The Professor of Outputmaxxing — Anjney Midha, AMP

so there's two types of utilization usually, right, that you're measuring in these clusters. One is node allocation and then the other is MFU. So node utilization is usually what percentage of cards in the data center are just used.

And that, if it's not at 95% There's no excuse. There's no excuse, right? Like I think 95% of Google, which is where my co founder Seb came from, he built the BorgX BorgGQM scheduler at Google.

And there, I think 95% was considered an outage. So 96% node utilization should be standard. And most single tenant clusters are not running at that.

So that's one. And then MFU utilization should be, I would say, the best in class today is somewhere between 6070%. I think this is a leadership question, right?

Fundamentally, it's an alignment question, which is, are the people who are funding the cluster and then deploying the cluster actually aligned? And sometimes, theoretically they are, but in practice the number of people in the chain, the supply chain between the capital and all the way to whoever's managing the cluster and whoever's measuring what the output is are just so many degrees of separation away know, that you heard that sort of, you know, radian metaphor, which is at the beginning of an arc, if you have two arcs that are two lines that are just off by a few degrees, that it spreads out, right, at scale. And I think what's happening is a lot of cluster implementations and infrastructure, a lot of frontier labs and other teams, that's what's happening is they initialize the plan, which is kind of like north star with a team that wants to do good, but then they're required to scale so fast instead of iteratively that the wastage just compounds really fast at scale.

And so I think we know the answer, which is just do iterative bring ups. Know, If you spend time with people who've been in the semiconductor industry or the data center industry for a long time, this is not new. And I don't think AI should be an excuse.

Like, sure, something What is new? Okay, we have a lot of new capabilities, but that doesn't mean just abandon common sense. Common sense should always be in fashion.

AI scaling doesn't change the In fact, if anything, AI scaling should be putting a premium on the value of common sense and infrastructure because the margin of error now is so much lower and the costs of wastage are so much higher. And the cost of wastage, by the way, not just economic. Obviously, I'm an investor.

Or I'm an investor by background over the last few years. Now we're running an AI infrastructure business called AMP. And I think that it's okay to say this time is different on the capabilities front.

Like, are genuinely getting capabilities of a kind we haven't had before. That doesn't give you an excuse to say this time is different for everything, especially infrastructure. So look, I love the hacker mindset and the hustler mindset.

Now that's great for the startup mindset. But you remember this moment where Zazak went from saying move fast, break things to move fast I with stable think now we need to move fast with responsible infrastructure. They're going to say, where is the impact?

Was a really in our class yesterday, Scott Nolan, who is the founder of General Matter, came by at Stanford to speak about energy bottlenecks. And he had a phenomenal idea. He said, if you look at the marginal unit economics of compute per hour, let's call it like $4 an hour.

If you're having to bring up a new data center in a new community, why not just say we're gonna charge $4.50 an hour and that marginal impact or that marginal increase, we just literally take that and give it to the local community as cash. I can tell you as a customer of that compute, I would love that.

I'd be happy to pay an additional $0.50 per hour at scale. Because if that means the public benefit is so clear to the communities that the data centers are coming up in, I'm going feel like that compute is much more reliable.

Up to 20% of all data centers this year in The US, my understanding is are at risk. Of community backlash? Correct.

Of not getting the community support they need to get brought up. Wow. That's a huge number.

Yeah. Now I think we should dig into what that number is. I think it's a little bit of overstated.

These things can get over reported.

If we're bringing up a data center in your community, we're actually going to reduce the cost of your electricity bill. Okay, now we're talking. Right?

The community is going, okay, now this is a deal. I feel like a partner in this. Yeah.

Right now that's not happening. There will be audits. There will be investigations.

And when regulators come, I don't know when it's going to be, The folks who are moving fast and breaking things in the name of AI progress better be prepared. That's certainly not how we're procuring compute. We're trying as much as we can to work with partners who have long term track records.

Many of whom, by the way, are not like AI providers. I think this whole idea of Neo Clouds being somehow this new category is a lot of marketing speak. There are really good, reliable, trusted data center providers in America who've been around twenty plus years.

I love those folks. They know how to sure. Are they sponsoring happy hours at NeurIPS?

No. Are they legibly bitter less and billed? No.

Are they hanging out in situationally aware parties? No. But they're adults.

I trust them. They can run land. They can run run They credit histories.

We sit down, we have a conversation. Many of them live in Silicon Valley. They've had to deal with the boom and bust cycles of the internet.

And I love those folks. They are stable infrastructure partners and thinkers. And I think there's a lot of short term thinking going on in the compute layer.

And it's going to catch up to us. It's not going to be good. You talk about aligning incentives.

two regimes of architecture, right? You have integration, and then you have pooling and utilization. Right?

Or rather, the way to increase utilization often is you can do systems integration where you collapse a lot of process into one node, or you can pull out a process from a node and share that amongst various that resource amongst several different nodes. And so we see the AMP grid, which is what what the system we're building here, which is basically a compute grid. You know, we're doing trying to do for compute what the electric grid yeah.

What the power grid did for electricity. This is a pooling and utilization layer across clouds. And so we're actually the opposite of a full stack integration.

It's much more horizontal. And it's multi cloud, it's multi silicon. The goal is to try to make flops flow like megawatts.

