you.
54:50
I mean, just about like the stuff that tech used to be about versus today.
54:54
I mean, have you found that your interests have changed as well from covering more I don't know.
55:02
It's It's gotten pretty technical, has it not?
55:04
>> No, it has, but I feel like I'm gravitating more towards the culture.
55:08
And when I say culture, I also just mean the personalities and the drama and the the the vibes.
55:15
Um I feel like that's a very intangible thing that AI is not going to be good at replicating anytime soon.
55:23
Like literally right after we're done recording this, I'm going to go to an Anthropic happy hour because they have their big Claude code conference tomorrow.
55:30
We're taping this on a Tuesday.
55:33
Uh and we'll have Mike Krieger from Anthropic on the show um next week uh from there which I'm really excited about.
55:38
And then I'm going to the Open AI 5. 5 party uh for GPT 5.
55:40
5 that Sam is throwing at their headquarters down the street.
55:46
And uh you know, it's a it's a party that 5.
55:48
5 design did all the invites, orchestrated, designed the party.
55:52
Like I'm I'm just going for the vibes to see what this is like.
55:53
And it's also, you know, the metal layer is like Open AI is really trying to capture the zeitgeist again with 5.
56:01
5 and reassert dominance and get the Silicon Valley bubble off the Opus hype train.
56:07
And that's the kind of reporting and like fly on the wall stuff that I think will become increasingly more valuable with AI doing a lot of content generation.
56:17
Uh for the foreseeable future, there's not going to be a a humanoid going to a party and doing a a scene report.
56:24
>> You are uh you are hardly a fly on the wall, my friend. >> That's true. I'm a giraffe.
56:29
It's easy to spot Alex Heath. I'll put it that way. >> That's right.
56:32
>> what I like about these parties though is that you show up and you get to the bar and I don't know who does the drink titles, but it's like, "Oh, perfect. I'll have the GPTini." >> Yes.
56:43
>> With uh extra with extra olives, please.
56:46
>> That's you know, I'm glad you brought that up because you and I need to come up with drink names for our party next week.
56:50
That's on the to-do list with the bar company.
56:52
So, if you're listening to this, send us good ideas for Access show cocktails for our live show and launch party in San Francisco next week at Notion headquarters, which we're super stoked about.
57:03
But yeah, we have to do that.
57:05
We have to come up with some drink names.
57:07
I was going to ask Claude for help.
57:07
I was going to I was going to ask Claude to look through every show Granola transcript that we've done and come up with some good cocktail names.
57:14
And then we can compare to our own ideas.
57:17
>> I'm trying to think of something, but all I'm coming up with is Heath Bar Crunch for dessert.
57:22
>> Oh, that's a good idea.
57:23
>> I'm sure you've heard I'm sure you've heard that one before.
57:25
>> to Oh, we'll have to have hamburgers.
57:27
God, I didn't even think about that.
57:29
>> Jesus, I didn't either. How's that possible?
57:32
>> Yeah, we'll have to do hamburgers.
57:32
I'll have to update the menu with that.
57:34
Uh we'll make sure >> With little toothpicks with a little sticker of my face at the top. >> Yeah. Yeah.
57:42
>> [laughter] >> But that's going to be great. Very excited for that.
57:44
Uh Substack CEO Chris Best joining us um for what will now be I think our It'll be our third or fourth in-person recording and I think our second live show with an audience.
57:52
So uh I'll always always love doing that with you. So that'll be great.
57:56
But yeah >> You've talked to Chris a few times in the past.
58:00
Do you have any questions left?
58:02
>> I do, but I can't share it on the pod.
58:02
I mean he'll listen and his team will listen and then he'll be too prepared.
58:08
Unlike Ange who lets his guest at Stanford have edit control over his the videos.
58:12
Uh we don't do that here at Access.
58:14
So uh got to keep it fresh and interesting, you know. >> Okay. I look forward to that.
58:17
I mean is he also kind of your boss?
58:22
>> [laughter] >> We're going to talk about that.
58:23
Uh in a way yes and he would say no.
58:25
Uh so we're definitely going to talk about that.
58:30
Uh he's >> they're they're very very much on the free speech train.
58:33
So I'm sure there's not a a boss relationship at the moment, but uh >> At the moment. >> I don't know.
58:40
Maybe you get a text late night and your signal chat says Alex, I saw open rate is down this week, man.
58:44
We got to We got to We got to get that back up, buddy. All right. Good night.
58:51
>> I can't wait for [laughter] your questions to him about my open rate. I'm preparing for that. It's good. It's good. Okay.
58:58
Well, [gasps] should we wrap it there this week?
59:00
I feel like we we have a full show.
59:03
>> That is it for this week's show.
59:06
Uh thank you to Ange for coming on. >> Yes. Thank you, Ange.
59:10
>> And if you like the show, don't forget to like and subscribe everywhere you get podcasts. >> We are access.
59:14
show online and you can find us in video at accesspod on YouTube.
59:19
>> If you really like this episode, we would just love a five-star review or even share with a friend.
59:24
That was Tim Rogers inspired if we have any fans out there.
59:29
>> [laughter] >> Help us chart, please.
59:29
Uh you can find my newsletter at sources. news.
59:32
Help me keep those open rates high.
59:35
>> You can find me at hamburger on Twitter and at meaning.
59:37
company for your startup storytelling needs.
59:41
>> Access is part of the Vox Media Podcast Network and this show is produced by Hooked Creators. >> Bye. >> Bye. >> [music]