0:00
You're watching TVPN.
Search this video
0:02
>> Today is Friday, September 11th, 2026.
0:05
Uh wallto-wall coverage in the Wall Street Journal.
0:08
Remembering 25 years ago, uh the cover of the Wall Street Journal.
0:13
Every single article except for one was about, of course, the World Trade Center.
0:18
The the the headline that day was terrorists destroy World Trade Center, hit Pentagon in raid with hijack jets. Uh Bin Laden is on here.
0:26
the only story that uh you know attacks raise fear of a recession.
0:32
Very interesting time capsule.
0:34
Highly recommend picking up a copy of the journal today and uh taking a trip down you know in in memory of uh the tragedy.
0:42
The one piece of news that broke through this day that was not related to the terrorist attack was Xerox reached an equipment financing agreement with GE Capital that will let Xerox erase about five billion of debt. Very very odd.
0:57
Every other story I mean the market was closed.
1:00
Actually the internet got a very interesting shout out here.
1:03
Says telecom systems were strained as terrorist attacks in New York and Washington knocked out telephone wireless services across the Northeast.
1:10
The internet proved most the most reliable way to communicate following the attacks.
1:15
As the phone system sagged from severed lines and a and an extraordinary volume of calls, corporate executives used email to find employees across town or across the country.
1:27
So interesting to see, but there's so much to go uh into and lots of interesting retrospectives across all the different media organizations.
1:35
I don't know that I have a particular uh anything to dive into there, but uh it's interesting for you to go and dig into. Anyway, moving on.
1:43
Let me tell you about ramp. com. Time is money. Save both.
1:45
Easy to use corporate cards, bill pay and accounting, and a whole lot more all in one place now with musicals.
1:51
Anyway, um, what are the AI doomers actually proposing?
1:57
That's the question I was trying to answer this morning for a while. >> 20ear sentences. >> That's one that.
2:04
So yeah, there will be this weird translation layer between the thought leaders and the people that are writing policy papers effectively and then what actually gets implemented, what gets, you know, votes basically can be wildly different because uh to get the populace to actually support something, you need to wrap it in a different different structure potentially.
2:25
Uh AI 2027 predicted that this month, you know, late in 2026, Congress would wake up and that is in the journal.
2:34
Congress is suddenly waking up to the AI doomsday threat.
2:36
And so, uh, this is happening all over the place.
2:41
Uh, was it Matt Damon who was caught on TMZ being invest interrogated about his thoughts on AI risk. There's now protests.
2:49
Uh, I believe AI 2027 predicts a 10,000 person a anti-AI protest by the end of the year.
2:56
Uh, I was trying to figure out how big this protest in Chapel Hill, North Carolina earlier this month was.
3:03
I think it came in sub 10k, but certainly tracking to it.
3:09
Um, it was about 150 people at this uh at this protest in Chapel Hill.
3:14
Uh, but just two oos away from the prediction from AI 2027 and AI 2027 in the sequel AI 2040.
3:24
Uh, frustratingly vague about impacts outside of the AI industry.
3:33
So there's a lot of really really amazing predictions about uh agentic capabilities and the amount of compute that will be marshaled and and like even lab revenue but there's not as much predictive stats and calls around like what will it do to GDP?
3:50
What will it do to employment?
3:52
What will you know how how often will people actually be using this?
3:55
What will it actually be good at?
3:57
The diffusion question is still sort of left unanswered.
3:58
Um but they're clearly taking it very seriously.
4:02
there's a huge uh press cycle around this and the journal says that uh Congress is suddenly waking up.
4:08
So, I wanted to dig into like what the actual proposal is because the joke for a while has been just like everyone when they're pressed on this, they just say we got to talk about it.
4:17
We got to talk about this. uh AI 2040.
4:20
Daniel, uh uh >> one thing that one thing that stands out, it's it's interesting to me that uh there's been so much more seemingly grassroots mobilization around the anti- flock movement, Dlock.
4:35
When you compare that and and you don't see >> totally >> you don't see videos of of people with, you know, 100,000 likes saying like, "Here's how to cut power to your local data center.
4:48
not for X risk reasons, but you will see a post that's just like I don't like AI image generation because I'm an artist and that will get a lot.
4:57
So there's a lot of anti but what I'm saying is like the sentiment the the anti- flock sentiment converted into actual physical actions by a bunch of just otherwise normal people. Yeah.
5:11
>> I'm saying we haven't seen that yet.
5:11
Uh with data centers maybe we don't. >> Yeah. >> Um but it's notable. Yeah. Much harder.
5:20
Data centers are in remote locations, highly fortified. There's fences. Totally. Just walk in.
5:25
>> Block cameras are in your, you know, on your street maybe. >> Exactly. Yeah.
5:28
So, yeah, there there's there's degrees there.
5:31
Um, but AI in in the what's weird is that Yeah.
5:34
I mean, you're you're calling out uh grassroots taking down the data centers.
5:39
Uh um the AI 2040 proposal is effectively the opposite.
5:45
it's like hardened the data centers even more.
5:47
Uh, and the actual proposal is super concrete in AI 2040.
5:52
And it's it's very interesting just to hear about how they want to slow things down.
5:57
So the the the thing that I think a lot of people online who are like, yay, an anti-AI sentiment that's going viral are going to be depressed about is that like this is not stop AI at all.
6:06
AI 2040 is like keep the inference flowing.
6:12
The current models are great.
6:12
Uh, also let's keep doing capabilities research, but we're just we just want to reach super intelligence by 2040 instead of 2028 where we're not necessarily prepared.
6:25
So they want it to be highly controlled by governments, nations, states, highly secured and there's a whole bunch of very very tactical recommendations that they make uh around that.
6:36
So the first mechanism is an AI pause.
6:39
They want to pause training.
6:39
They don't want to do any more new frontier training runs or R&D experiments.
6:43
And to enforce this, they're calling for uh to apply inferenceonly verification to essentially all major AI data centers.
6:55
So anyone who has more than 10,000 H100 equivalents, roughly a hundred million dollars of equipment if and and that is like pretty easy to figure out.
7:03
You're just like big building over there.
7:05
Let's send the inspector inside.
7:07
Oh, says Nvidia on all these chips. Count them up.
7:12
There's over 10,000 of them.
7:12
You got to apply for this permit.
7:15
You got to do you got to tell us what you're doing, right?
7:18
Very easy to enforce, at least in the United States.
7:20
Um, and with an international body, you could kind of do the same thing internationally.
7:23
So, uh, the whole goal, you can only inference the current models and you got to verify your workloads with an independent auditor.
7:30
Probably the government, maybe there's some sort of, uh, uh, non-governmental organization that's doing this.
7:36
You know, there's a whole bunch of different solutions that you can pull from across nuclear and non-prololiferation uh work that's happened in the past.
7:43
Uh major countries, they want them to declare AI compute inventories.
7:47
Tell everyone, not just your local population, but also the international community, how many how many warheads you got?
7:55
How many H100 equivalents do you have? Where are they? Uh everyone shares this.
8:01
That's going to be a really tough cell because international agreements are really, really tough cells.
8:07
It's much easier to have a ground swell of support for something that happens in America. America changes.
8:11
We have a system that we don't really have the international rules to to to quickly implement that in a way that uh doesn't doesn't allow for a lot of defection, but they want to know who has compute.
8:24
Uh so major data data center owners and semiconductor supply chain companies would be required to turn over sales records.
8:30
Who' you sell those chips to?
8:32
Where'd they go after that?
8:34
Foreign inspectors would do routine chip counts physically at site on site at facilities.
8:39
Large transfers of chips would only be allowed to go to registered audible counterparties.
8:43
Uh there's some interesting networking specific uh components to this proposal too.
8:50
They want to physically remove high bandwidth east west networking inside of data centers.
8:56
So you can't do large distributed training runs, but you can still do inference.
8:59
So again, anyone all the anti-AI people who are like yeah I I don't want LLMs around anymore like the these are not your guys.
9:07
[laughter] Uh they're not fighting for that.
9:09
They are fighting for stopping the next training run.
9:14
Uh which is probably a line.
9:14
Those are those are overlapping circles.
9:16
But uh it is not moving backwards in time.
9:21
It is merely slowing down at this current moment.
9:23
Um they also want to install passive optical network taps on anything leaving the data center to independently verify traffic.
9:29
Uh for any new AI R&D data centers they want uh entirely new facilities built from scratch with nation state level physical security and ver and verification.
9:40
security and ver and verification. So they're saying like okay we're we're like the the data centers that are built they can inference the current models your your Astros your fables like your your grocks like those can run because we we we can deal with those we can
9:55
harness those we have control over them we're going to continue to align them uh and and that's a solvable problem but for the next run and the one after that and the one after that as it gets crazier we want it in a new building built inside a Faraday cage so you can't communicate it from the side. Uh we want
10:10
Uh we want a bunch of physical controls.
10:13
It's like going to a nuclear facility.
10:15
Highly verified who gets in the building and when.
10:19
Uh air gap communications.
10:19
Uh this is one interesting uh like proposal that they have that really shows how deep they thought this through.
10:27
They want the R&D data center to be connected externally.
10:32
If you want to communicate with it and you want to tell it what to do, okay, train the next running or do whatever.
10:38
uh they will have a bandwidth capped connection at one meg per second.
10:45
So you can send little instructions but if you say send me the weights because I'm taking them somewhere else it would take you like five years to exfiltrate it.
10:53
So interesting like hardware solution to this I I you know how do you actually go and implement that?
10:59
There's going to be a whole bunch of other things but interesting [clears throat] that they're thinking about like the width of the pipe.
11:03
Uh, so it'd be very obvious if you're stealing the model weights because it's like, wait, this one meg pipe has been at full tilt for months. What's going on here?
11:11
Someone's taking the stuff out of the data center.
11:15
Uh, they want if front when Frontier model weights move from an R&D facility to an inference facility, they want it to be placed on physical storage devices encrypted independently by both the US and China.
11:26
So both countries have to sign off and physically escorted by representatives of both countries to the destination.
11:32
Uh, >> it's a tall order.
11:33
That one's a tall order for sure. Um, >> yeah.
11:38
>> And they actually want Frontier models to be made deliberately larger than compute optimal.
11:41
So, they want the weights to be a 100 terabytes instead of honing them down to something that's just one terabyte that could actually be moved around a little bit easier.
11:49
Uh, there's a bunch of public disclosure proposals in there, restrictions on various tradeoffs.
11:54
So uh labs would have to share the model specification, the fraction of compute devoted to internal AI use.
12:01
So you don't get a lab that's just internally using way better models than what's available externally.
12:05
Uh they want qualitative descriptions on how how powerful models are being used internally.
12:12
Uh restrictions on how big the gap can be between the best internally deployed model and customerf facing products.
12:19
This has been a common discussion point with like uh the roll out of Mythos and Fable and Astra and Astra Next and all these different models where people have said like ah it's really unfair that like this lab gets a better thing than I do.
12:31
We should be on even footing if we're both going to be competing in web design or we're both going to be competing in legal like what why can't I buy this product from you?
12:41
Um, and so there's some overlap there with like the broader business community, which I thought was interesting.
12:46
And uh, >> yeah, I mean, first of all, this I mean, a lot of this tracks with the kind of regulation that, you know, we had pushed for, you know, beginning about two years ago around podcasts. >> Yes.
12:56
>> You know, wanting podcast studios to be airgapped. Yes.
13:00
>> Wanting, you know, Faraday cages around podcast studios.
13:04
>> A locked briefcase with an SMB7, SM7B in it.
13:08
And in order to unlock it, Patrick Oshanosy and David Senra both need to give you codes to independently verify that this podcast is worthy of being recorded.
13:17
Yeah, >> I like that one. Yeah.
13:19
>> Uh it just makes sense.
13:19
Um so, uh high level Val Valve they see as being like the most effective in controlling the speed of capability improvements is compute caps.
13:30
Uh so there is a world and we're going to the Bernie Sanders thing because it's already getting like sort of twisted but um >> uh the the the the big hammer is just chip controls and data center buildout slowdown.
13:45
That's the easiest thing and I think that's like the biggest valve that they're going to that we're going to see twisted around uh to actually slow down capabilities.
13:54
[laughter] Um, and so the goal is to allow models to get better mainly by adding hardware rather than inventing better algorithms which can leak to secret projects.
14:05
So the goal is like okay well we know that this model is capable of this so we want this much compute over here.
14:12
Okay, you've done well we're allocating more compute as opposed to this one weird trick that >> AI doomers hate. >> Yeah.