And that is very hard to do today for many reasons. Like there's stranded pools of compute all over the place and there's no fungibility. And so right now we do it at the level of scheduling and we often do it at the economic layer.

But as we start to announce what we're working on, it's extraordinary how many folks are coming out of Woodworks and saying, hey, I'm actually working on a way to make compute fungible at this part of the stack and that part of the stack. And as a grid, we'd like all of these folks to participate on the grid. People often ask me, Andre, you're a new cloud?

I go, no, actually, new clouds are suppliers. Sometimes they'll ask, are you a venture capital firm? I go, no, actually, are demand of off takers of the grid.

We see ourselves as what's called an independent system operator. So if you study the history of the electric grid, once it became legible to a lot of factories and industrial participants that, hey, actually it turns out pooling is a good idea. We should pool our generators instead of all having a generator running at half capacity in our backyard.

There was a need for an independent entity who could coordinate all these parties, power generation facilities, transmission lines, factories. And that neutral coordination mechanism is very critical. If you study the history of grids, the most enduring ones were those that never owned their own assets.

They were ones that had often started with long term anchors who were uncorrelated sources of demand, a steel factory or shoe mill or whatever in a particular town who weren't competitive, where the steel factory want to spike up at night, the shoe mill want to spike up during the day. So then you pool and you share, right? So each of you is guaranteed some baseload, but then you kind of schedule your spikes to drive a peak utilization across the town.

The gold standard, so to speak, historically has been these utility companies like PJM Interconnect in the Northeast Of America, where they, over many, many years, became this what was called an ISO, an independent system operator of the grid. So that's how we see ourselves. Economically, that's what we are.

From a technical perspective, we started at the scheduling layer because Seb and Mihai who run engineering here built They that that at Google. Yeah.

so I was running the developer platform there. The internal infrastructure, was not responsible for. That was actually a guy by the name of Mark Smith, who was extraordinary.

And yes, Discord did pull so Discord is actually a counterexample. I guess I had the chance to learn a lot about fully full stack infer there because same thing. Yeah.

It's the it's the other architecture, which is Discord built its own WebRTC voice and video infra. So Discord did not use For the calls, yeah. Yeah, for communication, Discord did not use third party infra.

It was all built in house. And then the way you maximize utilization was you pull demand from the world's 200,000,000 plus monthly active gamers, right? And so that's how those stacks were constructed.

Again, in systems design, the two concepts that keep coming up over and over again are abstraction and composition, Bundling and unbundling. Bundling and unbundling, abstraction composition, verticalization and horizontalization. So in that sense AMP is an independent system operator of the grid.

We pool supply from a number of partners we trust at about 1.3 gigawatt scale over four years. And then we pool demand from some of the world's best research labs and so on.

We're sitting at one periodic labs who need extraordinary long term demand. And the idea is that each of them is guaranteed baseload on the grid, but they can spike up and down flexibly for compute with much shorter timelines as needed. That was roughly the design of the program I came up with at E16z called Oxygen.

That was the same design of the GQM, BorgExBorg GQM implementation at Google that Meehan and Seb had built, which is that how do you allow teams inside of Google on the internal infrastructure to be guaranteed capacity for their base workloads. But when they need to spike up on research, how could they ensure that that was sufficiently there? And of course, the big innovation that was discovered not discovered, but kind of implemented in this space, this infra space maybe three, four years ago at Google was the idea of interruptible demand, right, where you just queue up a bunch of jobs.

And through this sort of credit system, there can be a bidding mechanism. Priorities. It's prioritization, And jobs can get interrupted based on somebody else who's saying, you know what?

I have 10 tokens, 10 credits I want to spend on this job. Another team lead, research lead is like Genie three or whatever is only worth five credits and Nano Banana two is worth 10 credits. And so the Nano Banana job gets priority.

Like we structured ourselves essentially very similarly to Google. We are structured as a holdings company. So Alphabet Holdings is the Alphabet Holdings and then they've got these subsidiaries called Google and Other Bets.

Other Bets and so on. We've got AMP Holdings and we've got our infrastructure business. And then we've got a capital business called Foundry that incubates new Frontier AI labs and invests in them as venture capital.

Like Periodic, we put a few $100,000,000 into Anthropic from our fund earlier this year. So, wherever we feel like teams are making progress, especially researchers and so on who push the frontier inside of existing labs like DeepMind, I find there comes a point where they feel misaligned with the dictatorship of alphabet holdings. And at that point, sometimes the dictatorship doesn't want them anymore.

And they're like, thank you. You've done your job here. You've kind of helped us through the zero to one phase.

And for whatever reason, we're to deprioritize your amazing like omni model or whatever it is. And instead, we're going to prioritize coding. And I think that's a tragedy.

But I get it. Sergei and team are running their own business there. But that doesn't mean the rest of us should sit around waiting for that progress to get unlocked for the rest of the world and humanity.

I mean, you think about how much extraordinary research has happened inside of DeepMind over the last ten years. Mean, Demis and Sergei and those guys did such a great job. But at the end of the day, so much of that has never seen the light of day.

They're like papers only, but they never actually shifted to production or I mean, what's worse is the paper is actually not even being published anymore because there's a six month embargo inside of DeepMind, right?

