WEBVTT

1
00:00:00.140 --> 00:00:13.840
[gentle music] Okay, we're in the studio. Uh, special occasion because this week, uh, Diogo, my, my good buddy, uh, launched Jev, and, uh, it's been taking over the complete timeline. Uh, how do you feel?

2
00:00:13.900 --> 00:00:23.820
What, what's it like to be you right now? Emotionally? Yeah. Um, never been worse. [laughs] Um, [laughs] like, uh, I'm a ragged corpse of a person right now- Yeah...

3
00:00:23.840 --> 00:00:32.060
because there's so much going on, and I'm like a technical CEO, so I have, like, a lot of fires to fight. Yeah. But, like, mentally, it's--

4
00:00:33.800 --> 00:00:40.900
Like, I feel-- I, I say this all the time, and I've been saying this kind of for years in my over/under events.

5
00:00:41.160 --> 00:00:51.800
Like, I feel like the entire AI field is like one of those, like, carnival house of mirrors, and everyone is just insane and saying the weirdest stuff that doesn't make sense.

6
00:00:52.120 --> 00:01:06.040
And it feels like, for just this week, like I'm, I'm on a better in sync with reality and like, oh, people see it now. AI can be so much more than what was once thought. Mm.

7
00:01:06.180 --> 00:01:18.980
And like, yes, we are going to make-- Like, like, an AI-based economic revolution is back on the table, and, um, this is fucking awesome. You know, this is fucking awesome. [laughs] I'm so jazzed the developers get it.

8
00:01:19.320 --> 00:01:28.600
It's, it's, uh... Yeah, and I want to show my internal gratitude to the developers and- Yeah... I'm so jammed about the community and everything. It's so great.

9
00:01:28.780 --> 00:01:41.660
Yeah, you were saying yesterday that you decided to prioritize the dom-- town hall and not a bunch of like, you know, VIP inv- investor type people because you wanted to make sure that they are the people that you get, get your most, uh, a-attention, right?

10
00:01:41.720 --> 00:01:52.080
The engineers, the developers. Yeah. It felt a little like, "Oh man, I'm talking to, like, really important people right now." Yeah. I probably shouldn't reveal who. Yeah. But it feels a little bit dirty for me to, uh...

11
00:01:53.240 --> 00:02:05.120
I, I, I'm, like, perhaps overly genuine in things. Like, it, it feels, like, dirty if, like, in my gigantic calendar event of things to people to talk to, you know, the community isn't one of those. Yeah. You know?

12
00:02:05.160 --> 00:02:13.280
And actually, in my ideal world, it would be, like, community all the time. I was thinking, "Should I host a town hall while walking to your studio?" And I'm like, "No, that's too crazy." Sure. Yeah. Yeah.

13
00:02:13.360 --> 00:02:21.820
Uh, [laughs] well, you, you, you guys have been hosting town halls on the Discord. Discord is now at a hundred thousand people. Um, your, your Twitter is kind of- I don't follow these stats. Yeah. So holy shit.

14
00:02:21.960 --> 00:02:23.020
Your Twitter is blown up.

15
00:02:23.100 --> 00:02:29.520
Uh, you know, uh, it was, it was really funny 'cause, like, at AIE, you were like, "Uh, follow me, please," and then you didn't, like, provide even your, your handle, so like- [laughs] I'm a noob.

16
00:02:29.880 --> 00:02:35.360
I'm a noob You're such a noob. I'm a noob. But no, but that, like, that's, like, positive aura- Cool... that, like, you don't know how to pro- [laughs] promote yourself. Yeah.

17
00:02:35.360 --> 00:02:43.480
Well, someone, like, called me out when I posted, like, "Holy shit, we're all three trending, trending topics." And then they're like, "That's a personal feed." That's a personal- And I'm like, "Oh,

18
00:02:44.400 --> 00:02:50.860
no" Of course, of course it would trend to you. Cringe. Yes. Yeah, yeah, yeah. 'Cause it's what you clicked on. Um, so okay. Uh, let's, uh, yeah, oh, so congrats on everything. Thank you.

19
00:02:50.860 --> 00:03:02.740
We'll talk about more, uh, details a-as you have them, but let's, for people who are, like, living it under a rock or just- [laughs]... just want, like- Mm-hmm... the definitive thing, what is Jev? Oof.

20
00:03:04.320 --> 00:03:12.940
Let me think about-- That's a hard one. [laughs] Um- Okay, and, and I'm happy to, like, uh, re-ask if you wanna kinda address this one. No, no, no. I'm happy to- Okay... I'm happy to, like, just jam on it. Yeah.

21
00:03:13.020 --> 00:03:21.660
I will say, like, the first thing that I'm relieved about with this question is now I don't have to answer that question to my parents anymore, 'cause ChatGPT can just explain it. [laughs] Nice.

22
00:03:21.900 --> 00:03:33.740
Um, so, uh, the way I see it is we neew- need a new class of models. Um, we're not attached to naming that class of models. Um, our, the most accurate name we've come up with is System One models. Yeah.

23
00:03:33.780 --> 00:03:43.620
There will be reasons, but it's, there's a reason why we don't call them decision models because, like, they will be... Like, System One is beyond that. Uh, that-that's all I can say.

24
00:03:43.980 --> 00:03:54.280
Um, we didn't expect this to be our big launch, so we have stuff in the tank. Um- You should've said low-key research preview. [laughs] Um, it kind of was, right? It kind of was.

25
00:03:54.700 --> 00:04:07.260
But, um, we, so there's a class of models that we describe them as, like, machine native, System One, large programmable. Mm-hmm. I think these are, is the class of models where the goal is for code to be the consumer.

26
00:04:07.800 --> 00:04:23.220
So, um, as opposed to, um, lo- um, pre-trained large language models, which are meant for, like, auto-complete of the internet or RLHF models like chatbot instruction-following models, which are meant to re- like, reply to text, um, or RLVR.

27
00:04:23.260 --> 00:04:30.500
It's in a weird gray area with R-RLHF. Like, these are meant to have things that directly are consumed by code, hence the name type safe.

28
00:04:30.860 --> 00:04:48.340
So the thing we really, really, really want is to have, like, AI, like, be as powerful as possible, and we think the way to do that is to integrate it with software, and we are designing everything, you know, beyond just the outside the deep internals of the model to be optimized for software.

29
00:04:48.640 --> 00:05:02.500
So, uh, number one, Jev is our first large programmable model, um, or, uh, System One model, whatever you want to call it. Um, Jev is meant to be optimized for intelligence per dollar, uh, hence the name of Jev.

30
00:05:02.680 --> 00:05:07.940
You know, uh, I mean- Jevons paradox... Jevons, Jevons paradox, yeah. Um, and it's optimized for intelligence per dollar.

31
00:05:08.300 --> 00:05:16.240
I love this debate with people about what is the most important between reliability, cost, calibration, and speed. Um,

32
00:05:17.220 --> 00:05:23.280
and Jev is meant to be-- Jev will be the name of models that will be on the frontier of intelligence per dollar.

33
00:05:23.480 --> 00:05:31.780
There's other ways to optimize it, like, uh, ML, or at least if you're good at ML, it's all about trade-offs, and, you know, we are just going all out on that. Yeah.

34
00:05:31.840 --> 00:05:37.480
And, and to me, like, calibration is one of the new things that people weren't talking about as much. We've done an episode- Mm-hmm...

35
00:05:37.580 --> 00:05:44.120
uh, in the past, uh, with Clementine Forget of Hugging Face, uh, where they were like, yeah, a-actually, um, y- they're just...

36
00:05:44.640 --> 00:05:50.200
Or, you know, and this is your, your argument about RLHF, is they're model collapsing towards what you want to hear the most- Ooh...

37
00:05:50.220 --> 00:05:57.400
or what is most likely, uh, instead of, like, their own internal confidence about a, a, a thing. Can I soapbox on that for a second? Go ahead. Yeah. Or- Yeah, yeah... cool.

38
00:05:57.720 --> 00:06:03.060
Uh, like, um, uh, I've been heard that your audience is the most technical, so I actually want to get into that. Yeah.

39
00:06:03.160 --> 00:06:11.028
Um, and if I went through extreme precision to make sure everything in our launch video is accurate and real. Apparently, that's very unusual.