14:21
So the goal here is so interesting because when you look at I mean anytime you have uh you know really really hardcore government regulation >> uh and international coordination around issues like this, you're going to have a bunch of unintended consequences. Yeah.
14:39
And one thing that feels obvious around this uh if if these policies were to be rolled out is that you would effectively create an incentive for millions of individuals or groups globally to be in secret like trying to find entirely new breakthroughs that are >> Yeah.
14:59
And and and again, this incentive already exists, but it kind of pushes a lot of the idea that humans are just going to be like, "Oh, I'm not I'm no longer going to try to create the god model because >> Yeah.
15:16
>> Like there's this, you know, big global organization that's that's sort of policing it." >> Yeah.
15:22
>> Um I just >> I mean that's the same thing we see with nuclear non-prololiferation.
15:25
There's always a discussion about what countries are getting the bomb and when and how far along are they and wars break out over this and uh it yeah like the game's not over just because you create a framework.
15:37
Um but there is at least a I mean we've avoided World War II so you could say that uh a lot of like the vast majority of nuclear non-prololiferation work has been successful even though there's been a ton of examples of people trying to divert around it.
15:53
In fact, it's like been the backbone of geopolitics for like 60 years has been like who gets the bomb and what what what chips are on the table and stuff.
16:02
>> Ultimately, uh GPUs and computers are much more wide, you know, infinitely more widespread than nuclear materials. >> Yeah.
16:15
But you still got to marshall them all together.
16:16
all together. Yes, there's some weird scenario where there's a Python script that's AGI that can run on your laptop, but I think most people are convinced at least in this crowd um that uh scale >> that scale is a prerequisite and I mean we were we were joking about like how it would be so like we SSI's new neolab is uh recently got a big cluster from Nvidia and we were like the most bullish
16:43
thing you could do if you're this secretive Neolab would be like we're actually selling our compute because we've discovered a more compute optimal way to reach AGI and we don't need a lot of compute but of course even Ilia is like it's time to scale up I need more compute because it seems like even if he's taking a completely orthogonal approach to you know uh innovation and
17:03
research he still needs a lot of compute and so it does it does feel like everyone is sort of with the consensus that it's going to be a big building with a lot of energy big heat signature definitely visible from space and and pretty simple to to track uh at least in the short term until people start building crazy underground facilities and then you're back to you know nuclear nonp proliferation but uh the their goal
17:28
is at least like you know try you know try >> yeah and then the the other side of this is does it does AI development actually become >> something closer to the Manhattan project where >> you know what a lot of people 100 researchers are working with the government in secret because you can't just assume that other countries are going to slow down or do any of these things. >> Yeah. Um and but in in general, I think >> Yeah.
17:52
Um and but in in general, I think the proposal is uh don't go back in time.
17:59
It's definitely not stop everything in its tracks.
18:01
Uh it's a slowdown with the goal of scaling gradually.
18:05
They do actually want to reach super intelligence.
18:08
They just want to do it by 2040, hence the name of the project.
18:12
So the goal is uh sca gradually scale into top human expert capability around 2035.
18:19
Now a lot of people are saying oh we might get this by 2029 2028 2027 uh and they see that as too fast.
18:27
So they want to push that out to 2035 then wait five years with AGI and then unlock super intelligence in 2040.
18:33
This is their initial proposal.
18:35
Of course there's a lot that could change over the next decade.
18:38
Um and uh and I think I think like to zoom out overall if you're worried about X-risk the AI 2040 plan does feel like a concrete path towards slowing down.
18:50
the conversation definitely gets dragged down into pdoom estimates and trying to narrow down exactly how a human extinction scenario plays out.
18:57
And that can be uh that can I feel like that's almost a sideshow because in a democratic society in amongst humanity like it it doesn't really matter the mechanics of getting to 10% poom or any of those.
19:12
It's just like if everyone feels that way something will happen.
19:17
this is a concrete plan of what that might look like and that's valuable to understand in this case. Yeah.
19:22
So, >> uh for the safety skeptics, it's easy to see how this level of control over what you can do with computers is authoritarian or anti-lbertarian.
19:31
Uh even if we're talking about hundred million dollar computers, there's a lot of people that say like, I should be able to do math on my computer.
19:37
I can do whatever I want. Let me do cool things. I'm excited about this.
19:41
Uh that limits your freedom.
19:44
Uh it might create regulatory capture.
19:46
uh for a few major players.
19:48
It might crash the stock market or delay economic gains that come in the good ending where alignment is solved and X risk plummets.
19:54
You can imagine a situation where uh in a few years if X risk fades into the background, you're like, yes, there's still a risk, but it's the same risk that we face every day with like an asteroid hitting Earth like no one it doesn't really change anyone's behavior.
20:07
Uh that would be sort of the good ending in my opinion.
20:09
Um, so it's a balancing act.
20:12
And for most of these slowdown proposals, I personally have a hard time blackpilling about them in the sense of like if the slow if all of this gets implemented, how frustrated will I be?
20:23
Like the models are good.
20:26
I would like better models. I want safe models.
20:27
But at the same time, like there is this massive capability overhang.
20:31
The current models can do a lot of interesting work.
20:33
We're finding new uses even for like non-leading edge models.
20:38
There's a lot that can be done.
20:40
So I I I don't believe the the the doom doomers who are dooming about what the doomers are planning.
20:47
I find that unconvincing right now.
20:49
But people are starting to lay it out more.
20:53
Brad Gersonner, Jensen Wong, David Sachs are talking about the other side of this equation.
20:58
Um but but I haven't just like it's hard for some people to concretize the Terminator scenario.
21:04
I also have a hard time concretizing the uh the we didn't we didn't race and were unhappy about that.
21:13
I guess you could say, you know, the housing scenario.
21:16
There's been other times when we brought in too much regulation, slowed things down too much and been like, ah, like this was really not the right move.
21:23
But uh at the same time, I think we we have a lot that we can do with the current technology that there's still cause for optimism even if something like this gets, you know, universally voted on.
21:35
Uh it wouldn't I I don't think it would be the worst thing for for for companies and consumers and businesses and all sorts of different folks. >> Yeah.
21:43
>> But >> there's also a big question around what uh does does do the 2040 people have a point of view on robotics and and physical AGI?
21:52
because it seems like if you even if you pause uh like you know efforts towards RSI well if we add billions of robots into the world that are just running on today's model that also presents today's like you know >> I don't think they're worried about that >> yeah but billion robots with GPT6 level intelligence and years of alignment work that is currently happening. That's fine.
22:25
It's the next next next thing, the super intelligence, the thing that might have its own goals.
22:30
Uh I think I mean we're talking to uh uh uh someone from OpenAI's robotics team hooking up Astra to a robot, a paintbrush, and a camera.
22:42
We talked about it earlier, painting.
22:42
Um I don't think we're at a point where that poses a risk. It's the next model.
22:51
It's the model with its own valition basically which a lot of people still aren't seeing.
22:55
They're just like yeah like the models keep getting better but uh they seem to follow your instructions sometimes too much and then you need to worry about the paperclipip scenario but um but but it's not that they they want to do their own thing necessarily.
23:06
I don't know but people are going back and forth on this.
23:10
Clem over at Hugging Faces sorry but asking Jacob about AI extinction risk is like asking your AC guy about climate change.
23:17
Not saying it's necessarily uninteresting or wrong per se, but let's keep things in perspective and he and hear from the full range of expertise across the ecosystem.
23:27
And Nathan Lambert says, "Banger." Uh, what was our take?
23:34
>> AC guy might be right about >> Well, my AC guy would be like, "I don't really know about that, but I just want to make sure when you're hot that we can run this AC cool."
23:44
Yeah, I guess it's Yeah, the ACI seems seems >> I don't know about all that.
23:48
I don't know about all that mumbo jumbo, but you know, when it's a hot summer day, I don't want you to be worried about the heat, brother.
23:55
[laughter] >> I like that.
23:56
I Yeah, this is this is kind of an unnecessary shot at AC guys.
24:00
The you know, AC guys are important. Uh I don't know. It's funny.
24:03
Anyway, let me tell you about Figma. Agents meet the canvas.
24:06
[cheering] Your AI agents can now create and modify your Figma files with design system context.
24:10
What did Bernie Sanders have to say? Is this is this real?
24:14
Entities will shall shall be subject to the corporate death penalty and persons shall be shall be subject to more than 20 years not more than 20 years in prison if they don't pause AI development.
24:31
That seems pretty easy to comply with. I don't know.
24:35
I guess how do you define AI development?
24:40
is prompt engineering AI development and then you get caught because your model sort of did a little prompt engineering in the final stage and then you're guilty of this like that.
24:52
Yeah, that could be a negative knock-on effect.
24:53
I guess it does seem aggressive.
24:55
Um but his overall proposal is banning artificial super intelligence so no person or entity may develop or deploy super intelligent AI systems.
25:03
Uh he defines artificial super intelligence as an artificial intelligence system that exhibits or can easily be modified to exhibit capabilities that match or exceed human cognitive performance and capabilities across a broad range of domains or tasks.
25:19
It sounds like the mission statement.
25:20
[laughter] It sounds like the explicit goal of of like 17 different companies right now.
25:25
Um AI AI system or AI systems that have sufficient capabilities to plan and execute the disempowerment of humanity. Okay, that's a good one. I like that.
25:33
I don't like overthrowing or undermining the US government.
25:36
Um so strongly in favor of banning that.
25:39
Uh pausing advanced AI development until a new federal AI regulatory body is up and running and then the new cab cabinet level federal agency will monitor frontier AI systems at all stages of the life cycle, supervise the removal of dangerous capabilities and supervise the destruction of artificial super intelligence.
26:00
were coming for it, which is similar to the >> corporate death penalty is a line that you don't hear a lot, right?
26:05
Usually pe us usually these companies just, you know, go bankrupt and wind down.
26:09
But, uh, corporate death penalty >> goes pretty hard. >> Kind of metal.
26:17
>> It's kind of It's kind of metal.
26:18
>> Yeah, you kind of got me with that one.
26:20
[laughter] Uh, yeah, rough rough rough situation. We'll see. We'll see where it goes.
26:26
Um Jamie Cox over at um uh over at Fluid Stack, the co-founder of Fluid Stack, a uh a compute provider, shared his convictions.
26:38
Uh sort of pushing back on a lot of this, saying that he thinks America should build more.
26:43
They're pro- freedom, pro-democracy.
26:45
Uh they believe AI will bolster human flourishing.
26:48
Uh we support simple clear enforcable regulation frameworks that set simple requirements proportion to capabilities and risk with clear responsibilities and no unnecessary barriers to competition. Yeah.
27:04
Yeah. the the real you're going to see a lot of push back from people who are like the like the regulatory stuff is going to be like these 10 companies and I'm going to be number 11 and I'm basically getting the corporate death penalty then because I didn't make the
27:19
cut to be one of the regulated one of the approved companies I'm still early in my stage so uh there's a lot of nervousness I'm sure but uh Will Manitis said we believe AI will make everyone rich healthy and free is novel and interesting comms from the frontier. Uh he's he's endorsing this and I I
27:37
Uh he's he's endorsing this and I I agree.
27:39
I like I like this uh I like these convictions.
27:41
I think it's generally like a positive direction to move in.
27:45
Not a direct response to the the proposals that are going out, but we're going to get a whole lot more of them.
27:51
Uh where do you want to go to next?
27:53
Uh over on Tik Tok, they're sharing a photo of the whistleblower >> and saying in every worldwide disaster movie, there's a dude that looks just like this that nobody listened to.
28:07
>> And he really does look like an act.
28:07
He does look like an actor here. >> Yeah, he looks good.
28:11
Um but people are people are all >> And please stop calling him scary Potter.
28:17
I've been seeing people I've been seeing people over on X calling him scary Potter. That's the goal.
28:22
The goal is to is to wake up China, wake up Congress, wake up everyone.
28:28
>> Door Dash has entered uh the conversation. >> Indeed.
28:31
>> They say two years at Door Dash.
28:31
I do not say this lightly.
28:33
We are extremely close to the burrito arriving before you decide you want it.
28:38
We are not ask We are not asking for a ban.
28:40
We are asking for a pause.
28:42
I don't know why they would ask for a pause.
28:46
>> Um >> yeah, >> that seems like uh very very aligned to humanity. >> Yeah.
28:51
and to their business which I think is fantastic.
28:54
>> The doom is very much contained to the frontier lab work.
28:57
Everyone in the application layer who's applying the models diffusing them they're like I just I can't get this thing to work right.