Again, I think it's a tragedy. I mean, I get it. They're running their business, but the rest of the I think there's negative externalities of research being hoarded.

And so there's a market failure, and somebody needs to unlock that research. And we can't do it on our own. We only have 1.

2 gigawatts of compute. That's nothing. That's about $40,000,000,000 of cloud spend.

We're going to need a little That's a new number. I haven't come across that gigawatt number. That's huge.

Yeah. And to be clear, haven't secured all of it. That's how much demand we have started to secure.

I think publicly we haven't actually confirmed how much we have for this year. Where do want to get to? I think the steady state would be that we have a baseload pool of 1.

2 gigawatts at all times of baseload capacity. For spike capacity, right now, my estimate is we need roughly six gigawatts over the next four years for all our teams to feel like they were able to keep moving the frontier, whatever they're working on, whether it's like superconductor discovery over here. There's a new investment we're working on right now, which is in the end of life prediction space in health care.

It's extraordinary how much give peep you know, this was actually my graduate school work. I went to grad school for bioinformatics at Stanford Med. Yes.

I know we Econ, MCS, Bio. So my I was this really weird cat where like I was never satisfied with my major options. At So one point I was an econ major, then I was a CS major, then I was a MCS major called mathematical computational science.

And they decided they were gonna end that major. So I took all that coursework and I applied it to grad school, my graduate degree in bioinformatics, which was the master's program. And then I thought I was gonna do a PhD.

Never ended up doing it. I dropped out and went to work at Kleiner. But I was lucky enough to apprentice with this professor at Stanford Med.

His name is Nigam Shah. And he was working on end of life prediction. Stanford is one of the only research facilities in America that has a longitudinal patient data set that's larger at scale.

I think it's at least 12,000,000 patient lives. The only larger data set is the VA, the Veterans Affairs of America. And to do research, like do any deep learning and so on that dataset, it was called the STRIDE dataset at that time, you had to be a Stanford Med School affiliate, which is why I went and enrolled in the bioinformatics department.

Wow. Deep learning was early. Nigam Shah had the vision to see that you could do end of life prediction to help palliative care.

In America, over 30% of all Medicare, Medicaid spend, at least at that time, was spent on end of life care. And we grew up in Asia. At least I won't speak for you, but I have a very different relationship with death than I find folks who grew up in America do.

In America, spiritually and culturally, especially in Western societies where Christianity, the Christian tradition sort of frames death as this terminal point. There's often a judgment day and so on. The way we view death is with a finality.

In Indian culture, in Hindu culture, death is one Buddhist as well. You're Buddhist, yeah. So it's one step in a journey of many lives, right?

And so I grew up in this city called Chennai in the South Of India. And when people die, you dance on the street. You know, there's like a procession where your body is carried to be cremated, and your family celebrates.

And there's drums and so on. It's this huge thing. And it's because the idea is that you're going to be reincarnated.

You've been liberated liberated from the responsibilities of this life, and now you're onto your next. It's a new adventure. It's like going off to a new college or whatever, right?

And so it was so alien to me when I got here as an undergrad that the medical system works backwards from that assumption that we have to view death as this terminal thing and delay it, postpone it. It's a bad thing. And so at the time, decision support in The United States was this very primitive field.

Even to this day, physicians in The United States often will tell you when you have a terminal disease, this is your we've diagnosed you, which is great. Our ability to diagnose you is extraordinary. You have somewhere between six months to six years to live.

What do you do with that information? The error bars are so high that then you in times of uncertainty, we default to culture. And when the culture is, this is a bad thing, I've got to prolong my life, then you start doing things like just to sort of from a systems perspective, what's going on there is physicians often feel like they need to provide such high error bars because there's always some uncertainty in end of life diagnosis.

And if you provide the wrong diagnosis or recommendation to your patient, you can be sued for medical malpractice. And then your license can be taken away. It can be catastrophic for your career.

In contrast, in countries where that's not the case, what you often observe is that patients like physicians are quite prescriptive with their recommendation. They say, hey, this is your condition. The literature says that you probably have this much time on earth left.

My expert opinion is that you are an outlier or whatever. And they try to be more prescriptive. And that empowers a patient, right?

Because that patient can say, I trust my doctor. They said, on average, I have six months to live. But if I do these things, I may have a shot because of my particular predispositions or my genetic history or whatever.

And that empowers you to go about your life in actually a more scientific way than leaning on religion, culture, spirituality, and so on. In contrast, here, because of that medical malpractice sort of thing looming over your head, a physician never gives you a clear recommendation. So instead, say, Okay, doc, well, let's try it all.

And then you start a whole regime of drugs and therapies, and then you often spend weeks and weeks in the hospital, and that deteriorates your quality of life. And when that deteriorates your quality of life, instead of spending your last few days doing the things you love with your family, you're spending on a hospital bed. And that ends up being 30% of Medicare and Medicaid.

So it's worse for the patients. The doctors feel terrible. The American taxpayer is paying a huge amount of money.

And so this is why Nigam Shah, who was this professor at Stanford, said, Anjith, there's I kind of sat down with him. I was this young eyed. I was 21.