40
00:06:11.428 --> 00:06:21.068
One of the things that no one paid attention to was the downsides of RLHF, in particular mode dropping. Um- Mode dropping or mode collapse? Uh, it's the same thing. Is that what you- It's the same thing. All right.

41
00:06:21.368 --> 00:06:32.268
And, um, I, I, I wanna have a blog on this eventually, but I, like, want to tell as many people this as possible 'cause I think it's a very interesting thing. So the spicy take, I believe in Jan LeCun a lot.

42
00:06:32.568 --> 00:06:41.288
Uh, I think Jan LeCun's takes are actually among the closest to- What, what about this? [laughs] Well, do you want to address this now, or should I let you go on mode collapse? No, no, no, later. Okay, okay.

43
00:06:41.328 --> 00:06:53.188
Go, go, go more, like, on mode collapse. I wanna- So I actually think that among takes, Jan LeCun's is among the most accurate. Um, but he has this very famous/infamous slide about, uh- The cake?...

44
00:06:53.208 --> 00:06:59.768
the la- LLMs are doomed. Okay. You know, like that one where he's, it, where he, like, has, like, a pie chart with, like, a s- tiny part, tiny little thing- Pull it up...

45
00:06:59.788 --> 00:07:07.188
and says that as you increase sequence length, um, the probability of it making an error goes in. Yes, this one. This one.

46
00:07:07.548 --> 00:07:15.568
I love this one, um, because it's one of these things that seems mathematically obvious but is obviously wrong, right?

47
00:07:15.708 --> 00:07:26.488
Like, it's mathematically obvious, but it doesn't empirically hold, and this is my favorite thing to teach people about. Like, where y-you- What's the disconnect? Exactly. And, um, may I or you want to tell me?

48
00:07:27.288 --> 00:07:31.468
Uh- About mode collapse? Oh, no, no. W- uh, so mode coll-collapse is related to this. Yeah.

49
00:07:31.828 --> 00:07:42.928
The disconnect happens because if you are in a mode covering or a calibrated distribution, um, you are, like, not, you are not overly punished about having outliers.

50
00:07:42.988 --> 00:07:50.368
So you'd expect, like, you know, something, some amount of the time you'd be out of distribution, some amount of time you'd be in distribution. That's what happens when you cover the distribution.

51
00:07:50.668 --> 00:07:56.868
This was, like, models before GANs. They made blurry images, right? Mm-hmm. Um, instead, GANs mode drop.

52
00:07:57.048 --> 00:08:03.228
They, like, drop the minority classes and just do the really common ones, and this is why this effect doesn't happen, right?

53
00:08:03.408 --> 00:08:13.708
Like, instead of b- in order to generate really long strings, um, without making errors, they need to, like, be extremely conservative because it- really easy to see when an error happens.

54
00:08:13.988 --> 00:08:23.208
It's very hard to see when, like, a subtle thing that looks correct happens, and that calibration is, like, total poison into, like, the probability distributions of strings. Yeah.

55
00:08:23.288 --> 00:08:26.908
And, uh, the, it's a, it's a nuanced take and, like, I, I think that

56
00:08:28.028 --> 00:08:37.788
th-this is why this doesn't happen, and this is why strings are so bad at, uh, decision-making or, uh, you know, overloading the string models are for decision-making is, is, like, a, a bad time.

57
00:08:38.108 --> 00:08:47.468
And while we're on the topic of Jan, do you agree that his fix i-with which is, like, a world model, like a JePA-type, um, embedding thing is the right solve?

58
00:08:48.508 --> 00:08:58.348
So bas-basically, like, the, one of the reasons that it could fail is because you're trying to reason over token outputs and then, and then just looping back again and, uh, going, keep, uh, continuing going until you reach, like, a end of sentence.

59
00:08:58.888 --> 00:09:05.468
Um, like, is that, uh... And his solve is JePA, right? Yes. Which is, like, joint ambition, uh- Yep... uh, joint embedding prediction.

60
00:09:05.888 --> 00:09:14.768
Uh, so, like, is that the solve or, you know, like, uh, do you have a, do you have a take on, on that? Oh, man, I probably shouldn't talk- [laughs]... too much about the insides of ML.

61
00:09:15.148 --> 00:09:27.368
But I will say that my, my brand, other than unhinged, is practical. Mm-hmm. You know, like, even my take here is practical, and, like, I'm, am I a scaling law fan? Um, depends.

62
00:09:27.568 --> 00:09:33.208
It depe- you know, it's, it's, it's, like, it's a, scaling laws tell you how much better you get at a thing for amount in. Mm-hmm.

63
00:09:33.368 --> 00:09:43.408
A scaling law does mean exponentially more resources for normally sublinear gains, which looks to be a bad investment unless those, like, linear gains are, like, really, really valuable.

64
00:09:43.968 --> 00:09:54.768
Um, but it's all, to me, it's all about, like, what can we do with what we have to make the biggest possible fucking difference? I can curse. Yeah. Yeah. Yeah, yeah, yeah. Yeah. We're, we're at, we're a poof for adults.

65
00:09:54.848 --> 00:10:02.928
Hell, hell yeah. Uh, and also we have a scaling law thing if you wanna go into that later. [laughs] Oh, I, I could if we, we... It's, see, that part is not super relevant- Yeah... right now.

66
00:10:03.028 --> 00:10:15.008
I actually, if you wanna go into my Bitters lesson, I think that that's more relevant. Okay. But, like, to me, I'm all about, like, pragmatics, and I think that the JePA stuff is really cool early research.

67
00:10:15.528 --> 00:10:26.788
I really love awesome research. Is it practical yet? Um, probably shouldn't say. Um, but, uh, like, there's just a lot of...

68
00:10:29.288 --> 00:10:44.808
I just think there's just, like, so many diamonds in the rough lit all over the research world right now that haven't been polished because people don't know how to, like, do the right task, and I think that what our launch did, it, does it kickstart us as a company?

69
00:10:45.168 --> 00:10:52.328
Like, yes. Will it be great for us as a company? Yes. I think it's gonna be, like, even greater for this direction of, like, programmatic AI.

70
00:10:52.648 --> 00:10:59.268
You know, there, there was going to be, like, a gold rush on top of us for, 'cause, like, software is super fucking charged.

71
00:10:59.568 --> 00:11:12.808
But I think there's gonna be a gold rush parallel to us as well on, like, all the different ways we can expose things to make software more powerful so people can make even cooler stuff, and then we are back to, like, early internet energy, you know?

72
00:11:12.828 --> 00:11:18.528
Yeah. And, and I think that's why, like, you know, the Twitter is just like, "Jeff, Jeff, Jeff," you know? Um, it's, it's like, it is like a party.

73
00:11:18.528 --> 00:11:27.908
It's inspiring because it's, it's, like, so different than what we're used to, which is, "I'm sorry you can't do this, but we do scaling laws and only the big labs can do it," right?

74
00:11:28.368 --> 00:11:35.828
That, that- [laughs] Uh, actually, c- if I... I'll, I'll make a tangent if that's okay. Yeah, yeah. I think you might enjoy this. Really? Might d- Five, five tangents in. No, this is good. This is fine, yeah.

75
00:11:35.828 --> 00:11:41.748
Oh, yeah, I get lost in all my tangents. This, this is gonna be horrible for the listeners to figure it out, but they, they're gonna figure it out. It's fine. Yeah, we could edit it if it helps. This morning, yeah.

76
00:11:41.848 --> 00:11:54.168
Um, [laughs] so popular thing on Discord, um, that people keep asking me, I haven't had the time to explain it yet, is why am I opposed to safety alignment and why do we not refuse?

77
00:11:54.748 --> 00:12:05.448
Um, I'm not opposed to safety as a principle, but I think that safety alignment is generally misaligned with users. Um, and, uh, refusal is just, like, obviously a type error.

78
00:12:05.708 --> 00:12:16.508
Like, if you're a human being and you're chatting with, like, a, you know, a bottle or whatever, you're cloud coding, and a refusal happens like, "I'm sorry, I can't read DNA.py"- Um, that's an annoying time.