29:06
I got to get I got to get forward deployed engineers to teach people how to use this thing.
29:09
Everyone deeper in the stack Jim Rainbow over Jim O'Reilly. Sure. Or Jim Riley. >> Yeah. Jim Riley.
29:16
>> Jim [laughter] >> O'Reilly Auto Parts.
29:17
Uh Jim Riley over at Charleston AI says it's a whole lot of mumbo jumbo. >> That was your words.
29:24
>> He just he's just happy to get, you know, eight hours back. >> Yeah. Yeah. Yeah. Yeah.
29:28
And then yeah, everyone deeper in the supply chain like Jensen and all the different semiconductor manufacturers are are not particularly on this side.
29:38
And then you also have Wall Street who's just like what's the enterprise acceleration?
29:43
So, uh, lots of different groups around the table that need to be brought on board to this movement.
29:50
Uh, the discourse truly is fascinating.
29:53
Uh, let me tell you about console.
29:54
Console builds AI agents that automate 70% of IT, HR, and finance support, giving employees instant resolution [applause] for access requests and passwords.
30:02
[laughter] >> That's a new >> too much diet coke. >> New uh new approach. >> New uh Okay.
30:08
So, what what is Tyler Cowan calling for?
30:12
Uh >> he's banging the table saying bet on this.
30:17
>> Tyler, are you named after Tyler Cowan?
30:21
>> Is that is that your namesake?
30:24
>> Tyler Cowan says, "If you have very pessimistic fears or predictions about AI, name the market prices that will support or confirm them.
30:30
This is what taking this seriously means." And I love Tyler Cowan.
30:35
I'm not sure this matters because if you if you you're not going to be around to collect it, right? >> Yeah.
30:41
This is what what deep deep dish iner says.
30:43
Tyler, why would short-term existential risk affect market prices in any meaningful way?
30:50
>> Spell out the exact mechanism.
30:50
Contracts that pay out if everyone dies aren't worth anything to me. >> Yeah. I don't know. Yeah.
30:57
>> I think he's just he's calling for like, you know, he wants people to like make a falsifiable claim.
31:01
Like is this am I able to tell if your claim is like true or false?
31:05
And so it's like very hard with these scenarios where like >> isn't it an unf unfalsifiable claim though just by definition and like you just have to like accept that and move on. >> Yeah.
31:15
But then it's like so hard to have any like real discussion.
31:18
>> Was what what was nuclear any any different?
31:20
Like the threat of nuclear apocalypse, the threat of World War II, this was a very motivating factor for decades, most of the 20th century.
31:31
People made real decisions based on it, based on where to do business and where where where the where the conflicts were going to be and the motivation for nuclear treaties and non-prololiferation and how we treat >> you can make financial decisions uh like think about nuclear war, right?
31:48
You have a you have a bunker that's like a decision you make.
31:51
Is anyone making like AI bunkers? No.
31:53
Because they think it's going to be so totalizing that the bunker actually doesn't do anything. >> Exactly. Yeah.
31:58
So if you think it's so totalizing, then you don't make the bunker.
32:02
And so saying, "Hey, you don't have a bunker," is not is not proof that the person doesn't believe what they're saying. >> Yeah.
32:09
I I like I don't know the answer to this question either, but like it seems like >> I don't know.
32:13
Maybe there's some question you can ask that is falsifiable. >> I don't know. I don't know.
32:16
I think you just got to you just got to believe these like this crew that like that's what they believe and uh you know like they they believe it.
32:25
There there is the other side of this which is Paul Cristiano uh who uh is has been worried about risk.
32:33
He recently just joined the board of OpenAI and um and uh Tyler Callen, you know, has this quote, if you're a doomer, why aren't you short the market?
32:42
And Paul Cristiano is uh 2x levered long, and he's short the isn't he the the the bond market or US treasuries, right?
32:55
So, he has 5% of his net worth in Tesla, 90% of his net worth in AI bets, and 100% of his net worth in normal investments. No Tesla options.
33:05
Uh, that sounds like a scary place with lottery ticket biases and the crazy Tesla investors.
33:08
Uh, and then you leazer Udicowski says, "Am I correctly understanding? You're 2x levered."
33:13
And Paul Cristiano says, "Yeah."
33:14
And so, um, he says he's personally short the US the the US 30-year debt.
33:19
I think that just means you have a mortgage.
33:22
I I I'm pretty sure that's like if you have a mortgage, you are effectively short the US uh 30-year because you have you have sold that debt and you got the cash effectively. That's how that works.
33:35
But um still it makes sense because you if you have enough money you could pay off your mortgage, go long that debt and short the market on a relative basis.
33:45
So um but but again that's not this doesn't seem like a doombased bet.
33:50
This seems like this bet also pays out in just like the good ending and like AI is real and delivers value.
33:58
So your AI bets perform well, the market performs well and money slides from US debt to data centers and AI buildout debt or something like that.
34:09
So uh he's putting his money where his mouth is, but it doesn't feel like a representation of like doom by any means.
34:14
Anyway, let me tell you about Shopify.
34:16
Shopify is the commerce platform that grows your business, lets you sell in seconds online, in store, on mobile, on social, on marketplaces, and now with AI agents.
34:25
>> Uh, speaking of AI agents, >> people have been creating AI agents of fruit flies. Have you seen this? >> Yes. >> Okay.
34:32
So, >> where should we start?
34:34
Because I want to set the table on what what is actually going on.
34:38
Um, Tyler, do you understand this?
34:43
>> Break down break down what people are cooking.
34:45
>> We've seen this before.
34:45
We've talked about this before, but what's actually going on with the with the fruit fly?
34:50
Because now people are taking the fruitfly all over the place. >> Yes.
34:53
So, my understanding is is Google basically mapped out uh all of like the neurons in a fruit fry fruit >> fruit fly.
35:00
So, they probably took a took a took a deceased fruitly and put it in like a mass spectrometer or something that can can investigate the brain at a very very like a deep highowered microscope effectively. >> Yeah.
35:14
like the entire >> 3D they got the entire structure and now people have been able to recreate that in software in a simulation. >> Yes.
35:23
So I think in theory you can like replicate all of the the flies like decisions or whatever. It's like movements. >> Yes.
35:30
>> And this is notable because many people have said I have I have the mind of a of a fruit fly.
35:36
>> I have the in they've said I have the intellect of a >> So now they're going to put it to the test to to see which performs better.
35:43
the simulated fruitfly or just Jordy Hayes we got.
35:46
>> This is organic farmtotable Jordy Hayes.
35:49
>> So, so now that this is out and the code is out and you can run this fruitfly in simulation however you want.
35:54
People are having uh they're doing all sorts of experimentation.
35:58
So, uh Kevin said he trapped the fruitfly in his rabbit R1.
36:04
When he shakes it, he can see its brain brain's escape circuit light up.
36:07
Um, by the way, our consciousness is physically defined, which is why our physical things like drugs, neurotransmitters or events alter or initiate or end our consciousness and things like sleep, hallucination, waking or death.
36:20
Computer simulations are also basically defined representations. >> Yeah.
36:24
I think what what what ends up being uh unsettling and uh weird about this is is if if it just just purely you being human and doing this uh is is probably not good for your own soul.
36:40
Uh and you know, imagine imagine you have a fruitly in a box and you're shaking it and you're like look it it wants to escape, right? Um, totally.
36:51
I'm not a huge fan of small insects.
36:54
I don't really want them around that much.
36:57
>> Um, but when they're in my house, >> I try to, you know, if a spider's in my house, even if I know it might want to take a nice bite of me, >> I'm still going to, you know, try to transport it out of my house and and put it back into the world.
37:10
I think it's I think uh it's it's not good for your for your soul to be, you know, a merchant of death. >> Sure. uh situation.
37:20
>> And yet, if you're playing a real-time strategy game and you highlight a bunch of soldiers in this simulation and you send them on a charge that will result in their virtual death, you might not feel >> those soldiers opted into riding and dying with you. >> Okay. Okay. >> Right. >> Okay.
37:41
What about fruitfly is just sitting there, John.
37:43
>> So, they made the fruitfly play Doom.
37:45
What about if you play Doom?
37:45
You are killing a demon who's simulated. Is that immoral?
37:53
Like, >> well, that is that demon trying to kill you?
37:57
>> A lot of it comes down to like the fact that it's simulating a the actual representation of the fly makes it a lot more concrete than just, oh yeah, it's a it's a 3D model in a Python script that just says if you see a if you see a character, shoot at them in the simulation.
38:12
simulation. But we're we're clearly starting to grapple with like these odd moral questions of if you're simulating something then the next step it's a couple order magnitude but you get there and you can simulate a human and you could talk to that human and it would do
38:27
everything the human does that human have rights and agency is an ethical moral agent or is it merely just a simulated uh just a really good computer simulation it's just existing on transistors so you don't need to feel any moral about anything that you do to it. Uh I agree with the just the vibes
38:44
Uh I agree with the just the vibes based analysis that like torturing a real fly, torturing a virtual fly, probably just don't be in the business of torturing anything.
38:52
You don't need to overthink it, but people are the the real debate here is like the question of you know, are LLMs sentient? Are they moral?
39:04
Do they is there a moral weight to the uh to to to synthetic intelligence to artificial intelligence?
39:12
That's what people are debating here and I'm sure the debate will continue.
39:14
Who knows if it will ever be ended, but they did teach it how to parallel park, which I think is cool.
39:20
Uh, interestingly, last night I had Astra use computer use to play a video game that I very much enjoy playing called Bellatro.
39:27
It's sort of like a modified poker game.
39:29
And I don't feel like I was torturing the LLM by that.
39:34
I feel like I was giving it a treat.
39:35
I was like, "Hey, instead of doing my taxes, you get to just chill and play a video game." Uh, it did very well. It won.
39:40
Soul was not able to win.
39:44
And it was really fun because I would like pop in while it was using the computer and kind of armchair quarterback and be like, "Is it making the right decision right now?"
39:50
Felt like uh felt like a coach coaching like a a kid on the soccer pitch or something.
39:56
>> I had a fun I had a funny moment uh last night.
39:58
Uh I was uh on my racing sim simulator uh comparing uh asking Chad GBT for uh for to to compare my times uh at Lagona. Yeah.
40:13
>> Uh to to just other other you know what what would bestin class be like?
40:15
What's beginner like, etc.
40:17
Um [clears throat] >> and it said if you want I can help you uh you know cut cut some seconds off of this time. Yeah.
40:25
Uh, and and it was like, "Why don't why don't you take a video of a full lap?" >> Yeah.
40:30
>> So that and I'll analyze it for you.
40:30
And I was like, "Yeah, yeah, dude.
40:33
I bet I bet you'd love to just chill back and watch track >> hang out and and watch track footage."
40:41
>> Pretty soon it's going to be like, "You want me to just get in the seat?
40:43
[laughter] You might just take over cuz like I would really >> Yeah.
40:45
No, computer use in iRacing is something I'm I'm going to experiment with this week. >> I'm down. I'm down.
40:50
Take half my quota, my monthly quota. Just play games. chill, do whatever. It's a nice treat.
40:57
>> I've got some bananked resets that I'll put to work. >> Yeah.
41:00
Anyway, we have our first guest.
41:02
Um, I don't want to mispronounce Tyler.
41:05
How do we pronounce his name? >> Tiss. >> Tiss. Let's bring him in.
41:08
How do you pronounce your first name? >> It's Ty. >> Ty. >> Like nice. That's right. >> Like nice.
41:16
>> Well, welcome to the show.
41:16
Thank you so much for taking the time. >> Great to have you.
41:20
>> Uh, tell us about your your role in robotics and also some of your recent work.
41:24
uh hooking these models up to robotic tools and and infrastructure.
41:27
Uh it feels like we're going to be entering a boom of like people they ordered the Mac Mini recently.
41:36
People are going to be ordering 3D printers and robotic arms and doing hack projects.
41:40
Super excited for like the DIY world to explode over the next couple months.
41:45
But uh let's start with just like your most interesting projects recently.
41:50
Yeah, I'm incredibly excited uh for that and and that world's definitely like just around the corner.
41:55
So, I'm I'm super excited about that.
41:57
Um yeah, I work here on the robotics team at OpenAI.
42:00
I'm an intern and have been doing lots of exploratory projects.
42:03
And this is just one of the projects I I I basically noticed that our models were really good at painting and doing various like computer used tasks in software like you were just talking about and learned a lot uh that you could actually basically connect these into physical robots in the real world and uh wanted to see how they would do at sort of drawing and tasks like that as well.