And I was like, I want to work on a big problem. And he's like, the big problem is end of life care. And so we tried to do deep learning to say so we start trying to run deep learning on these tried patient data sets to say, could you have an AI system make a recommendation that is orders of magnitude more precise about how much time you have left once you've been diagnosed with a terminal condition than a human.

And then if we can get that precision to be high enough, then you can empower the patient. And it turns out the tech works. Once you get the data set, RL works, honestly, even regression models work.

You don't need get that fancy. At the time, we were just doing very simple neural nets. Today, what we can do with RL is extraordinary.

The problem remains then and now is regulatory. Because you actually can't shift the burden of the wrong clinical diagnoses from the physician to the AI system. And so at that time, I got quite disillusioned ten years ago, twelve years ago, because I felt I just didn't have the resources to influence regulation.

Today I'm very lucky, I'm in a different place. I'm a lot older, so I've been spending a lot of time on my next incubation, which is how can we unlock patient empowerment by training AI models through end of life prediction with much more precision. And you're still focused on this I whole haven't been able to get this out of my mind a single day for the last fourteen years.

This is the hill I would like to die on. There's two, I would say. You know what?

Actually, prefer not to die. Yeah, exactly. But I think bipartisan issues I think two issues that should be bipartisan in America are how do we empower patients to make the right clinical decisions at the end of their life such that we're reducing the taxpayer burden with science?

It's just good old science, and AI can help here. And the second is net positive data centers. Because I think that's the biggest critical bottleneck on training good enough AI models to help people at the end of their life.

So there's sort of two sides of the same scaling bottleneck curve. But those two we formed AMP as a public benefit corporation. My wife and I, who you've met, you you've met Viv.

Yeah. Her passion is is education. You know, her family is a long line of educators and so on and of physicists.

And so this this class is my attempt to stopping the black sheep of the family and be an educator. But if I'm not educating, the thing I would be doing is working on these two problems, whether on the political spectrum or as a researcher back in some lab. And my hope is, if anyone's listening to this podcast, if they're passionate about either of those two topics, I'd love to hear from them.

We can share the contact in the show notes. But we're looking for people to join both of those missions on the political side as well as on the medical side, on the research side.

It's the department of output maxing. Make the most of what we have. Exactly.

I'm a huge believer in optimal outcomes. I think both in America and other countries, we are losing our appreciation for nuance. This is the thing of AI is the same case.

Oh, the bitter lesson holds. Okay. Fine.

But that doesn't mean you just, like, throw 500 GB 300 500,000 GB three hundreds at your, like, suboptimal model scaling and you waste a bunch of compute. It also doesn't mean that the most optimals have like 50 different architectures where there isn't enough standardization. Like one of the reasons Anthropic has had extraordinary sort of velocity is because they picked the transform architecture and said this is simple, let's double down on it, right?

And now luckily, there's enough investment going into space that we can afford other architectures. But at the time, investment was just too fragmented into other architectures. So that arguably unlocked scaling.

So I think there's a philosophy. I think we all owe it to ourselves to do output maxing with a new capability called AI on a global level. I think if I was starting a new department at Stanford, depending on how fuzzy or technical I wanted to be, I'd probably call it the department of alignment.

really is a hard problem. And I think when you unlock it, full stack alignment is super hard in any organization, in any system, like in a venture capital firm. If you can have full stack alignment between your limited partners and the founders who are creating the value and ultimately the public that owns the IPO stock, that is a gift that keeps giving.

And when you study the history of these systems, when they start off, they usually start out small scale where the feedback loop is actually so tight that there's alignment. Yeah. And then the more you try to scale, the more division of labor happens, the more specialization happens.

And at each step, you add abstractions. And wherever there's an API interface, there's, like, loss. There's communication loss.

And so I think a really cool thing would be for us to figure out, is there a way for us to have our cake and eat it too as an engineering discipline? Is there a way to actually scale up and scale out without losing any alignment, without, you know, lossy transmission. You mean standards?

So standards is one way. The other way is you just have net new capabilities. So like soup you know, what we're trying to do here is discover new superconductors.

A room temperature superconductor would be a lossless transmission mechanism for energy. I mean, we would have flying cars. Yeah.

We are right within a few years of having a new room temperature superconductor. So I think those are two.

SF Compute is the kind of effort that I hope we can accelerate because what often happens is these exchanges are very hard to get It's hard to bootstrap them, right? Because they often require There's many inefficiencies between parties. There's trust boundary inefficiencies and infrastructure because you don't trust One part of the stack doesn't trust another part of stack to give them visibility.

There's capital markets inefficiencies. There's operational efficiencies. So if you can inject a single shock to the system of a ton of compute, demand, or supply, then you can accelerate these new flywheels.

And so my hope is one day, or soon, if SF Compute needs extra capacity, like has excess capacity, they just hook it up to the grid and they get flooded with demand from us. And on the other side, if they have a ton of demand but they don't have supply, they just, again, hook up to the grid and it's a two way protocol where they can just hook up to our capacity. And I don't think we're too far from that.

Today, our working implementation of it is mostly through a group of labs, universities, and a few sort of trusted parties who all feel like they're in alignment to borrow an over sort of used word. But our hope is to just have it be an open protocol that anyone can hook up to on Hook up for demand or hook up for supply?

what's happened in the last six weeks is we thought we'd have a bunch of excess capacity by the end of this year. It's all gone. It's exploding.