79
00:12:16.748 --> 00:12:24.308
It's anno- it's, it's annoying- Mm... right? But you can work with it, right? And you're forced to work with it 'cause of Stockholm Syndrome. Um- [laughs]... uh, d- I have stories about that t- too.

80
00:12:24.408 --> 00:12:35.128
I, I need another tangent deep in here. But, like, if you ever want this in a dependency running in the background, what happens if that refuses? What if someone else is using that dependency?

81
00:12:35.288 --> 00:12:45.348
They don't know what that system is. Like, like you want the software to just stochastically break because a user sent like a weird message in there? Mm. Like, that is, um, like straight up insanity.

82
00:12:45.608 --> 00:12:50.968
It i- it's, it's coming from a place of, like, people who do not understand software, do not understand programming- Mm...

83
00:12:50.988 --> 00:13:01.828
and like they are obsessed with like, I believe, this, um, horseless carriage of like AI coworker instead of unearthing like the full power of AI. Fair enough. Uh- Yeah...

84
00:13:01.868 --> 00:13:13.108
you, you want something that is the core kernel that is, is usable everywhere. Yes. E- exactly. Like the cognitive core, right? Yeah. And you need this thing to be s- like so general, so optimized for its use cases.

85
00:13:13.168 --> 00:13:22.848
You want it to be like, uh, you know, you want it to work on all the future use cases, all the weird shit that people are doing. Yeah. You know, we obviously didn't train on any of that stuff.

86
00:13:23.128 --> 00:13:32.408
Is it surprising that it works? No, 'cause we trained on weirder stuff, my friend. [laughs] Um, so but one tangent up about like safety alignment. Okay.

87
00:13:32.668 --> 00:13:39.248
Safety alignment makes sense for a product, in my opinion, for like ChatGPT and Claude. Like it, um, what, what safety...

88
00:13:39.288 --> 00:13:47.448
what makes safety and capability alignment different is capability alignment is like about doing what the user wants. That is sick for software engineers.

89
00:13:47.668 --> 00:13:54.968
They want their thing to do the thing, and the more predictable it is, the less they have to test it and play around with it. Jeb is not anywhere close to that yet.

90
00:13:55.028 --> 00:14:04.308
It could be, but like there's so many more nines of reliability that we want in order to make it so good, like a database query, that you don't even have to think about it. It is just there when you need intelligence.

91
00:14:04.888 --> 00:14:15.008
Um, but safety alignment is like the opposite of instruction following. It's when you want to follow someone else's instructions, like OpenAI Anthropic's- The lab's values- Exactly... stack. Yeah. Exactly.

92
00:14:15.188 --> 00:14:19.408
And this makes a lot of sense for our product. Again, like ChatGPT should do...

93
00:14:19.768 --> 00:14:31.348
Y- you, you sh- like if they don't want to like, uh, do like some, uh, not safe for work role play with ChatGPT, that's on them because like maybe that's, you know, f- what their users who have like parents and kids want.

94
00:14:31.688 --> 00:14:45.848
L- like that's fine. But in an API, that's nuts, right? Um, l- like that's completely unacceptable because like people need to like program around this and that is, that's so anti-user that it's... It, it, it-- I'm...

95
00:14:46.628 --> 00:14:54.928
Huh. I, I, I, I, I can be an angry person, so, uh, I should try to calm down. It, it's, uh, people get your passion, and I think it's really good.

96
00:14:55.028 --> 00:15:03.328
Uh, the, the one pushback I'll give you is like what if we use it to kill people, right? Like that, that is the actual like... N- the not safe for work thing, it's private, personal, whatever.

97
00:15:03.648 --> 00:15:15.728
But like, yes, like we will use it in war, and like that, that is, uh, something that companies can reasonably prefer their APIs not be used for. [laughs] I get that.

98
00:15:15.968 --> 00:15:27.748
I, I, I think that there's like pragmatic places where that opinion can be held. I don't think the foundation of like a general purpose technology is that place, uh, personally. Mm.

99
00:15:27.768 --> 00:15:39.208
Like would I prefer that our stuff is not used to kill people? Obviously. Would I prefer it's used for like all sorts of like great stuff in the world? Obviously. Will I put my thumb in the scale for that? Yes.

100
00:15:39.668 --> 00:15:45.128
Would, will I do it at the technological layer? Absolutely not, because that will fracture the intelligence.

101
00:15:45.608 --> 00:15:54.828
Every single time you mean it to over fit to some weird stuff, you're fracturing its intelligence more and more, and like these things are fractured to the like... They're, they're so dar- darn fractured right now. Yeah.

102
00:15:54.888 --> 00:16:04.028
So, uh, uh, and, and as a furthermore thing, um, to me it's like I think intelligence will be more like a database than a coworker.

103
00:16:04.348 --> 00:16:16.028
Like I don't think it's up to databases to add checks on whether or not they're used for like, uh, what's something that's not great? You know, like CIA... Actually, I don't know what the CIA does really.

104
00:16:16.068 --> 00:16:20.848
You, you can imagine. [laughs] You can imagine. Uh, k- killing people who are not even bad or whatever. Mm-hmm.

105
00:16:21.388 --> 00:16:34.048
And like I don't think it's the database's responsibility for that, and furthermore, like a thing that has been weird to me is when people like sign up for our thing on Slack and they're like, "Hey, we're gonna deploy this.

106
00:16:34.408 --> 00:16:46.388
Can we deploy this thing?" I am just like, "My brother, we are an API. You are a developer. It's none of my business," right? Like, uh, like you shouldn't know what the whole task even is- Yeah...

107
00:16:46.428 --> 00:16:55.668
because it should be decomposed into small things. We shouldn't be able to know what the downstream users are doing, and that is like a good boundary to give software engineers maximum power.

108
00:16:56.068 --> 00:17:05.368
Ideally, they use it for the good stuff, and ideally, we can like help them, and like we've talked about like doing open source and charity and all of that. We have absolutely no time for anything else right now.

109
00:17:05.668 --> 00:17:11.928
But like they will get any of that bias out of the technological layer as long as I'm in charge. Yeah. That's great.

110
00:17:12.008 --> 00:17:21.088
Uh, while we're on the topic, let's also briefly talk about your privacy stuff, terms of ser- terms of use, which, uh, got a little bit of- Ooh... uh, misunderstanding. Uh, I just wanna clarify that upfront. Hell, yeah.

111
00:17:21.108 --> 00:17:27.848
I think it probably takes two sentences from you about like you will not, uh, y- you're not being that restrictive about your API. Like clearly- Oh, yeah. Obviously, yeah...

112
00:17:27.848 --> 00:17:38.288
ideologically, you would take a, your role as a platform very seriously. Yes. Uh, yes. Uh, uh, I don't know what you're referring to, but like this was... Uh, I, I've seen a couple of things about like benchmarking. Yes.

113
00:17:38.348 --> 00:17:45.848
Like obviously we're not stopping people from do-- Oh, man, I should be careful about what I say. I'm realizing- [laughs] No, you said it, you said it publicly that- Yeah... that was in the preview period.

114
00:17:45.888 --> 00:17:52.628
You didn't take it out for the launch. Yeah. Okay. And now you're gonna take it out. So the team is doing stuff that- Yes... I'm not even aware of, so it's great to know the team communicated that. Mm-hmm.

115
00:17:52.708 --> 00:18:02.828
I asked them to check in with the lawyers about that. Yeah. Like we are obviously not stopping people from doing that type of thing. I am extremely in favor... So I'm extremely anti-public benchmarks.

116
00:18:03.228 --> 00:18:15.028
Um, I'm extremely in fav-- I'm medium about private benchmarks that are proxies. Um, you know- So are, are you worried about, uh, sat- saturation or like, uh, training on public benchmarks? So it's like e- easy to cheat.

117
00:18:15.468 --> 00:18:27.848
Um, not only is it easy to cheat, there's a lot of ins... So Uh, I think that we are, or, or anyone who's, like, competition with us that, you know, vaguely there is...

118
00:18:28.248 --> 00:18:34.208
Like, you could say, like- There's like 50 Jeff clones, yeah. Well, uh, sure, sure. Yeah. Uh, oh, the, these-- Let's say that there's some competition. And we'll talk about those. Yeah.