42:25
Yeah, if you can if you can paint in Google calendar with calendar invites, you can probably translate that uh to to the real world.
42:33
>> Yeah, it's really like everyone is doing everything.
42:35
Like I've seen people paint the Mona Lisa in Microsoft Excel or Google Sheets and then you t take it back and you can do Excel in MS Paint now if you want with computer use and it's like everything becomes everything.
42:48
Um what what what is actually the the important like precursors to a good experience?
42:54
experience? We were talking to a YC founder who built a a humanoid robot with some basic claws for like under two grand and it feels like there's there's an importance of some ondevice you know API or some ondevice models in some case doing some slam ondevice uh but then there's also just uh tools that
43:16
give you a very primitive interface that might be kind of clunky but it doesn't really matter because you can just vibe code your correct interface but what what have you liked what are you excited to to put in the repertoire of tools >> like within robotics and >> yeah within robotics. >> Yeah. Um I think that of course like a >> Yeah.
43:33
Um I think that of course like a lot of these demos and stuff is very early.
43:37
I think like the whole space and exact modeling and methods are still definitely being figured out.
43:41
Um, but I did really like the fact that you're sort of able to hook in the intelligence of like what the models are able to do as you can see with all that painting stuff and plug it into something physical in the real world and get it to do things.
43:53
Um, there's probably going to need to be like some combo right now like uh running basically my experiment here is is plugged right into codeex.
44:02
Uh, and that's of course probably quite expensive.
44:04
Um, definitely pretty slow.
44:04
I think these paintings took between 1 hour and like 2 hours depending on uh the methods exactly that the model like chose to use here.
44:15
Um but yeah, there's definitely a lot of improvements to make, but it's quite cool to see what you're able to do already. >> Yeah.
44:21
How are you balancing trade-offs between speed and reasoning effort?
44:23
I was I was I've been testing computer use on a bunch of video games that I just mentioned and uh there's like the you know like part of what the feel the AGI moment is is when the cursor is moving at like at least near human speed.
44:40
Uh and but then also you don't want to be making a bunch of mistakes and just uh so that's this like balancing act is really key.
44:49
Did you try multiple reasoning effort levels across different paintings and see like noticeable results?
44:55
Can you see qualitative differences in speed and quality based on models that you pick?
45:02
>> Yeah, there's like a lot of different ways to sort of go about doing this.
45:04
Um, in I initially started out these experiments by just telling the model like here I as you can see I actually have the bot behind me over here.
45:11
It's currently painting a TVPN logo [laughter] >> I love it.
45:19
>> But it has it's it's doing its best. >> There you go. Good start.
45:22
We got basically this camera up here that the model uh which is connected right into the codeex over here is connected to. >> Yeah.
45:32
>> And the biggest challenge is that images are like quite large. >> Yeah.
45:37
>> Uh to process especially like for the model there there are a lot of tokens. >> Yeah.
45:41
>> Um >> and the loop of like taking an image doing taking an action taking another image taking an action.
45:46
That's the thing that has taken a lot of time.
45:47
that has taken a lot of time. So what I ended up doing like it's not directly doing like image like one small movement image one small movement and I think like that would probably result in the best performance for like general
45:58
robotic tasks and stuff but like at the moment it's basically taking an image uh writing a plan in code of where like what it should do and where it should go next after it has done like a lot of the calibrations maybe a slower sort of methods at first. >> Sure. And then after it has that plan, >> Sure.
46:10
And then after it has that plan, it executes it and then monitors it in the background like taking images every second or few seconds and watching like to make sure things are going well and like making a small adjustments to the plan throughout.
46:21
And that I found that has been like a good balance of speed here.
46:25
>> Have you been uh thinking about like compression on the input?
46:28
You me you mentioned that the images are lots of tokens.
46:32
I I I have this monitor that renders at like 5K resolution and I was like >> this is probably going to be really slow for computer use.
46:39
I should like run this game in a window and give it like a 720p input because that's probably enough information, but I'm wondering like how important the the the resolution to speed and quality is that you've seen.
46:54
>> Yeah, I think it's very important.
46:54
Um, in this case for like a lot of the early demos, I was using like 512p just compressing it just to make Yeah.
47:00
make sure because I think the model is just pretty much uh it's pretty much solved like a lot of the perception challenges here.
47:07
it's like really good at uh understanding what's going on even with like a lower quality photo here.
47:10
Um but in this case like with the longer planning and and sort of like it plans sort of a minute of action out uh it's less important the like specific resolution probably but yeah as we go to more like faster much more smaller loops of like control is definitely going to be important.
47:28
Jordy, >> I I'm [clears throat] super excited about about this project mainly because it feels like we're right on the precipice of of sort of like what what I what I think is going to be a big breakthrough is like a sort of a deep research moment for robotics like simple robotics use cases where deep research for so many people was their first time using agents.
47:49
And the idea that you could type out a prompt and then get back what would have been maybe, >> you know, at least hours of human work, right?
47:59
Somebody reading all these different sources and combining that information into a document that's uh that that has a consistent narrative and and understands the the right information.
48:09
That was just such a big moment because a lot of people were saying, "Wow, I can't believe that that uh that the AI was able to do something that would have otherwise I would have hired had to hire somebody to do or just taken a bunch of time."
48:19
But there's like very simple tasks that I feel like you could probably start working on sooner than later, which is like an example, at least for me, would be like if I could just take all the mail I get and dump it in front of a robot like that and have the robot sort it. >> Interesting.
48:36
shred, you know, take all the, you know, 50% of everything I get is probably some sort of advertisement.
48:42
So, like figure out what's an ad and shred that and then uh actually, you know, basically like photograph and respond to if I have like a utility bill or or or any any number of things that I actually need to respond to.
48:56
Theoretically, you could close that sort of like IRL to digital loop where the agent would like actually get a task from the real world and then close that loop online >> uh with with just normal computer use.
49:07
And like that's the kind of thing that you don't need the you don't need like a $50,000 humanoid robot.
49:11
You theoretically could have a actual desktop robot that was able to do this thing that otherwise takes me I dread going and like okay I have to like sort through all this mail and figure out what's important and make sure I don't miss things.
49:24
Uh, but I feel like there's a bunch of other use cases like that where people are like, "Okay, >> I didn't just generate a pretty picture or or answer some question that I had.
49:31
I like actually saved myself." >> Spam filter. >> Yeah. An IRL spam filter.
49:36
The spam spam filter robot. >> Yeah. >> Yeah.
49:41
I think there's like two really really cool things about this project which sort of shows that direction that things are going.
49:46
One is that this arm I don't know if you know about like too much about the prices of like classic robotics equipment.
49:51
It's like thousands and thousands of dollars >> um >> at the moment.
49:55
And like this arm that I'm using here, like right behind me, this is a hugging face so 100 robot, which is like an open- source, fully 3D printable.
50:04
You just need to get the actuators um which are like much cheaper I think at the moment, which is I think more expensive just because of supply chain issues.
50:11
It's like around $200 or so, but like that's like incredibly cheap for robotic equipment and what you're able to do with it.
50:17
you're able to do with it. So like I can see a world like quite soon where similar to and someone put this really well on Twitter similar to how there was this whole 3D printing craze where everyone went and bought 3D printers and ran software like everyone's going to buy these uh sort of cheap before uh we
50:31
get like really industrial equipment for like personal use and personal product like buy these uh plastic cheaper uh robot arms that you can just like sort of clip onto a table and just put stuff in front of and plug them into like agents that are already you can already do things with that are like already out there. like Astra is something that
50:46
like Astra is something that people can just pull up codecs and start controlling uh robots with right now which is super sick.
50:51
Um and we've also been seeing like at least on Twitter I've noticed a lot of like sort of academic researchers at at different institutions start to like realize the that you can do this with these models and start doing like uh initial explorations and things with it which is super sick to see.
51:08
Uh and there's so much more space.
51:09
There was a YC company on the show yesterday that has managed to build a a humanoid robot for under $2,000 and go and that feels like a price point that people would experiment with.
51:21
So I if I if I know I can get a robot for like two grand and connect it to codeex >> and then tell it >> like it basically lowers the stakes a lot where I can be like, "Hey, go every weed like this that you can find in my yard, go like pluck it out and and try to put it in a in a in a bag or whatever."
51:39
And like that that's basically like the gardening robot and it could like flounder and fail.
51:42
But I have like pretty high confidence that even right now >> Astra would be able to identify like hundreds of like a specific type of of of weed and probably and there's of course like safety safety concerns with that. But yeah. Yeah.
51:56
The the Nat Freedman like leaf robot like that feels like we're we're here. The models can do it.
52:03
It it it's not it's not cheap.
52:03
But uh we're >> What is uh what is the next medium that you want to explore?
52:10
Are you going to get into whittling >> ceramics?
52:13
[laughter] I want to see you whittle a bench as a benchmark. Whittle bench.
52:16
[laughter] >> That's really good.
52:20
>> But whiddling a spoon or something. I don't know.
52:21
It just feels like 3D is the next thing.
52:23
>> You really want to give the robot knife.
52:25
John wants to give the robot a knife.
52:25
I don't think that's I don't think. Read the room. Read the room, John. Read the room. [laughter] Okay. Okay.
52:31
Maybe we'll stick to uh to >> inter OpenAI intern gives robot >> play. We'll do some ceramics.
52:38
Maybe get it on the pottery wheel, make a nice vase, something like that.
52:42
But I mean, there you're like, "Oh, yeah.
52:44
Okay, this is where this goes next."
52:48
>> I think I was having the exact same thought, which is like giving a robot a knife is like probably a very bad idea.
52:53
Um, but I I do want to like sort of see if I can help do like cooking tasks or do uh various tasks that are like things that you do in your life.
53:00
That would be really sick if you could have robot help you out here and there. >> Yeah.
53:04
Yeah, that's interesting.
53:04
Um, yeah, I I I'm so I'm so interested to see the because there's a whole class of tasks where you can't wait a full minute.
53:14
I was testing on like a real-time strategy game and you can pause the game, but if you're not moving at a certain APM, even on easy mode, like you will just get smoked.
53:23
And so, uh, but but it feels like with new chips and cerebrous and spark models and stuff like the the the speed up is going to come, but it's just and that's going to unlock a whole new knowle whole new host of of capabilities.
53:36
Uh, what what what advice do you have for for young people that want to get into DIY?
53:40
uh you know like this this type of work.
53:44
I mean you mentioned that one hugging face uh uh device that you have behind you.
53:47
Uh are there any other uh devices or tools toolkits that you recommend as places to get started?
53:58
>> Yeah, I think that the hugging face robot is like an incredible tool.
54:01
Uh you can also 3D print like a completely new embodiment and stuff.
54:04
There's lots of open source projects online where people have like changed it around to get better grippers and things you can do with that.
54:10
I just like also just don't give up after the first attempt.
54:12
Like this was the very first painting that the the Oh, that's upside down even. I couldn't even tell.
54:17
You can see the Golden Gate Bridge here.
54:18
You can kind of see the ground it was trying to do.
54:22
>> Um and just like if you keep going and and you can you can sort of see the progression as it as it improves. >> Wow, that's amazing.
54:28
Yeah, you need to frame those next to each other.
54:29
That's that that's incredible.
54:31
Yeah, I'm thinking of putting them all all all five of the progression in like a frame and calling it self-improvement cuz yeah, the model this was this was a thread.
54:40
The model just was able to figure things out with a few pointers here and there.
54:43
Uh how to make get better and better at painting.
54:47
>> Who's who's going to sign it? Is it you?
54:49
Do you have codec sign it? Astra sign it?
54:52
Who >> I've also given the robot I've given the robot a a pen and it's going to try.
54:56
I I don't know how well it's going to do, but we'll see.
55:00
>> Just an axe or something.
55:00
Uh, well, thank you so much for coming on the show. >> So cool. >> Very, very cool. >> Come back on soon. >> Yeah. >> Yeah.
55:07
>> I'm feeling I'm feeling the physical AGI. >> For sure. For sure. >> Very cool.
55:10
>> Have a great rest of your day. Goodbye. >> Cheers.
55:12
>> Let me tell you about public.
55:12
com investing for those who take it seriously.
55:16
They got stocks, options, bonds, crypto, treasuries, and more with great customer service.
55:20
And let me also tell you about Cisco.
55:22
Critical infrastructure for the AI era.
55:24
Unlock seamless real-time experiences. New value with Cisco.
55:29
We have two guests with us in the TVP and Ultradome. How you guys doing? >> We're good. >> Welcome to the show. >> Partners. >> Yes, >> partners. >> Thanks for having us. >> Yeah, welcome.