Yeah, it's all gone. And so I have my text messages are full of friends. I mean, we know many of these people.

These are founders who've raised billions of dollars in San Francisco going on. Any chance you have, like, 50 nodes in the next few weeks?

I don't think so. So actually, Rainer's a great example. Rainer's the CEO and founder of MatEx.

I actually had him buy for office hours in the class earlier today. And there was an insight he brought up that I hadn't considered before, which is when they decided to pick the standard for their data center, they picked the NVIDIA reference architecture. So the MATx chips just plug in to any site that has an NVIDIA bring up plan.

Well, from an input and out IO perspective, it's the same footprint as an NVIDIA rack. Where they have done innovated a bunch from what I can tell is on systems co design, which is where a lot of the gains are to be had. And so he picked he was like, Anj, there's just so much work to do when you're building a new chip company.

Can't fight on every front. You just can't fight on every front. So my question to him was, well, you're working on this new chip.

Their tape out is next year. Who are you going to partner with to host the chips? And he said, whoever will host them, that's not my focus.

And I said, but how did you back to earlier systems design question, he decided that he didn't want to be a fully integrated chip provider. The bottleneck they're focused on is the logic die. And he feels they can crank out a ton of performance gains to co design there.

But then that means you delegate, to our question earlier, he's like, data center provider is a different part of the stack. And so then he's dependent on that part of the ecosystem to host his chips to get the performance gains to the customer. So now you have another abstraction, and you might have loss.

So I asked him, how do you prevent loss? And back to your point, he said, just picked the NVIDIA standard. Because I didn't want to I wanted to piggyback off of an existing protocol.

And what's great about NVIDIA is that reference architecture is known. It's open. They've published it.

So Jensen's actually enabled someone like Rainer to build a chip company like Matix. I don't see them as competitive. The compute demand is so high.

Like, I don't I think NVIDIA is not able to meet the demands of production. So we just need more chips. And I think it's very smart what Matix has done, which is say we're just gonna we're not gonna innovate on the data center design because actually, thank you, Jensen, you've done all the hard work.

Where we can innovate is somewhere else. And I think that's very healthy. I think that's how we unblock new bottlenecks.

And my view is these chip teams like Matix who have arrived at the insight that co design is the way, the primary bottleneck for them is trust boundary. To do co design well, you need visibility into the next model generation as soon as possible because it takes two years to tape out. So if by the time I bring my chip to market, your model architecture has changed, I'm host.

Now when he was inside Google, he was sitting next to the Gemini team. He was on Palm or whatever. His co founder was one of the Palm guys, I think.

Yes. Yes. Exactly.

So when you're inside the trust boundary of Google, then your systems co design loop is super tight. When you leave as a founder, one of the biggest risks you take is now you're outside the trust boundary. And so what I love doing is helping chip teams who can help us unlock more capacity for the independent ecosystem access to trust.

Because if I've been involved with a lab from day one, and I was lucky enough to work with Anthropic, and then I'm on the board of Mistral and Black Forest Labs get started. I think at this point, I'm on six or seven different teams.

five month gig where Weilin and Anastasios were graduating from their PhDs, and they didn't need a product team. So I helped recruit the head of engineering product and design. But Anastasios has always been the CEO of that company.

I played a pinch hitting job. I'm an intern. I was CEO intern for five months.

but also very quantitative and mathematical, is such a unicorn. See, you know what's amazing about him? If you look at his output, he's an output maxer.

Like, by the time he was graduating from his PhD, which he only graduated last year, he had published more work with a citation count than people twice his age. But at the same time, he'd already started a project called LM Arena that was being used by millions of people as a side project. And time and time again, I've realized is venture capitalists suck at seeing human beings as like dynamic agents where They want to put you in a box.

They want to put in a thing. So the first time I got introduced to Anastasios, somebody had told me like, oh, he's amazing, but you know, he's a researcher. Yeah.

I was like, what? What what do you mean he's a researcher? That that is what Not a CEO.

Not a founder. Not a CEO. Exactly.

I was like, are are you crazy? Do have you met Dario? Dario is a scientist.

He's gone from zero to what will soon be a trillion dollar company in four years. Being a CEO, nominally speaking, is not that hard. Being a good CEO is hard.

Being a great CEO actually requires a level of performance that scientists who have already published at the top of their field have accomplished. It is super hard to be a competitive scientist, to publish in academia over the last twenty, thirty years, to make it at top of your discipline at a place like Berkeley, you are a star athlete. You are an athlete of the mind.

And you perform at the highest levels. And to get there, whether you're Anastasia or Weyland at Berkeley, or you are Robin who BSL here. With Black Forest and created stable diffusion, or if you're Guillaume at Meta who created LAMA before he started Mistral, like, the amount of human leadership you have to demonstrate to get the resources, like get the trust of the organization, publish it, put it up.

I mean, I would just fund researchers all day, right, who have contributed already to the field. If they put SODA out there, star athletes already. If they haven't done SODA, look, they can still be good CEOs, but then I find the failure mode is that they just don't want to be CEOs.

They primarily want to publish, and that's okay too. One of the things we do with the AmpGrid is we donate excess compute we have to nonprofits like university labs. We carved out like a couple thousand H100s.