119
00:18:34.228 --> 00:18:40.288
Or let's just say that there's l- there- let's just assume that there's an industry two years from now of people who are doing similar things to us.

120
00:18:40.628 --> 00:18:51.648
The thing that we are selling is intelligence per something, per, like, dollar or per second. Um, the-- No o- like, people obsess about the cost and the speed. Um, I believe that that is...

121
00:18:51.828 --> 00:18:57.368
It's cool, but, like, the thing that matters is the intelligence. Like, the cost and the speed are, like, th- are bad things.

122
00:18:57.748 --> 00:19:06.388
You know, you're paying them for something, and you need the thing back, and the intelligence is what truly matters. The problem with intelligence is that there's a je ne sais quoi to it, right?

123
00:19:06.748 --> 00:19:09.048
Like, the, like, uh- Mm-hmm... the, the good model smell.

124
00:19:09.088 --> 00:19:17.968
Like, the thing that happened after we launched of, like, two hours later that actually went way bigger than the video, which was like, "Holy shit-" This is actually usable. It, well- Yeah...

125
00:19:18.008 --> 00:19:22.608
it, it's, you know, you know, like, beyond that. Yeah. You know, like, the... Hoo.

126
00:19:22.668 --> 00:19:33.988
The, the, the launch was crazy, and people could really sense how hard we care about that, and that's truly what I think the long term of this is. And I think public benchmarks are antithetical to this.

127
00:19:34.328 --> 00:19:47.588
Like, they are a way to get people trust in intelligence because intelligence has a, has a je ne sais quoi. But the public benchmarks are extremely, extremely gameable. Even if they try not to, they still will.

128
00:19:47.868 --> 00:19:58.928
You know, like, l- back in the old days, every lab had a team to collect data that looks like MMLU to make it look better, which is just benchmarking, benchmarking with extra steps. Mm-hmm.

129
00:19:58.968 --> 00:20:14.428
So, um, I, I believe that in the long run, it needs to be vibes and trust until you put it into a workflow and evaluate it for that workflow and measure it and have your own sense of, like, how it does on the exact workflow that matters.

130
00:20:14.848 --> 00:20:18.188
And our job is to keep moving the nines of reliability.

131
00:20:18.528 --> 00:20:28.967
This is, like, an ever-present part of our, l- l- like, of what we need to be doing as a company, and we need to do everything to have people know that this is something we care so much about.

132
00:20:29.288 --> 00:20:40.028
You know, like, if we wanted to, we could have released Jeff, like, a year and a half ago if we wanted it to be dumb. Oh. Uh, it, like, like, the, you know, my bitterest lesson, right? Like, architecture and, uh, yeah.

133
00:20:40.228 --> 00:20:54.188
I'll bring it up- Hell yeah... since you, since you talked about it, uh, here. Hell yeah. Um, t- like, you know, Suchin says that algorithms beats compute very roughly. Um, data matters way more than compute, obviously.

134
00:20:54.528 --> 00:21:05.768
And doing the right task, having the North Star is, is, is the hardest, most important thing. This has happened, um, in LLM land twice so far, right? Maybe 2.2 times.

135
00:21:06.048 --> 00:21:17.448
You know, there's RLHF, which, like, shifted the task to instruction following. No one realized that that was possible. Um, RLVR did, like, a tiny little, like, edit to the, to the direction, and now us, right? RLCD.

136
00:21:17.668 --> 00:21:31.428
We have a new task, and the goal is, you know, programs in the loop. And, um, yeah, data matters so, so, so- Right... unbelievably much. So, so- It's like I can't, I can't emphasize it less.

137
00:21:31.508 --> 00:21:37.608
Yeah, you consider yourself a data lab rather than, like, a model lab. Is that- Absolutely... something-- That's the wording that you guys use? Yeah, yeah, yeah.

138
00:21:37.608 --> 00:21:50.568
Uh, we are-- We will always, like, care so much about data. Um, to me, model capabilities means data. Data is so unbelievably complicated, and that is what gets nines.

139
00:21:50.888 --> 00:22:04.388
Like, you have no idea how, how much data can shift everything. Data is so important. Yeah. Holy crap. Um- So if people are looking for a job, um, we are hiring infinite data people. Actually infinite.

140
00:22:04.388 --> 00:22:15.968
[laughs] What is a good data person? Like, um, you know, clearly somebody who cares about reading through the, the transcripts of, uh, whatever. You said, for example, that y- all your data is synthetic. Yep.

141
00:22:16.008 --> 00:22:21.408
But, uh, that's only, like, the scratching the surface, right? Yeah. Like, it's not, like, synthetic, so what, right?

142
00:22:21.708 --> 00:22:30.308
Synthetic, but we have a people with a lot of taste and a lot of care looking at, looking at these, articulating what's wrong, going back, regenerating. Is, is that what a good data person is these days?

143
00:22:30.328 --> 00:22:44.308
[sighs] Let me try to figure, figure out how to... Like, it's, it's super complicated, and, like, I literally onboard the data people with a talk that I assume is longer than this podcast will end up being.

144
00:22:44.668 --> 00:22:56.468
So I will try to say, like, the high level of it. So number one, we don't do the kind of synthetic data that people ki- well, I'll do a... Actually, number is zero. Um, data and synthetic data depends on your task.

145
00:22:56.728 --> 00:23:06.448
Like, the shape of your data, uh, s- the shape of your task changes the data. Like, RLVR's data is kind of environments, right? Yes. Um, RLHF's is the human feedback, you know.

146
00:23:06.848 --> 00:23:15.708
Um, each task has its own unique kind of data, and we, of course, have our own unique kind of data, right? Um, so number one, we have that.

147
00:23:16.168 --> 00:23:23.448
Uh, number two, um, the thing I l- the reason why we don't want to train on our users' data even if we could, right?

148
00:23:23.488 --> 00:23:35.968
Like, we could probably ask for any terms right now, and it'll, w- w- I don't know if it would make a difference. We truly don't want that, um, because no matter what, the real world data has so much bias.

149
00:23:36.328 --> 00:23:42.248
There's, like, a power law of, like, people, like, asking the same things where you'll end up, like, over fitting to it and, like, fracturing to it and all of that.

150
00:23:42.588 --> 00:23:56.368
And number two, we are, like, aiming for, like, a complete sci-fi future years from now where, like, these models are going to be, like, the general infrastructure, layers and layers and layers and deep down the stack to, to, like, things people can't even imagine.

151
00:23:56.628 --> 00:24:02.708
Like, I would like to think of our model, like, kind of like, like, you know, UDP as LLMs and TCP as our models.

152
00:24:03.148 --> 00:24:12.148
All sorts of stuff can be built on top of that, and we need to be able to nail those futuristic use cases such that software developers can actually build that futuristic stuff.

153
00:24:12.668 --> 00:24:20.668
And the way to do that is even if we had all of the data of the present, um, we would just over fit to the present, and then it wouldn't work. What we need is to, like-

154
00:24:21.896 --> 00:24:28.136
It, it, it almost feels like, uh, like, I-- they're, like, artists, you know? They study this cognitive core.

155
00:24:28.256 --> 00:24:39.016
Our cognitive core is, like, way less jagged than anyone else's, and then they find the jaggednesses, and then they address them surgically in a way that, a-and you can never perfectly do this, right?

156
00:24:39.256 --> 00:24:47.336
But they do it in such a way that it addresses it in every single possible, like, dimension, past, present, future. The general case rather than the specific case. Exactly.

157
00:24:47.356 --> 00:24:56.436
And, like, that requires a lot of intelligence every time. Um, okay, so we mentioned a little bit, you, you, you sort of criticized my thinking as R- like, very RLVR influence, which is, like, very fair.

158
00:24:56.916 --> 00:25:06.516
Um, let, let's, let's actually mention RLCD. Ooh. Um, which obviously you have some secret sauces. To my knowledge, you've never actually published a paper or anything like that on it. No, right? No, not yet.

159
00:25:06.576 --> 00:25:15.036
Um, but, like, what, what should people get from this? Like, what, uh, can you give people some confidence that you're just not just making up jargon for the sake of sounding cool, right?

160
00:25:15.316 --> 00:25:22.936
Like, uh, one thing for me is, like, calibration I do think is, uh, to me, like, well understood because we've covered it in, on the podcast. Yeah.