55:40
Uh, introduce yourselves. Introduce yourself.
55:44
>> My name is Guy Siri um from Maverick and also Sound Ventures. >> Yeah, welcome. >> And I'm Alex. Yeah.
55:49
>> Now with sound, >> also running sources still. >> Very excited.
55:53
>> Yeah, >> it's been a big week.
55:54
>> No, you don't know how big it hat on your head at all. No hat.
55:56
Right now we we were debating our >> every time for the last year every time I've seen you have the capital >> capital Journalism hat on. >> It's off. >> But it's off now. >> It's off. Okay. >> Yeah.
56:08
Uh but getting him here guys journalist. >> Um >> 10 years. >> 10 years. >> 15. Yeah.
56:15
Started around when I was in high school. So yeah.
56:20
>> And then now Capital Venture capitalist. >> Capital V. >> Right. Capital I investor. >> Yeah.
56:26
It it feels we're we're aligned on the vision and uh >> really excited to get started.
56:30
He's this guy's a force, but also >> a talent. >> Yeah.
56:36
>> You know, like you guys, I mean, there's you guys you guys understand how to work with people and talk to people and help them tell their story.
56:42
>> And I think it's so aligned with what I've been doing my whole life as well, which is helping people tell their story.
56:48
>> And so when we got together, it was it was just magical. >> Yeah. Yeah. It's awesome.
56:51
What are you interested in investing in?
56:56
We're spending a lot of time on on different things. Um personal agent space. Okay. Very interested in. It's very hot obviously.
57:03
>> Are you a daily driver of anything yet? >> I'm using everything. >> Everything. >> Uh yeah.
57:08
>> What's the last agentic thing you did? Did you book a flight?
57:12
>> Um >> did you email the CEO of Walmart for a refund on $5 of raspberries? Did you hear about this? >> No. >> Oh yeah.
57:19
Some of these agents are very persistent.
57:21
There's very p persistent personal agents where you know a lot of like one one of the seemingly now very obvious use cases of agents is just like hey like do a bunch of things that would take me a lot of time that could maybe save me some money right and when you're using a free agent you don't care if it's spinning its wheels for for 24 hours to get you a $10 refund if the ref if it's not costing you anything.
57:41
So apparently somebody was trying to get a refund on a $5 pack of blueberries they got at Walmart. It >> was raspberries. >> Raspberries. Raspberries.
57:51
And the agent actually reached out to the EA of the CEO of Walmart trying [laughter] this.
57:59
So that was like the last place to escalate it is like I got to take this right to the top.
58:03
>> These agents are getting crazy. They're getting crazy. They're swarming.
58:07
>> But but but I guess like re Yeah.
58:08
Rewinding a little bit guy.
58:08
you invested in OpenAI and Anthropic like years ago and so >> and hugging face. >> Wow. >> And hugging face. I didn't know that. Nice.
58:19
Um >> and so I in some ways like you probably the last few years have been just sitting back in a just basically just getting to like getting to uh experience your own conviction and and and watching the space evolve.
58:35
But I think like everyone has got to the point in the last or at least has consistently been feeling like nothing is like settled yet.
58:43
We have these new kinds of businesses, labs, some of the labs, you know, are making products, but there's still tons of room for other players to come in and make things.
58:53
Um, so are you feeling like renewed excitement around [clears throat] early stage?
58:59
>> When when we did Enthropic and OpenAI, I think we were the only fund that that went in. so deep back then.
59:07
>> Um, >> and it was confusing to some people, but to to to us it felt like >> this was it. This was the time.
59:13
These were the companies.
59:14
These were going to be long-standing foundational uh uh platforms.
59:20
>> And today, it's a lot more confusing.
59:20
I I you know, there's so much going on >> and every single week, every single day, you guys are announcing people's raises. They're raising now. They're raising now. They're raising now.
59:30
And >> it's hard to tell.
59:31
You know, it's not as easy as it was for us to to really decipher these are going to be the things people use in 10 years.
59:41
>> And and now it it feels like there's so much going on, but I'm also as excited. I'm also as inspired.
59:48
I I don't I want to be part of of these exciting companies.
59:53
I we just have to pick, right? Yeah.
59:55
>> Um I'm I'm meeting with some really incredible founders and um visionaries.
1:00:01
It feels really exciting.
1:00:01
It's so it's like when it's starting in the music business. >> Yeah.
1:00:06
>> It's like the early days where I got going and you're just getting demos everywhere.
1:00:09
You know, you're walking out of a club and you're like, "Hey, hey, you're the guy at the here's a demo. Here's my demo. Here's my demo. Here's my demo."
1:00:14
Like, how can I tell which one of these artists?
1:00:16
But if you just if you put enough work in, enough time in >> and you're diligent, they starts to like they start to become a little more obvious. Yeah.
1:00:25
>> And and there wasn't initially I had a hundred demos.
1:00:27
I literally I was like 17 years old with a 100 demos.
1:00:29
But as you listen to a hundred my first three demos >> at 17 they were you >> these were people that I would given them to you. >> Yeah.
1:00:39
People that I would go out and go let me hear your music hear your music.
1:00:42
And you know the first three demos I had I had my favorite one my favorite two favorite like in in order >> but 100 in those three are not even in the top 20 right?
1:00:50
So you have to just you have to you know pattern recognition.
1:00:54
You have to do a lot of work. You have to listen.
1:00:56
You have to meet a lot of people.
1:00:57
And then through this crazy time there's a it's it's pretty freaking crazy.
1:01:02
I think the right things appear and then and then you just have to be there to be part of that. And so still is excited. Yeah.
1:01:09
I have we have been able to sit back a little bit and watch our our you know we we've also not we didn't just invest in those companies.
1:01:16
We also invested on the way up. >> Yeah.
1:01:18
>> So that also keeps you busy.
1:01:18
We put a lot of money into anthropic on the way up.
1:01:22
We put a lot of money into OpenAI on the way up.
1:01:23
And um I think we have close to a billion dollars worth of of of money uh invested into those two companies.
1:01:31
>> Um and so so we're not just like laying back, you know, but but it is a lot harder today to decipher between what is real and what is not.
1:01:39
real and what is not. make when you make you know when you have like two effectively recent investments that are now two of the most important companies in the world I do feel like the bar goes up on other on other investments because it's like it's suddenly is like is it is
1:01:56
it as thrilling to invest in a company that that can only be a10 billion company right when when in when >> a lot of VCs were very excited to underwrite a company to 10 billion >> you know six years ago go even even during the period that that you were making those two investments. >> Yeah, I I think about it differently. I
1:02:14
>> Yeah, I I think about it differently.
1:02:14
I I have heard some people say, "Hey, zero to 100 that's like it's not a big deal anymore, you know."
1:02:20
Um I just like I I I've always been attracted to talent and and visionaries.
1:02:30
So, I don't start with, okay, this, you know, I we're we're fortunate to be in the these two incredible companies, but there's a lot in between and there's a lot that's to come and I I just love sitting with a founder and >> problem solving and figuring out how we're going to get from A to B.
1:02:47
And sometimes it's sometimes it's where you get in.
1:02:50
I've saw a lot of people, you know, uh John from, you know, Beta Works did really well on Hugging Face.
1:02:54
you know, we came in, we did well, but he came in where, you know, at seed, so he did really well.
1:03:01
And um sometimes it depends where you get in um as as well.
1:03:07
But for me, what excites me is the same thing.
1:03:09
It's been constant my whole life, which is surround yourself with really really incredible, brilliant people who are trying to change the world.
1:03:16
How does uh identifying creative or musical talent differ from uh startup entrepreneurial talent?
1:03:27
Because I'm sure there's some some common threads, but whereas in music, you might back a musician that's that that has like uh you maybe know they have a drug problem and and and that's part of part of the music.
1:03:37
But in in in startups, you know, a founder that's like has like some crazy crazy crazy stuff going on in their personal life, maybe it's like, hey, you should figure that out before you build a, you know, massive team and you're managing people and stuff like that.
1:03:52
Um, but there has to be like a bunch of common common ground between the two.
1:03:57
>> Well, I was able to transition seamlessly because of music.
1:04:03
um my job was to identify um artists before anyone had heard of them and to sign them very quickly and to then help them reach an audience.
1:04:13
So when I meet a founder I I I actually feels the same.
1:04:18
I I always say founders are the rock are the rock stars too because when they walk in they also have their music that they want to share with the world.
1:04:25
So I have to identify that founder same way I used to identify music artists. >> Yeah.
1:04:30
and then go, "This guy has or she or whoever have music that is so good. Oh, I love that course. That's a great idea.
1:04:40
You mean a car shows up and it picks you up and it takes you or you know an apartment people could share and you're like, "Oh, wow. That's a hit song." >> Yeah.
1:04:48
>> So, I I I always listen to um every pitch like it's like like it's a song or or an album or a music artist and I just go, "That guy's got the talent. He's the rock star.
1:04:59
We just need to make sure the world knows it.
1:05:01
We need to make sure that people are aware of what he's building. Let's go get a base. Let's go.
1:05:07
Let's go create find that audience first and tell this story.
1:05:09
And so for me, it's I I feel like I've been doing the same job since I was a teenager, which is, you know, identifying talent and helping them helping them reach an audience.
1:05:18
But music is the constant.
1:05:20
I'm always listening for the chorus. >> Yeah.
1:05:23
>> You know, and if I don't hear the chorus, I'm like, I'm not sure about the song.
1:05:25
song. this is uh you know or the performance of the song or I don't think we can I don't think this one work you know so so it all it always comes from that the DNA is is is being around uh music artists >> is that uh in in in tech world you see
1:05:45
entrepreneurs that are truly visionary as in they're seeing opportunities before they are obvious and pursuing those and they have an idea of the way that the world should be and they're trying to sort mold the world into that state. Uh and then you have actually the
1:05:57
Uh and then you have actually the majority of entrepreneurs which are just like they don't know something's an opportunity until they see someone else pursuing it and they're like that seems like a good idea. I'm going to do that.
1:06:09
is same thing in music where you know you have somebody that has like truly a new sound and they have a life experience that they're trying to >> like they need they feel like they need to create >> art out of and then there's the the follow on of like
1:06:24
>> you want to cover band entrepreneur [laughter] >> yeah cover band entrepreneurs that might be a good >> when we started the record label you had Jimmy Ivan here the other day >> um when we started the record label it was just like four of us small companies Madonna's company. So that's cool. But So that's cool.
1:06:39
But was still there just no one knew what to make of it.
1:06:44
>> And >> Jimmy was on fire.
1:06:46
Innercope Records was on fire.
1:06:48
And and it I was always thinking if I don't act quickly, he's just going to pay the more and get them. >> Yeah.
1:06:55
>> So, and we were competing with big labels, but Jimmy was the the guy who I always looked at like, okay, he's he's he's he can he can just come in here and just >> Yeah.
1:07:04
>> wow them and get them.
1:07:04
So, not only do I I have to I have to hear your song >> and decide right then and there >> I want to do it.
1:07:12
So, I don't have any background.
1:07:14
>> I don't have Oh, they Oh, you know, my biggest successes were always the things no one else wanted. >> Yeah.
1:07:21
>> But, uh, you know, uh, Alanis Morset um, I mean, she tells the story where every single label passed.
1:07:27
I didn't have any of that history.
1:07:29
>> She came in, she was in my office with her producer Glenn Ballard.
1:07:33
>> They played me one song. which is called Perfect.
1:07:35
And within 30 seconds or 40 seconds, I think I was like, I'm in.
1:07:38
And and and and so that Muse, you know, the band from England, they came to to LA >> and I flew them in because I like their demo.
1:07:49
>> They flew they they did one after their there to perform a few songs.
1:07:52
After the first song, I stopped them.
1:07:53
I said, "We're ready to go."
1:07:54
And they're like, "We flew all the way from London.
1:07:55
Can we just play out the next few songs?"
1:07:57
I'm like, "Of course, but I just want you to know."
1:08:00
So, I developed that that actquick intuition and it really just came from >> I had to or else someone else would just figure it out and overpay and then I couldn't do the deal.
1:08:12
so many of the the market dynamics that you see in music.
1:08:16
Uh friend of mine uh Zach Beio was telling me about like some of the process of signing the artists that he works with where these artists are like you know the same thing that happens on on like with in in tech where like some X account pops up and maybe there's like a team attached to it.
1:08:34
There's no launch video yet, but you see a bunch of people following this person and then you hear that they're meeting with this firm and this firm and that firm and and and the whispers start going around.