But I do think there's extraordinary research being done on university campuses. My father-in-law is a physicist. He's a professor.

Extraordinary work in physics. And we need that. But if you want be a CEO, what you need to be willing to do is be super confrontational outside of science.

Within the scientific community, some of the best researchers are very confrontational about their convictions. This architecture is right. To be a great CEO, you basically have to be willing to be confrontational up and down the stack.

To your own team. To your own team, customers. Well, I would say, yeah, pretty much to everyone.

Stakes are personal scaling vectors, right? Like the stakes that seem so low to you, like having this podcast where you can talk to somebody and just have a I mean, you're an extraordinary communicator, right? Like already in this conversation, you've pulled more out of me than most people.

You know? And I've been on 12 podcasts in the last two weeks. I think we've just seen each other enough that there's some base trust.

There's some trust. And I know that you you know that I've done my homework. And like, I I know that trust is a big deal for you.

So Right. Yes. I think trust is about consistency.

And you and I've seen each other in the community for years, right? Like I remember the first time we met was at NeurIPS and New Orleans. I don't know if you remember that Oh my god.

Reiko had set up this you know, Reiko's amazing and he set up this luncheon. Yeah. I was like, who's this Discord guy?

I'm like, okay. No. Weren't no.

Made some investment. You were much less polite. You were like, who's this VC?

but then I I think I I heard your accent. Yeah. And I was like, are you Singapore.

Are you Singaporean? You're like, yeah. And I said, went to high school JC in Singapore.

And then the ice broke. Okay. Right?

Yeah. Yeah. Yeah.

But this is the you know, there there are in the scientific community, sometimes the stakes are very high for people who haven't had the emotional we know what is called EQ, coaching and mentorship. Right? Which is like, to have scientific impact, you often need to be an extraordinary emotional emotionally in tune person with the folks you're trying to influence.

And so what comes so naturally to you is actually a super high stakes thing to other people. And so I wouldn't assume that Dario is more stressed out than you. Know?

These these things are like, you'll be surprised how similar and small sometimes the problems are to you that some of the world's biggest leaders are facing. And that's what I've learned from this class. The guest speakers are Sam, Satya, Jensen AI Coachella.

Yeah, it's AI Coachella, right? So we've to get all the headliners. And I'm very lucky that some of these people have either mentored me over the years or I've done business with them.

And when you take the performative stuff out and any assumptions you may have about these people that you read in the press or on Twitter, And we're all just humans here, all trying to get along. And what's so special about this moment is AI is forcing like scaling, the bitter lesson, is forcing a lot of people to revise their assumptions for how the world works and go back to first principles, or go and educate themselves. The kind of people I won't name who this person is, but I was at an event last week in Texas and ran to somebody who said, Ang, I came across the class.

What do you think about real time action prediction models? And I don't know how happy it made me feel when they asked me that question. I know they've done the work.

They've challenges. Didn't ask me, what do you think of world models? They said, what do you think of real time action prediction models?

World models, don't get me wrong, are cool and everything. But you and I both know that that is a layer of abstraction that is sometimes not usefully precise enough. Yeah.

Right? There's like four different kinds of world models. Exactly.

We've done the part with general intuition, by the way, which is very focused on Oh, cool. Yes. I love PIM.

PIM is great. And this is what I love about people who've done that level of work. They realize they're not in competition with people who the rest of the world thinks they're in competition with.

Yeah. Because they're not in the category. They're in the specific thing they're trying to do.

They're focused on their mission, and they have a systems understanding of the bottleneck they're trying to, like, solve. And when somebody else says, I'm working on real time action prediction models too, Tim goes, oh, I love that person. I want I can learn from them.

But the minute they're like, oh, that person's a world model person, it's like, ugh. Like, which type of world model person? But mostly, they're just trying to figure out if it's a waste of their time because we don't have enough time.

So PIM, for example, is super loves this other company I work with we've talked about called Black Forest Labs. And he's mentioned to me multiple times that he's so he thinks what Flux is doing is really cool. Andy Blattman came by and spoke in the class.

And what I find over and over again is for people who do the work, who can be usefully precise enough about what is actually going on in the world of frontier research, the sense of camaraderie is still well and alive. But it gets lost sometimes when you have to abstract the technical complexities in business terms. And then the VCs are like, are you different from that world model?

Come and say, where do I even start to explain this stuff?

idea of winning is so

No. I think you wanna lead. You want soda?

No. I think you wanna lead. Yes.

So you you wanna push the frontier. You wanna push the state of the art. You wanna do something that hasn't been done before.

You wanna capture value. We don't wanna capture so much value that people think you're unaligned with your mission or trying to do what's best for the world. You wanna capture enough value that you can keep innovating.

Right? And I think that people want to lead. This idea of winning and losing again, I love Jensen.

He's a leader. The mindset that he talked about on Dwarkish's podcast, he was like, I didn't wake up with a loser mindset. I think that was awesome.

Right? Because he's an engineer. Dwarkish has done the work.

So there's at least even though to me it was very obvious they're talking about the same thing just past each other. Basically Jensen has this five layer cake abstraction of how the industry works. And Durakesh had I I think from that podcast had more of like a pre training, mid training, post training systems loop concept.