161
00:25:22.956 --> 00:25:31.916
But I, I don't know what you mean when you say RLCD versus what people are familiar with. It's a great question. Yes. And actually, I will give a related question. Okay. Um, what is RLHF? Okay. Right?

162
00:25:31.936 --> 00:25:41.456
And actually, RLHF means multiple different things, right? Okay. Like, there's the RLHF of the original, I think it was, like, Paul Christiano teaching a robot to backflip or something like that.

163
00:25:41.756 --> 00:25:51.456
That, wasn't there something- Was that it? That, that was the original R- I, I referenced the PPO paper, but I don't, I don't know. Um, and, uh, so, uh, PPO is not necessarily from human feedback, if I recall. Okay.

164
00:25:51.956 --> 00:26:03.136
That's true. But I, but I believe it was, like, an OpenAI alignment work that could teach hard-to-specify outputs, like a backflip. I'm not 100% sure. Um, and then there was actually learning to summarize.

165
00:26:03.656 --> 00:26:16.556
You know, this was work, um, by a bunch of the, the team that helped with, um, instruct and, and co-authored, um, the instruction following paper, which was teaching doing PPO on language models. This is the, uh,

166
00:26:17.476 --> 00:26:26.776
uh, sorry. I'm trying to, trying to- [laughs] Yeah. Trying to manipulate this, this thing. Uh, this is 2017. Yeah. All right. Um, I, I'm not 100% sure, but, like, that looks quite right. Yeah.

167
00:26:26.816 --> 00:26:36.916
If it has, like, a robot doing backflips or something like that, that might be it. [laughs] Yes. Okay, cool. I guess I, I got it right. Hell yeah. Um. There you go. Yeah. That's the one.

168
00:26:36.976 --> 00:26:41.396
So the idea was kind of like can you do, like, ill-specified things with it? So that's like version one.

169
00:26:41.676 --> 00:26:52.256
Version two was, uh, the learning to summarize work, um, uh, that, uh, like, OpenAI did, which is actually, like, PPO on language models to do something somewhat ill-specified.

170
00:26:52.616 --> 00:27:03.016
This is, like, a-another thing that people refer to as RLHF- Mm-hmm... which I did not co-author. Mm-hmm. Oh, Dai was there. Cool. Um. [laughs] Hell yeah. And Radford. Uh. Yeah, yeah, yeah.

171
00:27:03.096 --> 00:27:15.036
Uh, shout outs to Alec and Ryan. Love them. Yeah. Um, but, uh, the thing that I refer to RLHF is the, um, oh, man, of, I, I'll get into that. You have comments on that, yeah.

172
00:27:15.116 --> 00:27:25.316
I have comments on that paper, but, like, we're so many, uh, tangents deep. Yeah. Um, so the thing that really got, uh, to, to me, the thing that I'm calling to RLHF is the task of instruction following.

173
00:27:25.616 --> 00:27:36.856
It's not about the PPO. That part doesn't matter. It's about, like, setting a North Star of this is a valuable direction. It's kind of like the Bitterest Lesson, Lesson North Star. Ah. And for us, RLCD is this new task.

174
00:27:37.456 --> 00:27:46.616
Um, and it is not-- I, I don't see it as jargon. Like, I, I try to communicate with precision. It's just that, hey, here's another North Star.

175
00:27:46.816 --> 00:27:54.476
Just like DPO and all of its like, um, you know, descendants also do RLHF despite not using the algorithm in that paper. Mm-hmm.

176
00:27:55.896 --> 00:28:06.836
And so cl-clearly stating the North Star is, um, being program, uh, programmable AI is, is, is one, uh, word that I really ca-catch on to, uh, removing the human in the loop- Yes, yes, yes...

177
00:28:06.836 --> 00:28:17.876
uh, from, uh, because RLHF is tuning- Yes... for this so that you can automate everything. Yes. Everything that we can. Uh, did I miss anything else in the, in the thesis of, like, what the North Star is? There is--

178
00:28:18.876 --> 00:28:30.396
That is, that is right. Uh, I'm overly nuanced in my communication. The one nuance is that we need to be practical. We need to be aware of what language models can do really well. You know, like what AI can do. Mm-hmm.

179
00:28:30.436 --> 00:28:42.056
Right? Like, um, there could be programmatic types that are, like, sick AF, but if you, if the technology is not ready for it, it to, it's not a tragedy if that's not out in the world. Yeah.

180
00:28:42.096 --> 00:28:50.836
But to me, like, the pre-Jev world was a tragedy because it, it, it sounds arrogant. Uh, uh, hear me out. No, no, no. I strongly believe you. Cool.

181
00:28:50.896 --> 00:28:56.836
Uh, it sounds arrogant, but, like, I felt this way since long before I even had a company. Yeah. And I can, I can vouch that, uh- Yes, I've been talking about this for so long...

182
00:28:56.856 --> 00:29:02.636
you've said this at Rounder for, like, three years. Yeah, I've been talking about this for so long. And, and I've been saying it because I thought it would have been easier.

183
00:29:03.036 --> 00:29:09.676
Um, they say they do not do things because they, it, they're easy. It is. They, 'cause they thought it was easy, so something like that. Yeah, exactly, exactly, exactly.

184
00:29:09.756 --> 00:29:20.836
I, I thought it, this whole project would take a week. Um- [laughs]... and I was unbelievably wrong, so I am so sorry to everyone at OpenAI that I thought, I was like, "Man, I'm solving this right now."

185
00:29:21.376 --> 00:29:37.556
Um, but, uh, like, I think that the tragic thing is when-- Well, I think overpromise and underdeliver is tragic too, and, like, AI is super extreme on that axis, and I think RLVR is, like, the main, well, both RLVR and RLHF are extreme perpetrators of this.

186
00:29:38.096 --> 00:29:46.376
Um, but, like, it, to me, it's like it's just, there's just so much potential there. Like, AI is clearly so smart. I, smart.

187
00:29:46.476 --> 00:29:58.736
I love this in my talks, you know, when I ask people, like, "How can AI be so unbelievably smart? How can we, like, solve Millennium prize problems in math, but still not automate even the most basics of works?"

188
00:29:58.976 --> 00:30:07.716
Like, like, really basic rote stuff that, like, the, you know, it, it, it doesn't take, like, extremely smart people to do this. It's not a satisfying job.

189
00:30:07.936 --> 00:30:26.424
Like, there's other things these people could be doing, but yet we need them to do, like, this ba- like super basic Non- unsatisfying stuff because, you know, like we can't automate it yet, but we have this like supercharged engine of automation that just does not have like the right plugs and stuff to plug into all of this economically valuable work.

190
00:30:26.784 --> 00:30:36.064
And, you know, like if the whole company of TypeSafe disappears, like maybe it'll take like a year or two for people to like truly catch up. I actually don't know how long it'll take.

191
00:30:36.124 --> 00:30:49.164
If, uh, if model quality matters, then we are gonna be in a very good position for a long time. But, um, it, it, like, it, it, it's done, right? Like there-- Like this has changed the path of like technological history.

192
00:30:49.444 --> 00:30:57.544
Yeah. And like, we will be exploring that space as a field. Yeah. I, I think-- I, I def-definitely agree with that. Um, you've created possibilities.

193
00:30:57.584 --> 00:31:08.484
So I think, you know, if I can paraphrase so that people can un- also understand, uh, you should not take the success of TypeSafe and Jeff as just like, well, you know, that is a new, new model type.

194
00:31:08.724 --> 00:31:16.984
Now we're done, we go back to business. Like, no. Like, actually there's-- there are like five other model types that you should be exploring and like let a thousand flowers bloom. Ab- absolutely. Right?

195
00:31:17.024 --> 00:31:22.984
Like, like- And some of that- Early Internet- Some of which you will probably also be- Of course, yes. Yes. Early internet energy. [laughs] I think it's back to tech utopia.

196
00:31:23.264 --> 00:31:34.344
You know, it's no longer like, oh man, like sometimes my coding agents work, but the-- all of the best ones are hoarded internally. Yeah. Right? It's like creation is back on the menu. [laughs] You know?

197
00:31:34.664 --> 00:31:42.184
Though it's gonna be a wild-ass world and, you know, buckle up. [laughs] I'm so, so jazzed about that. Yeah.