1:08:44
Same thing in music where like an artist might one day have no followers on Instagram, be totally totally under the radar, living like in their parents' basement, but they have some little bit of magic.
1:08:56
And then soon enough they're like doing a road show basically with different labels.
1:09:03
And then you as a label need to be like, well, how much can we how much can we invest in this person?
1:09:07
How big do we want to bet?
1:09:09
Uh we need to get to them.
1:09:11
We need to get to them first.
1:09:11
But and then you're also sometimes competing on price, but other times you're just competing on like, well, how how great a partner can I be to this to this artist.
1:09:23
So like I can see how music translates just so well into venture because the exact same thing. Venture is not a game.
1:09:31
Once somebody's talented and they're known, it's like very obvious.
1:09:34
Obviously, you want to be >> it's harder to get in then. >> Yeah.
1:09:37
>> And it's the same with I have competed when things are big.
1:09:39
I remember when Prodigy, everyone wanted them and I I flew to London like in twice in four days to try to get that and I got it. >> Yeah.
1:09:48
>> And every VC will have a story like that where they're like, I flew to this back place. >> We have those.
1:09:53
We we we have all of those.
1:09:55
But you know, we [laughter] you know, you look at um when we did Enthropic, >> when we did we did SPVS and Enthropic, we couldn't a lot of people were not >> Hey, you guysough. >> Yeah. tough to fill.
1:10:08
A few times people didn't get it.
1:10:10
Of course, now you know we're we're begging to get more of it.
1:10:14
So, >> I just again I always go back to don't listen to anybody.
1:10:18
Um you know um you know you talked about I think the other day also you talked about blinders.
1:10:23
Did you talk about blinders or something on the show? >> Yeah.
1:10:26
Jimmy I has that concept. >> Yeah.
1:10:28
So that horse blinders blinders.
1:10:30
>> We have a a mutual friend and he's my mentor.
1:10:32
His name is David Geffin. >> Yeah.
1:10:34
>> And David said to me when I was like 21 Yeah.
1:10:37
Um, he told me that story. I didn't know this.
1:10:39
I He said, "You know, guy, you need to be a racehorse."
1:10:43
>> And And I was like, he goes, "You know what raceh horses do?"
1:10:45
And I go, >> "Yeah, they race." I I had no idea. >> Yeah.
1:10:50
>> And he goes, "No, no, they wear blinders." >> Mhm.
1:10:53
>> And so just race your own race cuz if you don't wear your blinders, if if horses don't wear blinders, they could literally they could kill they could die.
1:11:00
They could trip over their they look over and they could trip over.
1:11:02
They could break their legs.
1:11:03
And and so I really stuck with me that you guys talked about it that really that that concept stuck with me and I I really try to just not pay attention.
1:11:12
You know, when we did anthropic and open AI, a lot of people doubted it and and um we you know, we we didn't have any doubt.
1:11:23
>> We were we were determined to do it and and um I I I want to stick with we're we're we're I'm really trying to continually connect to that >> approach of not listening to all the noise.
1:11:33
Of course, data is important and we want to get more details and more information and we're structured, >> but that gut that has gotten me here, uh, I need to continually respect.
1:11:46
>> Alex, on your side, you've spent, uh, >> how is it 15 years, a decade, 15 yearsish?
1:11:53
Uh, >> did it take did do you feel like it took a few sort of cycles to hone your your intuition around around companies?
1:12:02
Because in in our first conversations, um I was always impressed with your ability to just see directly through the marketing [laughter] on so many different companies.
1:12:11
Like some people like marketing just works on them.
1:12:14
Marketing works on all of us.
1:12:16
Advertising just works period.
1:12:16
But like marketing works, marketing and good comms work like too well on us.
1:12:22
>> Or some people that are just like this is a good story.
1:12:23
So I'm chasing it from the capital.
1:12:26
the capital. But for you, you'd be like you would I would I would we would be talking about something and you would >> be aware of like a dynamic around a company that no other journalist had talked about and and at times like you would be like yeah story is not for me but you were like clued in on a on a story and you knew exactly what was going on with the company and then and in in in the example I'm thinking of I
1:12:49
won't won't name the company only >> 6 months later did it did it has it even started percolating up that that that dynamic is going And so I feel like for me and and at least personally like I had to see the cycle of like company like starts gets
1:13:05
hots a bunch of capital but sometimes you have this intuition around the company where like something doesn't really like feel right about this company even though it has a lot of momentum. >> Yeah that's it. Yeah. I don't I don't >> Yeah that's it. Yeah.
1:13:14
I don't I don't know where that comes from except that I've I've been fortunate to spend time with a lot of like the best founders in the world.
1:13:22
I mean I just had Zuck on the podcast Sam Alman before that.
1:13:26
incredible lineup coming up. Yeah.
1:13:28
>> Uh and I've gotten to know these people over years and years and years.
1:13:30
So when you see like the people at the apex who are crushing it and who are have high integrity, are beasts at the game on the field, >> um you you can quickly see when someone is pretending. >> Yeah.
1:13:43
>> Uh and I just try to stay really close to like what's actually happening and ask around, do my diligence.
1:13:48
Um leave no stone unturned.
1:13:52
And that's got me well so far.
1:13:55
But like the thing you said about like 6 months later you saw it.
1:13:56
I have that a lot where I'm like, "Oh, this seems really interesting.
1:14:01
This seems like everyone's going to be talking about this and then it happens and it's happened enough and times to where I'm like, "Okay, >> you got to figure out a way to make some money on it." >> Well, yeah.
1:14:11
And like what guy was saying about his gut and this is where I think we really hit it off is yeah, you have to trust your gut.
1:14:16
Like if you can get enough pattern matching recognition in, um, it's just instinctual.
1:14:23
How do you think the podcast will evolve in this new role? >> It's going full tilt.
1:14:28
I mean, uh, first two episodes again were were Sam and and Mark.
1:14:33
Uh, I can't share the names, but, um, it's it's going it's >> I just mean like like there's there's interesting ways when you have position in a company.
1:14:44
The the critique is always going to be like, oh, they're only having them on the show because they have a bag or whatever.
1:14:48
But when I look at like what Dwarf Cash has done with Maddx and Rainer, like explaining his expertise, it doesn't feel like a sales pitch for that company at all.
1:14:56
It's actually just tapping the network at a deeper level.
1:15:00
And at the same time, if I'm a founder and you're the place that I go to hear Mark Zuckerberg talk about his vision, that adds value and attracts even if it's not a company that you're actively investing in because you're not doing Publix.
1:15:15
Um, so I'm I'm wondering if there will be more like less like this person's on the funding track or more like 360 views.
1:15:24
Do you want to climb the mountain and do all the mag seven CEOs?
1:15:28
Is that the goal or is it more like go deeper with certain experts, build this community of people with particular philosophy?
1:15:35
There's like a whole bunch of different ways I could see it evolving and I'm wondering if you have any particular direction.
1:15:40
>> The Max 7 I feel pretty good about.
1:15:40
Uh, >> you're on that track for sure.
1:15:43
It's Yeah, I feel great about it.
1:15:45
Um >> I always love the like I love I love putting the people at the top with the people who are upandcomers for what I do.
1:15:53
>> So, uh I've got two and from base 10 on next week. >> Amazing.
1:15:57
>> Like >> legend >> uh legend, incredible company.
1:16:00
>> Uh he's incredible and >> you know he's he's obviously crushing.
1:16:04
He's he's huge, but like he's not he's not Zuck yet.
1:16:06
[clears throat] >> So, but I I want to bridge that world because like this is all one world we're in.
1:16:11
like everyone talks like the Zucks want to know what the Tins vice versa want to learn from each other.
1:16:17
So I like I like building that cinematic universe and it's like my taste.
1:16:20
It's like I wanted to have that convo with Tuin which is coming out next week because like >> inference is just so important right now and it's like everyone's trying to figure it out. >> Yeah.
1:16:31
>> And so you're also getting my POV of what I think is interesting with my guests and I'm booking everything myself. >> Sure.
1:16:36
Um, and I think that's only going to get better because again like sources is separate.
1:16:41
I mean obviously I'm with D and I'm with Sound, but you know we'll have I'll have people on the pod that you know are we're we're not investors in.
1:16:49
I'll have I'll have other VCs on >> um it's about the ecosystem. >> Sure.
1:16:55
>> Um and I I think the brand is important.
1:16:58
Like I want to invest in the brand.
1:16:59
>> What's the future of the writing newsletter?
1:17:01
I imagine that you're not gonna be able to put the pen down forever.
1:17:04
Like, >> well, yeah, it's interesting.
1:17:07
>> I mean, there's gonna be times when you just want to get something out. >> Yeah.
1:17:09
I think I think, you know, I want to use the newsletter, which has just an incredible audience, to to share >> what I'm seeing.
1:17:17
And it's really like you're getting like an even deeper >> sense of what I'm seeing because now I'm like in the room in a way that I was kind of in, but I was always like when you're a journalist and you're in the room, you like get brought into the room and then escorted right back out.
1:17:30
[laughter] And now it's like I get to hang out in the room.
1:17:31
And so you're I'm like I'm sitting with it and I'm marinating on it.
1:17:35
And so when I write like a a piece like I'm thinking on something on personal agents actually right now.
1:17:41
>> It's informed by a lot of conversations.
1:17:43
I'm not going to share all those, right?
1:17:45
Like obviously confidentiality is very important.
1:17:47
>> But I think it's going to make the perspectives I'm sharing a lot better.
1:17:50
But look, I you you started this like hanging up the the capital J journalism hat, right?
1:17:54
There is a sense of like journalism in the traditional sense. >> Yeah.
1:17:57
breaking ski of the and we we were talking about this when I was on the show before like the leaks and all the things that I've been known for over the years like obviously I'm not going to do that anymore >> but but you know what like I've done that for 10 years. >> Yeah.
1:18:08
If you have uh like I'm very excited to read this take on personal agents.
1:18:12
I think a lot of people have been thinking about the the way this category evolves.
1:18:14
Um do you have any interest in turning that into a video essay direct to camera talking to the camera putting it on the same feeds like what Daresh does when he writes an essay.
1:18:26
He also has a video version. >> Yeah.
1:18:29
>> Could just be a good product but also reach more people.
1:18:32
>> People have asked me to do that. >> I have another job.
1:18:34
>> I don't know if that makes sense for like other reports where it's like here's here's some facts.
1:18:37
It's more of a quick hit.
1:18:40
But if I'm like getting like a quarterly thesis from you like maybe is a 20inut video.
1:18:45
I I would watch that and it just gives me more optionality to like I can browse it in the email.
1:18:49
I can also like with Dwar I get his emails.
1:18:51
I also see them on YouTube and then I get in the podcast feed and sometimes I'll be in the video mood sometimes and I I'm multiplatform with a lot of these creators and so >> I need AI to help me with this guys. I'm going to be honest. >> Yeah.
1:19:05
I don't >> but but you want the rawness of the person which of course but like I think the pod is conversations for now.
1:19:10
Maybe maybe I branch it out. Maybe it's more things.
1:19:14
I just think a lot of the platforms are very receptive to uh multi-product feeds like like having multiple media products within a feed.
1:19:23
>> I thought a lot about this.
1:19:24
>> I've been surprised that we've been able to do it with a a 20-minute version of the show and a three-hour version of the show dropping in the same feed every single day and it hasn't been bad.
1:19:30
No, people just pick whatever they want.
1:19:33
And then if there's a hero interview, we get an, you know, big interview with someone that goes out as another one and that's its own thing.
1:19:40
>> This is maybe a phase two thing.
1:19:40
I mean, I have I have eight eight incredible guests lined up over the next few weeks.
1:19:47
So, it's like I got to get those out. >> That's great.
1:19:49
>> I'm in SF with guy next week doing some I'm doing four next week.
1:19:51
Like, >> so I want to get all those out and then and then yeah, maybe like the personal agents thing, maybe that's like a >> has the lens changed when if you think about a mag seven CEO, there's the the getting the scoop in the interview, the Capital Journalist interview in that conversation.
1:20:09
And I think that you can there's a bunch of different ways to do that.
1:20:13
But then there's also uh the you know what is it what will the next generation of great founders get out of this particular conversation with this mag seven CEO?
1:20:22
Are you starting to put on that hat of like >> I don't really think of it that way.
1:20:25
I think about like what do I want to know and >> and >> I really care about strategy.