It's just a factor of who he talks to. Right? Again, it's very clear.

The It's the abstraction, the mental models, the it's the whole dude, so much of the problem in the world is reasoning by analogy. And then the assumptions that are held invisibly.

And the venture capital community is, like, notorious for this where people look in times of uncertainty, they, like, cling to axioms that ended up being true from the previous era. And they they kind of, like, proclaim them with confidence as if they're truths, but they're not. And it's very important to see the distinction between a heuristic and an axiom.

Yeah.

Heuristic is a way you a shortcut. And my god, the number of people I have had to put up with over the last few years who proclaim, like use heuristics as axioms to judge people, to judge which companies are going to succeed. I mean, number of people were like, oh, yeah, yeah, yeah, Anthropic, they're just training models right now.

But this won't continue. Things like to B2B SaaS? Yeah.

Which which over the fullness of time, if you squint at it, maybe. But the way you arrive there is so important that you can you just you can dismiss people. Here's what happened, right?

What happened is Anthropic basically achieved takeoff in October of last year. That training run Whatever, three seven? I forget the numbers now.

But whatever that checkpoint was We saw the cognition. Yeah. Right?

You probably to those of us in the community, especially once post training was done and it was released in December. Yeah. Can I sneak a sneaky question in there?

I don't know if you have a perspective, maybe you don't. I just the the number one question is how did Anthropy correct coding?

I had this very annoying teacher. Yeah. I I went to this boarding school called Rishi Valley in India, which is like this bird preserve.

It's like 350 acres of of bird preserve in in rural India. And there was no technology for seven years. There was this teacher, I won't name them, but they would have this I hated it every time he said this to me.

He was like, luck favors the prepared mind, which is like a common saying, but the way he delivered it, like always great me because he was always try. Like I was always one of those kids who got like a good grade without trying very hard. Because like high school, you know, middle school is not that hard if you're generally like paying attention and so on.

And there was this one time where I but then I would get an 80% grade. And he would keep pushing me to say, like, the reason you didn't get the 95 plus percent is because you're not that lucky. And I would say, do you mean?

Because I I would think that I deserved that grade, and I would sometimes argue with him. And he'd say, you didn't have a prepared mind. If you want to get lucky again there was basically one time where I got like 95 or 96 on this subject.

And now that I felt entitled, I was like, I'm gonna keep doing this. And I didn't. And then he was like, luck favors favors a prepared man.

You got lucky last time, but you gotta stay prepared. And I didn't understand what he meant. Now as I'm older, I'm like, okay, these adults actually knew a thing or two.

Anthropic has been the most prepared company for four years. And so then when the right context data comes in, the right developers start sending in the right context diffs, sure, you could say you got lucky. But if you ask me, they're pretty damn prepared with paranoia for like four years.

And if you remember, it was so hard for them to get going early on that they had to do so much more with so much less that you just have to be prepared to be so efficient.

Yeah. But it's so clear. Right?

Like, how to output Macs for the world. They have been prepared.

You're saying culture is the ultimate moat? Yeah. But the thing about culture is it's very fragile.

So moats, I I I don't think that there's very few moats I found that are actually moats. There it's it's a nice concept, but in reality, you have to replenish your culture. You know, Ben Horowitz was the speaker in CS one hundred fifty three on Tuesday.

And I asked him this question about the culture bottleneck in Teams. Because, know, there are several AI teams book, like, Hard Things About Hard Things. But more concretely, there are so many AI labs today that have all the cash they need.

They have all the compute they need. And they're still not able to ship anything soda. And then you start seeing people leave and so on.

And my diagnosis is it's the culture. And so I asked him, Ben, you know, there he's been one of the most aggressive investors in AI Labs. Goes back to this thing which resonates in my mind a lot.

When I used to work at A16z, I would book a conference room. And right outside the conference room, which is closest to the toilet because it was the fastest way for me to go use the bathroom between Zoom meetings. Oh my god.

I'll put Maxine my toilet optimization. Okay. Never mind.

It was not healthy in hindsight, but maybe this is TMI. But anyway, outside that conference on the wall was this quote that was printed that said, culture is not a set of beliefs. It's a set of actions.

And it's by Bushido, who's a Japanese philosopher. And if you stop taking the actions that demonstrate the mission alignment to what you've said to your team and to the world matters to you, then your culture starts to fray. So it's not actually a moat, I would say.

It's a very, very brittle, fragile thing that requires daily tending to like a garden. But if you figure out the system to keep that garden tended, which I think ultimately comes down to knowing yourself because you'll most naturally, if you're authentic and so on, you'll naturally make trade offs that seem effortless to you, but that reinforce your culture. And then that becomes this very hard thing for other people to catch up to.

And at Anthropic, from day one, you know, there was this mission missionary like zeal and belief that, hey, these capabilities will scale. These systems are stochastic, not deterministic. There will be error bars.

And until we crack interpretability, there's risk. And at some point, people will stop using Claude just for coding. They'll use it in some mission critical context where there's it'll throw off a bug, and then people are gonna come blame them.

And they wanna be on the right side of history where they said, yes, this is a powerful technology. We think it's going to change the world. And we want to be very measured and scientific about the fact that, hey, these are statistical models.