198
00:31:42.244 --> 00:31:54.304
Uh, a- I mean, and, and now you have the, the funding and the, the momentum to do whatever you, you, you envision there, uh, which I, which I think is like very gratifying to see you have after, you know, so long of, uh, of saying these things- Yeah, yeah...

199
00:31:54.344 --> 00:32:00.564
but like actually, actually show the world. I know. I just... Such a, such an interesting thing to be a tease the whole time.

200
00:32:00.824 --> 00:32:12.204
Like my talk like felt like it was a cliffhanger 'cause I didn't say how the automation would occur. Yeah. Um, uh, Sean reviewed our manifesto, and he's like- [laughs]... "It's a little bit vague in these parts."

201
00:32:12.264 --> 00:32:18.764
[laughs] And you know, like, "What's step one? What is, uh, what..." You know, "What is the intelligence product?" Well, I asked you for a model, and you were like, "Yeah, model coming." Yeah, yeah, yeah.

202
00:32:18.804 --> 00:32:28.704
And like, uh, well, I, I just-- I mainly objected to the word composable- Mm-hmm... but build product on God is fantastic. Thank you. Yeah. I, I-- we, we've really rallied around that.

203
00:32:28.764 --> 00:32:32.184
I'd like to think we're not entirely a cult like some companies are.

204
00:32:32.204 --> 00:32:43.524
[laughs] But like, we are like jazzed about what, what we're doing and like we are, like by brand is being practical, and like we are all like so super-duper practical. Yeah. It's really great. Yeah.

205
00:32:43.564 --> 00:32:51.624
So here-- and by the way, here, here is the, uh, the step, the, the, the secret master plan, right? Yep. The shape of machine native composable AI. It was your idea to make a secret master plan. I suppose.

206
00:32:51.664 --> 00:32:57.184
It's a, it's a Elon thing. When he started Tesla- Yeah... he was like, "Here's just what we'll do." But I, I, I'm giving official credit to you. Oh, thank you, thank you, thank you. Yeah. Thank you.

207
00:32:57.564 --> 00:33:08.384
Um, but like, uh, you know, you, you should have told me you're [laughs] you're, you're also gonna do this model launch, uh- [laughs] 'Cause you, like you told me, you told me half of the story, and then the other half, you didn't have the Doom demo at the time.

208
00:33:08.464 --> 00:33:14.644
Yep. You didn't have any numbers to give me. Yep, yep, yep. I was like, "Oh, what?" Well, the problem is I don't believe in benchmarking. Exactly. Right? Exactly.

209
00:33:14.684 --> 00:33:22.404
So like it is a thing that you need to feel, and like I think that this is the way to build long-term trust, even though it like hurt, it hurt us a lot, you know?

210
00:33:22.664 --> 00:33:33.264
Like, like last year when we did fundraise, no one believed us. Mm-hmm. You know? Like, like and they wanted just benchmarks and stuff, and we're like, "We're not gonna do that. We are principled.

211
00:33:33.304 --> 00:33:44.184
We're gonna stand by our guns. That rewards bad actors. I don't give a shit, you know, like what you want. Like this is who we are, and we are standing by that." So sorry is my- No, yeah. Yeah.

212
00:33:44.204 --> 00:33:50.444
Well, in, in some ways, I think like, uh, choosing the hard path, it-- but it-- you end up making the company that you wanna work in. Yep. Right?

213
00:33:50.584 --> 00:34:01.364
Otherwise, if you sell out, then you're just working in like OpenAI but with my people, right? Which is like [laughs] Yeah. Yeah. I mean, like I'm-- I, I don't have too many regrets on that, obviously. Yeah.

214
00:34:01.384 --> 00:34:13.424
Like it worked out so unbelievably well and, you know, like I, uh, uh, I was emotional last night when I was talking about like the reasons I left OpenAI.

215
00:34:13.884 --> 00:34:17.684
Um, and because like it actually had to change my wording after the launch.

216
00:34:18.184 --> 00:34:42.084
Um, my, uh, my, my phrasing was, "If an AI winter did happen and I did not do every fucking possible thing I could to like avert that, I would see myself as personally responsible both for, you know, the, the RLHF direction," which I think really widened overpromise versus underdeliver, and also not going all in on this because I think this is, this is where value is going to just be like printed.

217
00:34:42.524 --> 00:34:53.204
So, and it was really cool because I feel like the, the AI winter I'm worrying about is averted. You know? Like AI will be useful. It'll be used for automation. Mm-hmm.

218
00:34:53.244 --> 00:35:03.444
It's been less than a week, and like the numbers are already undeniable- Yeah... that it's like being used for real work and like there's... It, it's, it's the Wild West. Yeah. Yeah.

219
00:35:03.484 --> 00:35:13.824
Can you sh-- Uh, just, just if you have top of your head, what numbers are you seeing? Like what's, what's like sign-ups? Like what-whatever you can share. I'm actually not super on top of everything.

220
00:35:13.964 --> 00:35:19.164
Like the team is the ones who are telling me all of these things. Yeah. And I'm sure it's like changing every day, right? It's, it's, uh- But like- It's kinda nuts...

221
00:35:19.244 --> 00:35:27.464
if, if there's a milestone that you're like, "Well, yep, that's something we're hoping for. We reached it." I will say a milestone that we've passed is [censored] tokens per day. Nice.

222
00:35:27.564 --> 00:35:40.244
Um, and this is not like fleeting [censored] tokens per day. Yeah. This is like even at night, like it's constantly churning, so you know machines are calling it and not just people trying things out. Mm-hmm. So that is,

223
00:35:41.484 --> 00:35:53.024
that is so cool. [censored] tokens a day is a lot. Yeah. Um, so surpassing that is awesome. Sign-ups to me don't really matter, and actually, this was like a bit of a mistake we made, if I'm like totally honest.

224
00:35:53.384 --> 00:36:01.844
People on Twitter were calling us like marketing geniuses and all of that, and, um, that was just us. We don't have a marketer, also hiring.

225
00:36:02.204 --> 00:36:13.468
Um, and we were just being our genuine, goofy, like irreverent selves, and we w- we were just like offboarding people off the wait list so hard. Um, I Our platform team is so unbelievably, belly cracked.

226
00:36:13.548 --> 00:36:23.188
I think we have more n- up nines of uptime than Anthropic while having- [laughs]... the most unprecedented launch ever. Like, that is kind of nuts. Yeah, yeah. So, like, props to them. Yeah.

227
00:36:23.308 --> 00:36:31.848
Um, and, uh, the thing we didn't reali- So number one, wait lists, wait list signups don't matter for, like, a developer platform, in my opinion.

228
00:36:32.148 --> 00:36:42.108
You know, uh, I would guess that a large number of them are not even developers. So they go in, they try some queries, and a lot of people don't get it because they are not programming, right?

229
00:36:42.148 --> 00:36:45.768
Like, they're just like, "What? This is not a chatbot. Where, where's my ChatGPT 2?" Mm-hmm. Right?

230
00:36:46.148 --> 00:37:01.468
But if like, I, I haven't exactly calculated this, my sense is that if every single human being in the world, like, just wrote a couple of queries, that would be a rounding error compared to, like, one power user's for loop that is just, like, creating value.

231
00:37:01.768 --> 00:37:08.228
Yeah. And the thing we are- didn't realize with a wait list is, like, we can just w- off- offboard anyone off the wait list. It doesn't matter.

232
00:37:08.628 --> 00:37:17.548
The scary part is rate limits and then once people start getting value from that, uh, then they just want tons and tons of rate limits because this is what software is, right?

233
00:37:17.808 --> 00:37:26.848
Like, you spend effort upfront to specify your rote task, and then this rote task creates more value than it takes to put in, and then now that you have that- Set it and forget. Yeah. Exactly, yeah.

234
00:37:26.888 --> 00:37:35.668
You run it in the background. You make it a dependency, uh, to, like, other things. You can make, like, higher level stuff and, uh, like you just create so much value in the world.

235
00:37:35.868 --> 00:37:47.648
You know, early internet people probably did not imagine, like, the wonder of early 2000s internet, which is still not early internet. But, like, it's, it's through, uh, no offense, composability- No, no...