1:20:31
>> I really care about strategy. I really care about connecting the dots like getting them to say something they've never said which you saw with you know the last two pods >> and that's still going to be the thing and that's that is journalism like that
1:20:44
that you know you're getting interesting in fact out in the world that good [clears throat] >> and like you know these people are are doing a lot and so it's like they're they're out there but like I I don't know I think I get a lot out of my conversations with them and I don't I don't think I'm going to change much. Um, I mean, I think the only thing is
1:21:01
Um, I mean, I think the only thing is like, yeah, you I'm not going to be like leaking memos anymore.
1:21:04
But, [laughter] uh, I used to do leaks about companies worrying about leaks, and that was one of my favorite kinds of stories.
1:21:12
>> I'm not going to do that anymore.
1:21:12
I'm not going to do that anymore.
1:21:14
>> One more leak in 2020, you get a good one.
1:21:17
>> No, a decade of that, and I'm I'm good. >> You're hanging it up.
1:21:21
>> Is the leak >> is now a good time to become the next Alex Heath?
1:21:25
Somebody's like 20, 20, 22. >> I think that's great. Yeah.
1:21:29
I mean, I think it's really hard right now if you're early because um it's just a the media the traditional media environment is so challenged structurally >> and uh places doing really well.
1:21:40
I feel like I feel like the the the ads product, the sponsorship product.
1:21:47
>> No, I'm saying traditional. >> Yeah. Oh, but okay.
1:21:48
So, he would start in the news. >> How would you start?
1:21:51
How would you how would you how would you you know I've been fortunate people care because >> I've broken a lot of big stories and I've gotten a lot of big interviews >> when you and I would came up in an environment where like I was in a newsroom learning from incredible people who've gone on to do incredible things and run now in many of these news rooms.
1:22:07
So >> I don't know how you do that now.
1:22:08
Like a lot of places aren't hiring.
1:22:11
>> Uh their traffic's declining.
1:22:11
They haven't made the pivot to like what we're doing this direct thing like subscriptions like streaming. I >> it's really tough. I've thought about it.
1:22:20
it. I don't know how you would break out right now unless you just kind of are >> you got to start >> you maniacally focus on one thing and become the best in the world at that which is how I started which is like I'm
1:22:28
going to be the best in the world at social media covering Snap back in the day during the IPO and then I was breaking a ton of news on Snap and then I got noticed and then I was like I can shift this into other companies and keep shifting it and shifting it. >> Yeah. >> Yeah.
1:22:40
>> Um so I would still say that's it.
1:22:40
You have to maniacally focus on one thing >> I mean defin but but it's the intersection of niche and matters. It can't be catch.
1:22:50
>> What do you think about uh >> the possibility of Instinct having a a bigger valuation than Snap?
1:22:59
>> I think I mean I wouldn't be surprised.
1:23:02
I uh it's crazy out there, guys.
1:23:04
>> It's also a new category.
1:23:04
And >> what do you guys think about Instinct?
1:23:09
>> I think that they're in a unique position because they're a startup.
1:23:12
So they can like like if there's like rough edges like those can get uh those can get ironed out and they there's more forgiveness I think as opposed to >> if you're a startup. >> Yeah.
1:23:24
As opposed to like the Muse agent is going to be like congressional hearing if it something goes poorly whereas instinct is going to be like look it's a startup you knew you were you were an early adopter let's give them the benefit of the doubt here.
1:23:34
them the benefit of the doubt here. And the other side yeah to me the the most interesting dynamic right now is uh because of the people's fear around AI there's like way greater willingness to try new products because you don't want to be left behind >> right and so you may somebody may have been trying and being a daily active user of a variety of AI products for
1:23:59
years now and still they're like they want to try the new thing because one the space is progressing there's so much room for new products and new categories, but then there's also this fear in the back of your mind of like if I don't try the new thing, then like I'll, you know, the the permanent underclass meme and that's just consistently created this sort of second mover advantage, third mover advantage. And then AI brands once they're big,
1:24:22
And then AI brands once they're big, they accumulate, they've been accumulating baggage, right?
1:24:26
And so people are like, you know, maybe they have some maybe they're just like they're excited to share and talk about the new thing in a way that that they wouldn't be even products that they're using day-to-day.
1:24:40
I think it creates a big opportunity for um for startups, but I think that we're I'm very interested to see how the personal agents market ends up comparing to just like the frontier model inference market because it seems like every company is going to build a personal agent.
1:24:59
personal agent. many of them already have especially if you count LLMs which do have agentic uh or just like chat apps which do have agentic capabilities but it's going to be an absolute it's going to be an absolute >> blood to network effect and you get to some sort of take rate on aentic
1:25:15
commerce like you buy your car through it and they make 500 bucks like that's a very very >> which Zach told me will be the business for Muse is a take >> he's trying to do and he has the network effect he's bringing to it so instinct certainly challenger >> there are I'm really interested in the idea of network effects with agents Yeah. >> And instinct's doing it. Town is doing
1:25:30
>> And instinct's doing it. Town is doing it.
1:25:33
>> Meta is going to do it.
1:25:35
>> Uh and maybe that is the next >> at the same time.
1:25:37
It's tricky if you can point an agent and say like get me off of this thing, right?
1:25:42
>> Well, they're not people.
1:25:42
So, it's like, do you care?
1:25:43
Do you care if your agents are in a network if it's >> But if you're like this one is the one that's never had a leak or never had a crash or never had a hack, then you do stick around.
1:25:51
In theory, like the time to build a new social network would be today because you could say like go through open up my Snap account or LinkedIn and scrape out every single person I have.
1:26:05
>> They don't get a say in it and go add them on this new network, right?
1:26:08
Because like that that was the export the contact book was like an arbitrage that closed and it's kind of opening back up. >> You think? >> I think so.
1:26:16
I'm I'm waiting for, you know, the the criticism of social media was always like we created social media to be social and it's made us less social than ever.
1:26:24
social than ever. And with personal agents, it's like it's like less well less like personal like we don't have we're not going to have personal relationships with like service providers and and variety of things because it's like >> even even people some of these new functionalities which is like sorry
1:26:41
grandma I don't want to talk about the road trip that we're going on just talk to my agent you know and and so the new criticism will be like we're no we're no one's talking to each other it's only agents talk you know we're communicating through like uh you know can on a string or whatever. Um
1:26:58
Um >> well [laughter] we we got to hop on with Natasha from Positron. Uh we >> this was great.
1:27:04
I'm super excited for you guys. Big big fan of of both.
1:27:08
>> We'll let you we'll let you hop. Yeah. Sounds good.
1:27:10
>> You guys are going to absolutely cook together.
1:27:11
[clears throat] >> Thanks so much.
1:27:14
>> Let me tell you about MongoDB.
1:27:14
What's the only thing faster than the AI market?
1:27:18
Your business on MongoDB. Don't just build AI.
1:27:20
Own the data platform that powers it.
1:27:22
And let me also tell you about Crowd Strike. Your business is AI.
1:27:27
Their business is securing it.
1:27:27
Crowd Strike secures AI and stops breaches.
1:27:28
We are joined by Mitesh Agarwal from Positron AI building a new chip for the AI era. Welcome. How are you doing?
1:27:39
>> Hey John Dhy, pretty good. How are you guys?
1:27:41
>> Thanks so much for hopping on the show. >> Good.
1:27:44
>> Uh great to have you here.
1:27:44
I I just want to start by saying um I've been in the background of Stephen doing this multiple times now and Stephen Steven Balivan from he loves he loves you guys.
1:27:55
So this is my first time on.
1:27:55
So thanks for thanks for having me. >> Yeah.
1:27:58
How how direct is the lineage from from Lambda.
1:28:00
You're working at effectively Neocloud.
1:28:03
You see the problem you go solve the problem externally with a new startup.
1:28:06
Is is the story that simple? >> Yeah fairly.
1:28:10
I mean for me I mean like look I I didn't found positron right?
1:28:14
was co-ounded by Thomas Summers and Edward Kamett. >> Sure.
1:28:17
>> Another lineage uh Grock lineage from from from before and and you know they they designed the actual silicon and and and the system setup and uh uh part of it is just like great luck.
1:28:27
You know I've known both Stephen and Thomas for over a decade.
1:28:30
I've been close friends with both of them.
1:28:32
I've worked with them, been roommates, everything all of all those things.
1:28:35
And Thomas has been uh wanting me to join Positron since day one since he started the company in 2023. >> Yeah.
1:28:41
>> Yeah. Um but lambda was just starting on its hockey stick growth then and I was like look I'm not leaving lambda uh started the company with uh Stephen there and lambda cloud but uh uh early 2025 man like late 24 you know reasoning models had come out when uh just started to to to get in the zeitgeist and then
1:28:58
video generation um you know Sora first Sora came out um not a lot of people saw it but I I got to see a little bit behind the scenes on on the video generation models the amount of memory I remember looking at the Google VO model back and you know it needed four H100s to run like a 10-second clip. >> Yeah. >> Yeah.
1:29:16
>> Was completely memory bound on bandwidth and capacity >> and knew what Thomas was building.
1:29:19
I was like this is actually very interesting.
1:29:23
Memory uh is going to get a big part of the story for inference.
1:29:26
Uh uh someone is building something about it.
1:29:28
is building something about it. let me go get and and and work at the at actually the fundamental technology layer like >> Lambda builds technology on the cloud and and services uh you know and I'm chemical engineering studied fabrication never used it ever uh so I was like all
1:29:44
right I'm going to go back a little bit to my roots and and come back to it >> amazing >> how much is AI actually accelerating semiconductor design semiconductor fabrication uh we saw one of your investors Dylan Patel at semi analysis talking about that the opening a jalapeno chip seemed like ahead of schedule or very very quick. Uh for a
1:30:06
Uh for a long time we've been hearing oh new chip that's 3 years that's 5 years feels like it's 18 months now.
1:30:13
What are you actually feeling? What are you seeing?
1:30:17
>> Yeah to start from reverse like to answer your last part about it like >> man new chip every 12 months.
1:30:21
Nvidia is the absolute king and if they're coming out with a new silicon every 12 months, >> you better get in that game or or you know like you like don't even be part of the conversation kind of thing, right?
1:30:32
So so that's that's for sure.
1:30:32
Uh in terms of utilization of AI, I mean look I I was just like focused on Positron itself, but you know we are a small team.
1:30:40
I mean um >> we we got to we we right just now I mean just yesterday or today crossed 100 people but over 50 of that is over the last three months.
1:30:48
So we got our first gen product out with less than 20 people.
1:30:53
Um and that's built on FPGA.
1:30:53
So it's already pre-taped out silicon and we we're deploying and implementing architecture.
1:30:59
And our second gen those are the 50 atlas racks you have at Oracle. >> Oracle. Yeah.
1:31:05
>> So those are interesting. >> Yeah. Those are FPGAs.
1:31:07
And it's kind of like hearkening back to a little bit of uh of previous times when you used to design and and build silicon.
1:31:13
You would actually test it on FPJ before going into the the tape out kind of thing.
1:31:16
So we just wanted to get a product out as quickly as possible.
1:31:21
Like that was the whole thing.
1:31:22
It's like you know from the start of the company we got our first shipment to a customer in 15 months and it was you know it was built on FPJs obviously but to get the full u >> you know bit file ready getting it deployed getting models running on it was all done in the first 15 months and then over the last time has killed it out.
1:31:39
out. Um but yeah I mean like look we have to use a lot of the AI toolkit especially on uh on on uh verification um design less so I would say I mean look obviously we use a lot to to kind of interact with like now Astra for example this phenomena right you know to to interact with it [clears throat] but
1:31:57
you're still not going there and saying hey like come up with this like new design yet although like you know anas recursive and others they they're obviously built out and you know they raised a big round for for that as well right so it's going to come you know you're going to see very soon. It's like
1:32:09
It's like one person uh in Astra or one person in Astra's kid taped out a chip kind of thing.
1:32:16
But uh >> we have to use it a lot.
1:32:17
I mean if you think about 100 people or you know very recent until very recently 50 people >> for a company that is targeting tape out end of this year to do with like 50 60 people is it's very tiny amount in in a silicon world.
1:32:29
>> What is the demand side of the equation work?
1:32:31
You're already working with uh jump trading i3d. net.
1:32:36
Uh is this something where you like if you can get capacity, if you can get performance, some solid benchmarks, uh you think that sales isn't going to be a problem or are you going to have to find and work very closely with a customer to sort of co-design a solution for a particular problem within the AI stack? >> Yeah.
1:32:59
I I don't want to trivialize or make it sound simple [laughter] like like like that.
1:33:03
Your sales guys might be listening and they're like, "We work very hard."