That's how statistics works. Ultimately, when you're training neural nets, it is just a statistical system. And I think that that belief that safety is important and that it might seem toy like in the early days.

And sometimes you could say, Anj, they totally over exaggerated the risk like two years ago when they said, let's not launch Cloud one or whatever. Well, Okay, maybe in hindsight. But hindsight is twentytwenty.

And at the time, they didn't know how that model would be used. And to them, felt existential if somebody came and said, you weren't responsible. This wrote a bug.

The liability associated with that is massive. So how do you prevent against that? Well, in, day out, you say safety, safety, safety, safety.

And when you start deviating from that, you have the team hold you accountable. You have the world hold you accountable. And I think that becomes a moat over time.

At some point, that moat will get challenged and so on, and then it becomes fragile. I hope it endures because that's the beauty of having founders run the show. Because they can make really hard trade offs to do mission alignment.

The hardest part is in the earliest days when you don't have a group of people who are going through difficulty, stress, crisis together, then your culture doesn't get defined sharply enough. And that's what I'm worried about right now, is there's so much money going to these labs, there's no hardship. There's no 21 nos.

There's no 21 nos. And that, in hindsight, was a feature, not a bug for Anthropic. The number of people who said no, the number of people who said, sorry, we're all the investors in OpenAI, that is competitive difference.

It forces you to really understand what is the hill you want to die on at the expense of everything else? What's the p zero? And there, p zero from day one was coding.

The reason the mechanism system there was if we crack coding, then we will crack AGI. You know, our mission is AGI. We wanna get there safely.

If we focus on coding, it's such a generally powerful capability that it can accelerate all kinds of work on a computer. And if we're going to accelerate all kinds of work on a computer, we can get to AGI. As a result, they've had to say no to so much other stuff.

Here, superconductivity is the mission. Coding is not the mission. So we use Claude.

We'll use Claude. We don't care about that. The mission defines everything.

And I think teams who can raise too much money too fast, too early, who don't have to define what the P0 is, because that's the only thing when you have scarce resources you've got to invest in, those cultures end up being the most fragile and brittle, and they almost don't even make it to take off. So let's apply this to periodic since we're here. Sure.

What is the constraint or the hardship that they were forcing themselves to go through? Dude, here? Are you crazy?

No. Well, Okay. So on a technical level, it's physics.

It's literally reality.

one that's like the company building? Yeah. Mean, When, Liam was a co creator of ChatGPT.

And Doge was skip level from Demis at DeepMind, had created genome, so one of the important tools to come out of DeepMind. You know, at the time, I was a visiting scientist at the Stanford Physics Department, we had started benchmarking frontier models on physics and science capabilities. They were not very good.

They were good at, like, doing things like summarization of papers. But if you said, hey, could you analyze the scientific data coming out of a condensed matter physics lab I was in the condensed matter physics group at Stanford that was terrible. So it was not popular twelve months ago.

Peer record and I won't go into details, but there were people who said, recently as a few months ago, who said they wanted to join the company. And they, for whatever reason, took a job elsewhere. They kind of reneged on their commitments.

They took a job elsewhere that offered more money. Then we had a technical breakthrough. Create a SOTA system and like Okay.

I'm excited. Yeah. We'll we'll be doing a separate part on on periodic And then they wanted to come back.

And I said, no. Yeah. No way.

You you come here. You had your shot. You you had your shot.

Because it's actually about culture. Of course. And first principles.

Yeah. You know, and look, I believe in second chances and so on, but time will need to heal. Some of those wounds were they will leave deep, deep for them, leave deep scars.

But because I started my company at 24, 25, I went through the whole cycle of betrayal and drama. And so you realize know, Silicon Valley is both a very missionary place. It's also a very mercenary place.

Sometimes people lose their minds with when big money gets involved, which is and the grand scheme of things, quite small money.

will sleep on a hard slab of stone. Like, my mattress was this thin, you know? I mean, you grew up in Singapore.

When I got to Singapore, I used to sleep. I was part of the scholarship program, but which which was amazing. I'm very grateful to the Singaporean government.

was a transition dorm because they're building this beautiful, like, residential campus on-site at SAJC in Potong Pasir. But we were the last, I think the second last batch to be in the transition site, which was some old, like, I think was like an immigrant laborer.

So I lived in a, my eleventh and twelfth grade, I slept in a bedroom the size of this, like, literally from from there to here. Yeah. Right?

They were like bunk beds. And so one bunk bed here, one bunk bed there, one on top, one on top, one more here, and then here was where our like, we kept our toiletries and clothes and stuff. And when one guy would climb onto his bed there, this one would shake.

Oh my god. And one of my roommates who was from and when he was amazing. I loved every minute of it.

You know, my my roommates were a guy who was a a top ranked DOTA player from PRC from China, didn't speak a lingo Loved him. Amazing guy. I mean, all the Singapore scholars are fantastic.

And honestly, we should treat you guys better because of what you're going to do. But cool to know. No.

I mean, what I'm saying is I don't need much to be happy in life. You know, when you've lived through that, money is a way that I think sometimes we measure ourselves. You know, when it stops becoming know, it's more of Goodhart's Law.

When it stops becoming just a byproduct and more of a measure, it stops having meaning.