236
00:37:47.728 --> 00:37:57.128
um, that all of the crazy stuff happens, and I just really wanted to emphasize that in our manifesto. We are going for emergence. We are going for, like, being the catalyst.

237
00:37:57.388 --> 00:38:05.128
We're wanting to empower people, and we are going to do whatever we can for that, be it, like, Discords in our town hall with me wearing a garbage bag or not.

238
00:38:05.348 --> 00:38:12.268
[laughs] Um, and, uh, and podcasts and, and, you know- Hell yeah... getting, getting- Absolutely... like, 'cause I want the long form, right? Yeah.

239
00:38:12.308 --> 00:38:15.648
It is like, yes, we'll get past the, some of the superficial things, and then we'll go deep. And- Hell yeah...

240
00:38:15.808 --> 00:38:25.748
and people will really trust and understand your mission and, like, you know, the, the people that, uh, will resonate, that will end up joining you or, or, you know, uh, uh, buying you, uh, and- [laughs]

241
00:38:25.788 --> 00:38:31.268
No, no, b- uh, sorry. As, as a, as a customer. Oh, as a buy- [laughs] Okay, okay. As a customer, as a customer. Yeah, yeah, yeah. Sorry, sorry. That was funny. I'm sorry. Uh, sorry, I didn't, I didn't mean to say that.

242
00:38:31.608 --> 00:38:42.808
Um, but no, any- one, one, one version, one very flattering version of this, like, 36 million views of your launch video. Cool. Um- Up to 38 now. Uh, yeah, rounding error. Yeah. Uh, you know, uh, Navio still has got 74.

243
00:38:42.908 --> 00:38:52.548
Fable 5 got 57, so a- like, as far as, uh... And, and, and I, I didn't, I didn't do the stats for, like, original ChatGPT, like- Yep, yep, yep... which there was no video. Yep. So, like, up there, right? Yep.

244
00:38:52.568 --> 00:39:01.128
Like, as, as far, as far as, like, if you were to launch a NeoLab in 2026, I think you're, like, number one right now, which is, like, pretty crazy. [laughs] Yeah. Well, I actually would rather...

245
00:39:01.548 --> 00:39:09.708
I do have the shirt, like, your favorites NeoLa- favorite NeoLab's favorite NeoLab. Uh-huh. Um, I don't give a shit about being a NeoLab. I think being a NeoLab...

246
00:39:09.868 --> 00:39:20.188
Actually, we have a lot of, like, swag that's being a parody of a NeoLab. One of them, one of them I have is, like, NeoLab with product. [laughs] Which actually is not a NeoLab. Like, I don't care about that, really.

247
00:39:20.228 --> 00:39:25.628
Yeah. Um, what I care about is being a reliable dev platform. Yes. So, uh, appreciate the comparison- Yeah...

248
00:39:25.648 --> 00:39:35.608
but, like, hopefully we transcend past them, and we go back into, like, a thi- you know, like, you know, a revolutionary moment for developers and, like, this stable thing that people can rely on and trust. Yes.

249
00:39:35.868 --> 00:39:43.988
Uh, I mean, to that end, um, I mean, I, I think that's one thing that, uh, really impressed me about you guys is, is that, yes, you, you do talk about reliability.

250
00:39:44.328 --> 00:39:54.188
I thought it was mostly about calibration, which like we, you know, we talk about RLCD. Uh, but actually it's also about just, like, uptime and, and, uh, scalability and- Mm-hmm... all those things, right?

251
00:39:54.248 --> 00:40:04.608
They, they all sort of, of a kind. And nines. And nines. It's, uh, like, like- Which uptime is in my- But that, that's part of it. But, like, there's reliability in, um, like, how intelligent the thing is.

252
00:40:04.888 --> 00:40:15.488
Like, how consistently does it do the thing that you want? And I think that, like, the, the big reasoning models are very smart. In my opinion, they still lack reliability. Mm. I think there's many use cases where you...

253
00:40:15.708 --> 00:40:26.648
they look like they should be smart enough to automate their work. There is economic incentive to automate that work, yet still they're not reliable enough as at an intern because they're optimized for different things.

254
00:40:27.088 --> 00:40:38.108
Uh, so, like, I think that there's the reliability of being able to, like, trust the outputs, and also we are... Like, like, there are dimensions of reliability that we are not yet at that I'm, like, so excited by. Yeah.

255
00:40:38.148 --> 00:40:47.268
You know, like, I want to automate the easy work before the hard work, you know? Like, I think that that's just a common sense thing to do. Um, but to me, we will be sufficient...

256
00:40:47.608 --> 00:40:53.428
I don't know if there's such a thing as sufficiently reliable, but I wanna get so good that people don't even need to try the model to know that it'll work.

257
00:40:53.488 --> 00:40:56.288
It's like, that's, like, what flow state is in programming, right?

258
00:40:56.328 --> 00:41:08.568
Like, I'm just, like, writing queries because I need intelligence in here and, you know, like, when, for non-trivial branching, I can just write it in in, like a, like, like a typesafe system one query, and then get the result out, and it just branches accurately.

259
00:41:08.908 --> 00:41:19.128
Like, that would be so, so good. Like, that's the, that is the dream. Yeah. And that is, like, going to be, like, a long, long slog. Yeah. We're gonna go into your API design in a little bit- Ooh...

260
00:41:19.168 --> 00:41:28.668
just, just to give people examples and, like, maybe path not taken, that kind of stuff. Mm-hmm. Um, one thing up the front that I do wonder about in terms of reliability is I notice that there's no seed.

261
00:41:29.288 --> 00:41:42.348
Uh, there's no, uh... And, and so basically, same input, do I always get the same output? Hmm. So- And if not, why not? Oh, great question. So this is actually, like, a common question we have between, um...

262
00:41:42.668 --> 00:41:52.508
So reliability is actually a catchall. Like, whenever AI can't automate something, it's due to some form of reliability. Could be, like, type safety. It could be determinism.

263
00:41:52.868 --> 00:42:00.228
It just could be, like, it's, uh, it's jagged, right? So, um, reliability is a catchall. It- I just think that it's also a catchall for, like, what the North Star is.

264
00:42:00.828 --> 00:42:14.968
Um, re- determinism is, like, same inputs, same outputs. I do believe that this is, like, slightly interesting for unit tests, but I believe that to be the wrong North Star. I, I believe robustness is what people-

265
00:42:16.640 --> 00:42:22.920
I, I don't wanna tell people what they really want, 'cause that would be a little arrogant of me. [laughs] I believe that that is, like, the more important property.

266
00:42:23.320 --> 00:42:36.440
Um, uh, you want, um, given similar input, get similar outputs, and it's kind of wild how unreliable LLMs are. Hmm. Uh, like a, a way that we test this is you put, like, UUIDs in, in like little- Yeah...

267
00:42:36.460 --> 00:42:37.700
I think they're called nonces- Mm-hmm...

268
00:42:37.800 --> 00:42:52.040
in, in the prompt, and what you want is similar outputs from all of those, 'cause it's truly semantically the same question, and that is the part where you really want th- like, like, that robustness is where, like, people get, like, burnt with AI making decisions.

269
00:42:52.420 --> 00:43:04.200
So I think that is the- a super-duper important property. We could also have determinism. That is, that is a thing that can be available. As far as I can, like, mentally model for programmers, like it,

270
00:43:05.840 --> 00:43:21.140
it, it, it could be valuable for some use cases, so like please educate me, um, in comments or you. But, um, my- in, in general, it's easy... Determinism is something you can, like, trade off for better cost, you know?

271
00:43:21.179 --> 00:43:31.060
Like we are, we are constantly wanting to be on the intelligence per dollar frontier. We are doing, like, absolutely disgusting things to be there. You know, like this is, um...

272
00:43:32.640 --> 00:43:42.580
I, I shouldn't say this, but no one's here to stop me. [laughs] Um, you know- You have- you s- you s- you sign off on your own PR. [laughs] That is not how it works at this company.

273
00:43:42.580 --> 00:43:54.580
[laughs] Um, I believe for this week, um, my chief of staff, K, is the most powerful person in tech. Yeah, and shout out to K for organizing this. Holy sh- Yeah. Holy shit, she is so fucking competent and powerful.