1:33:05
Okay, [laughter] shut up.
1:33:08
>> But well, we we only have one sales.
1:33:08
We only have one sales, >> you know, like in that way, right?
1:33:13
>> you know, like in that way, right? But but the point I will I will make is um really around um >> the way we think about uh the demand curve is >> you're kind of hitting the nail on the head in saying that like look if you can make your silicon work uh show the performance is comparable especially in
1:33:32
the current ecosystem even within the niche of you know doing this or something like that but especially if you can make the entire inference kind of workflow have a good TCO um and or and generally people always assume assume it's an or that hey you have a good TCO or you have a great
1:33:47
interactivity curve but if you can do andor uh you know you're you're going to bound to have get demand more importantly you know when you when you step into the rooms of like not only just jump trading or you know hedge funds or or kind of influencer service providers but like really the big labs the hyperscalers kind of the two
1:34:03
questions that it boils down to is like hey like look guys can you fabricate this in enough quantities like you know is your supply chain and the way that you are using the technology components is it robust enough that you can fabricate it that we can be interested in it. And then the second thing they're
1:34:14
in it. And then the second thing they're asking is can we deploy it you know is is your power source like you know do you need this kind of liquid pool setup and and if so then you know we don't we might not have a data center because
1:34:24
we've already allocated to GPUs or TPUs or can you do something else so so the questions you can see there they are asking is not that hey like you know you know we we we'll see we don't we're not sure about the demand curve. So, so from
1:34:35
sure about the demand curve. So, so from that angle you are kind of spot on that look if you can make the frontier models run you know you're you're bound to find kind of adoption in in today's market and that is a really great like I mean I'm so l like you know as positron we are so lucky to be building silicon in
1:34:51
this environment uh and that's kind of what you're seeing for with silicon companies just raising rounds uh right now >> yeah I think $9 billion has flowed into silicon companies over just the last 12 months and >> yeah we were we were talking yesterday feels feels incredibly [laughter] low relative to the spend like the annual spending category. Gavin Baker, one of
1:35:09
Gavin Baker, one of another one of your investors, has this quote they where he says, "I see it as a 1% market share is a hundred billion opportunity."
1:35:19
And that sounds like a crazy bull take.
1:35:21
And you then you realize like, wait, no, Nvidia is a $5 trillion company.
1:35:26
Like it's going to be a 10 trillion market like any day now.
1:35:31
And so yeah, actually 1% should equal a hundred billion. But yeah, anything else? Sorry.
1:35:36
else? Sorry. No, you Gav like he said that in a board meeting to me like I don't even know like a year ago or something where it's it's basically like yeah he's he's like like look look look guys like Natash Thomas just just 1% of the market you know 100 billion enterprise value just focus on your architecture where you can do well you
1:35:55
know like one of the things like you know people always like whenever a new new chip company raises around the headline and luckily you guys don't have that which is like oh to rival Nvidia it's like guys like no no one is rivaling Nvidia like get to at least 10% of their revenue before before putting the tagline on, right? But uh but
1:36:10
But uh but [laughter] but like the point there is just like look Nvidia is everywhere.
1:36:13
just like look Nvidia is everywhere. got to work in that ecosystem uh to to both work with them but also like having a product that is >> differentiated enough like you have to have technical innovation obviously to
1:36:23
stand out and and show your performance TCOs and interactivity but then you also got to prove that like look in the world of HPM co-as constraint like for us our big stories are you know like look HPM and then cos bottleneck you have Nvidia
1:36:36
TPUs AMD is ahead of you in that line you know how do you get around that well >> again you know you you say okay we are using commodity memory well pro and on no free lunch in silicon land like you know commodity memory is slow how do you solve that that's where the technical innovation comes in and then second thing is like okay it's still not trivial to get commodity memory it's not like I can just show up to Samsung and
1:36:56
be like or micron we're like hey can you give me LPDDR5X you know you have to still figure out how to how to get that and plan it but it is more feasible to to get it and that becomes a story of that the company can then scale out and
1:37:07
saying not only we're going to have a product but we're going to have a product that will scale with the requirements of hyperscalers and and and kind of the frontier How how do these how do these customers think about like
1:37:17
the minimum scale when when they're when they're working with you and they're looking at ordering order making orders that will be delivered in let's say 2028 2029 right you need to be able to and and during that same time period right
1:37:28
you know we saw Microsoft yesterday wants to add >> over 10 10 g you know an extra 10 gawatt right and so for you to be >> 6 gawatt yeah 26 38 g by 2032 >> yeah so so you're sitting there it's like to really be worth a company at that scale's time. You need to be
1:37:45
You need to be thinking about like >> it's almost like hey the orders we want delivered in 2029 are like a proof of concept for like the 2032 order which will be you know at some scale to actually impact and and be able to scale the fleet in a meaningful way but but how are you thinking about like that feels like the biggest challenge is like minimum viable sort of like um deployment >> 100%.
1:38:08
>> 100%. I mean like literally you kind of circle back on like I said when we walk in these meetings and like look the scale depends on like if you're going into hyperscalers and frontier labs and I said the first question they ask is like guys can you fabricate like this like in in enough and the question there is like in enough quantities that it's
1:38:24
like work well to us and and that answer for hyperscalers and and uh frontier labs like honestly they will literally say gigawatt plus like come up to us with a proposal of a gigawatt plus which is kind of insane right like a gigawatt like even at Nvidia scale you're talking about 40 billion $35 billion worth of of of of revenue for them, right? uh even
1:38:41
of of revenue for them, right? uh even you know assume ASIC cheaper blah blah blah all those things you're still talking about tens of billions of dollars right but like at least you have to show a plan of like how do you get to hundreds of megawws in you know to to use your like your specific figure 2028 you know for us we're taping out this
1:38:57
year production kind of ramp up in second half of 2027 in 2028 we better have a plan of how do we get to like hundreds and I I don't want to just say cop out by saying hundreds as in just 100 megawatt hundreds means truly like you know three four five and four for those but >> yeah So that by the next scale up you're in that gigawatt range. >> Yeah. Exactly. And but also like I also >> Yeah. Exactly.
1:39:16
And but also like I also don't want to discount the fact that you have other customers like you know you have inferences service providers obviously sovereign AI clouds uh you know uh quantitative finance quant finance uh kind of spectrum and so they have like different magnitudes of kind of requirements that that come through with it.
1:39:32
So, you know, we although we do internally use kind of go big or go home as as a thing, like we have to attract one of these large customers to to really be a long-term viable company, you know, I I don't want to just like discount the fact that like look, you can grow the company through the ranks as well.
1:39:48
Like you you can grow the company, you know, get 200 million revenue for 250 million, a billion, two billion through through this other kind of channels as well, right?
1:39:55
I think that that that becomes a big part of it.
1:39:57
But yeah, like you if you really want to get to like the frontier labs and hyperscalers, you're really talking about hundreds of megawatts.
1:40:05
And that's why like you know like look when we you know we we we have venture techch alliance on our kind of cap table and and when we speak with TSMC for fab capacity they're also wanting to know kind of like can you scale like you know do you have the the balance sheet to do that like the one of the reasons we raised you know 875 million is not like we need 875 million to spend tomorrow or even in the next six months.
1:40:25
even in the next six months. So I mean look we raised 230 million in series B uh in February of this year untouched right we still have all that capital uh part of it because we have been making revenue this year but we do have plans to spend that very quickly [laughter]
1:40:39
thank you that that was actually >> untouched untouched because we're making revenue [laughter] >> yes uh the the main point that is there though is like look they they want to know like if if you actually get a customer you have the capital >> and and Even that capital is is that is
1:40:57
not enough equity capital to scale out to even you know 200 megawatt right and then you have to go to the blackstones of the world and figure out how do you how to finance that deal kind of what lambda's done right so >> what's the software side of the equation you're coming for Nvidia you're
1:41:11
challenging them you're going to drive their market cap to zero >> you have to drop that in [laughter] >> no uh obviously uh this is a market that can sustain multiple players and there's different tools for the job but interoperability is important And I'm [clears throat] interested in terms of software development. Are you
1:41:28
Are you going to lean more opensource with the software side of the business or more integration with just a few buyers and code design on the software side to make sure that integration is really seamless?
1:41:45
Is is there even do we even need to be having a software conversation in an era where AI agents can write code?
1:41:52
can write code? Yeah, I mean [clears throat] you you definitely need the software conversation because like you have to plan around how people want to use it and people want to use it how they're currently using it and going to continue to use it which is based on uh
1:42:07
NVDS stack but also primarily based on PyTorch and then you know you have VL MSG lang as the and and I'm I'm specifically focused on on inference like look training you know that's such a harder challenge like you know what Jensen says like remote around scale out and everything right like that's just
1:42:22
>> that's the only reason you have probably only TPU as as a potential kind of only other silicon that can be used for training right so sure >> um uh but but on the inference side of things for sure you have to have the conversation I will say this in the in the era of aentic kind of software development um the the worries around
1:42:40
like hey you know model drops if you don't have access to it it takes you days weeks months to bring it up is it's going away like you know you know we we had news glimmer drop and within our team you know on our atlas first and could get it up and running within hours, right? You know, and and that
1:42:54
You know, and and that that that Yeah, exactly.
1:42:55
Like I had the same reaction, by the way, in terms when when we had that and and people are are are making it even faster and more automated, too.
1:43:02
Like you don't even have to interact.
1:43:03
Model drops comes in can can probably do it in in a very near future of it.
1:43:08
But to that point, it doesn't give you the right away the efficiency, the optimizations like you know, you want to extract every dollar of it.
1:43:15
of it. So to your question around you know when when the customer is large enough you want to work closely with them to like literally extract every single dollar and also like I I'll be very frank like anthropic openi this frontier labs hyperscalers they are so sophisticated they kind of want to come in and be like look guys we're even if
1:43:32
you don't want it we are working with you to make sure that this is going to like you know this is this is optimized to the to the fullest right so you so the answer as always in in this scenario is all of the above uh you know even though it might sound like it's like Oh, it's a very cliche answer, but it really is that way. >> Sounds like the mafia coming in. Oh,
1:43:47
>> Sounds like the mafia coming in.
1:43:47
Oh, your software stack isn't open source, so you're about to open it for me.
1:43:51
I'm going to make some changes if I need them.
1:43:54
[laughter] >> The software stack is like going to be built around open source like in the sense of like if you want to make like every company to to use us for inference, you know, you have to build it on SG Lang and and VLM kind of setup, right?
1:44:05
So, so but you know, it's like when you're talking to SpaceX or Anthropic or OpenAI, they're not using the generic SG or VLM.
1:44:10
have all their optimizations built in and they're going to help you do that.
1:44:14
Then obviously there's disag then within this there's all the different domains that they they do and they're going to figure out it's like oh Jalapeno is good for this AI you're good for this you know and then they're going to they're going to say okay that's how we're going to use you guys. >> Amazing. Well exciting times.
1:44:29
Uh I want to hit the gun for you. You raised 875 million.
1:44:37
>> That was a following run. Thank you. >> Congratulations. Thank you. >> Thank you so much. >> Great stuff. Great to meet you.
1:44:42
>> Keep an eye out on Instagram because we're definitely dropping a a Nvidia Challenger slide later today.
1:44:47
[laughter] >> Please do not associate my photo with that.
1:44:52
But yes, >> well uh have a great rest of your day.
1:44:56
>> Looking forward to the next appearance. >> Have a good one. >> Great to hang. >> Goodbye. >> Cheers.
1:45:00
>> Let me tell you about Railway.
1:45:00
Railway is the all-in-one intelligent cloud provider.
1:45:03
Use your favorite agent to deploy web app servers, databases, and more.
1:45:06
Well, railway automatically takes care of scaling, monitoring, and security.
1:45:10
And lastly, the New York Stock Exchange.
1:45:13
Want to change the world?
1:45:15
Raise capital at the New York Stock Exchange.
1:45:18
>> You can see Positron over there pretty soon. >> Yeah, pretty soon.
1:45:22
>> Picking some fresh ones. >> Wonderful week. >> Wonderful week. >> Short week. >> Monday. Yeah. >> But >> Monday 11 a. m.
1:45:28
Do us a favor and go ahead and have the best weekend of your entire life.
1:45:34
>> Have the best weekend of your entire life. >> Just do it. Put the pieces together. Make it happen.
1:45:38
>> We'll see you on Monday.
1:45:39
>> Leave us five stars in Apple Podcast and Spotify.
1:45:41
Sign up for our newsletter, tbp. com. Goodbye.