274
00:43:54.820 --> 00:44:03.900
She's incredible. Um, uh, I mean, she sucks. Don't poach her. [laughs] Um, but, um...

275
00:44:04.140 --> 00:44:12.220
So I try to be a bit more filtered, but, like, people are telling me, "Don't call it a Frankenstein's monster of models," but because that has like negative implications.

276
00:44:12.600 --> 00:44:19.400
I think Frankenstein's monster was, like, the good guy in this whole... I mean, it was innocent, right? I didn't read it, okay. [laughs] I'll, I'll confess. Okay.

277
00:44:19.440 --> 00:44:26.660
That, that one facial expression, I- There's a- My cards are on the table... decent Jacob Elordi movie if you wanna see- I-... sort of the- You-... the adaptation. Anyway.

278
00:44:26.699 --> 00:44:36.340
Th- y- you have no idea how little time I have right now. Yeah, yeah, yeah. My priorities are sleep, you know? Um- Developers, developers, developers. Developers, yes. [laughs] Uh, developers, developers, developers.

279
00:44:36.340 --> 00:44:47.820
[laughs] Um, but, um, yes. I- I- we, we do, like, absolutely disgusting things to be on the Pareto curve of intelligence per dollar, and we are going to keep doing that. Yeah.

280
00:44:47.860 --> 00:44:59.240
We're gonna be doing crazy-ass stuff, and I think people really need to think outside of the box. Like, like, part of the reason why surprising is, like, people are thought inside the box, and we continue to do that.

281
00:44:59.640 --> 00:45:08.580
Um, as of right now, we are obviously the best at this, and we want to continue being the best at that whole thing. Yeah. So, um, wait, where did, where did we tangent from? No, so, so I- Yeah...

282
00:45:08.640 --> 00:45:16.100
asked you about, uh, will you have seeds and determinism? Oh, yes. So- And then you basically defined reliability and, like- And robustness, yes... from how you see it. Yes.

283
00:45:16.120 --> 00:45:23.020
But, like, determi- well, like- I have an, a robustness example that's, that's, uh, real quick I can show you. I would love that. I would just say one thing. Yeah. We can make a deterministic model. Exactly.

284
00:45:23.060 --> 00:45:32.000
Like, like we're hap- if people can convince us that that is a valuable thing to do- Yeah... and we don't have a gigantic GPU shortage- Yeah... um, we can happily make all of these models.

285
00:45:32.280 --> 00:45:42.380
We live to please, um, and rev- and, and rev- revolt, revolute. Uh, uh- [laughs] You will throw over everything, except you'll do it in a nice way. Yeah. Uh, and fun- and have fun with it.

286
00:45:42.420 --> 00:45:50.720
So, so like, uh, determinism could be on the cards. Yeah, yeah. It just gets you less intelligence per dollar. Yeah. Uh, well, just having seen the trajectory of OpenAI and Anthopic, uh, you will...

287
00:45:50.780 --> 00:46:00.300
Uh, uh, now just trust me now that you will be peer pressured into doing it. Uh, so, like, just people will want it even if they, if you tell them they don't need it, they'll still want it. So, like, yeah.

288
00:46:00.340 --> 00:46:13.600
That, that's the TLDR of- Okay. Yeah. Okay. Yeah, yeah. I will love to... Maybe one day we will see how that happens. Yeah. I've been told I'm, um... They, they, they say that part of our brand is being unshakable. Oh.

289
00:46:13.620 --> 00:46:20.480
And they say that that's just the nice way of saying stubborn. Stubborn, yeah. Yeah, exactly, and I'm a very stubborn person. I don't think we could have done it. Yeah. Yeah. No, but so, like, I, I...

290
00:46:20.520 --> 00:46:27.780
Okay, but I tr- I, like, have argued with you before. Yeah. And I know that- And you've been right about developers every time. [laughs] So okay, I give up. You win. You win.

291
00:46:27.780 --> 00:46:37.779
[laughs] I, I'm sold that I've argued with you before. [laughs] No, no, I'm just saying, like I, I think that you can hold your ground while also, like, if I give you the right evidence, you can, uh, not...

292
00:46:37.820 --> 00:46:44.980
You can sort of throw away your priors and be like, "Yep, like that actually makes sense to me." Yep. And so, like, y- you know, just, just trust your own gut on this. Yeah. Yep, yep, yep.

293
00:46:45.000 --> 00:46:56.200
Um, I'll bring out some- But I suspect, though, that we will be GPU constrained for a very, very long time. Very long, yeah. And anything that has less intelligence per dollar means it consumes more GPUs- Yeah...

294
00:46:56.220 --> 00:47:01.300
for the same intelligence, which is... You know, like our goal is not to onboard companies.

295
00:47:01.460 --> 00:47:14.380
Like, it's, it's valuable, but, like, our goal is to have people, like, experiment and do weird shit, and we need, like, we need to, like, get, uh, it to as many hands as possible and, like, starting, like, the California Gold Rush for that.

296
00:47:14.660 --> 00:47:21.060
Hmm. Yeah. I think there is right now, yeah. Yeah. Um, just a word of caution. I mean, I'll, I'll just say it- Ooh, hell yeah... because somebody's thinking about it right now. Mm-hmm.

297
00:47:21.120 --> 00:47:24.640
Which is when you say things like, uh, "We will not commit to deterministic models.

298
00:47:24.700 --> 00:47:32.980
We, we will, uh, we'll do whatever it takes for intelligence per dollar, and we are, are- we are facing GPU constraint," people are thinking you may quantize your models, right?

299
00:47:33.080 --> 00:47:42.760
Like s- like whatever you had at launch, you may quantize down in, to reduce the quality, um, in order to, to free up, uh, memory or bandwidth or whatever, right? Mm-hmm.

300
00:47:42.800 --> 00:47:53.720
And so, uh, you should probably, uh, have some kind of promise, which you don't have to make now- Yep... about, like, we will uphold model quality at launch. People, like... So this is like when people, when, when we...

301
00:47:53.780 --> 00:47:57.760
I, I mean, you were at OpenAI when you launched- Yep, yep, yep... all these, all these APIs, and even Claude as well.

302
00:47:57.780 --> 00:48:06.920
Like, uh, when they first launched the models, the model strings, um, did not stay the same model at all times. Yep. Right? You have versioning in your models. That's great. Yep, yep, yep, yep, yep.

303
00:48:06.940 --> 00:48:15.800
But, like, you should, you should publicly commit to some kind of, like, once a thing is launched, we don't change it. We will not change our models when we deploy them. That is insane. We care about developers.

304
00:48:15.900 --> 00:48:25.880
L- l- like it makes sense if you're... So- Doing something like that, a-again, this is the problem with a for-first party product and an API. It makes-- You can do whatever you want in a first party product, right?

305
00:48:26.119 --> 00:48:32.720
Like, like more power to them, whatever gets that experience. That is fine. With an API, you obviously can't do that.

306
00:48:33.100 --> 00:48:41.920
But I will say that we, uh, plan to move a lot faster than many people are used to model providers, uh, doing things.

307
00:48:42.340 --> 00:48:52.340
So we will be launching new models a lot faster than people think, and we are not promising long-term support for the models because we think that there's lots of improvements to have.

308
00:48:52.820 --> 00:49:07.380
So, um, there is a world that we might temporarily LTS what is right now Jev 1.13.0. We might do that 'cause so many people are using it, and I know developers hate breaking dependencies.

309
00:49:07.740 --> 00:49:13.900
The alternative is fracturing our fleet, and that is a very bad vibe for everyone. It's gonna be- Yeah, you can't have, like, a hundred different versions of the model.

310
00:49:14.000 --> 00:49:25.240
Exactly, and if we're iterating very fast, there would be a lot of those versions as well. Yeah. So, um, we, we do want to have not just a LTS-supported thing eventually, long-term support.

311
00:49:25.600 --> 00:49:34.980
Um, we want a really sick way of doing that. We have, like, research stuff cooking in that direction, and I think it's gonna be the most pro-developer thing ever. Yeah.

312
00:49:35.060 --> 00:49:43.520
Um, but it is not yet our current models, and I'm not promising that we'll be able to keep the exact same models. They will get smarter every time, for sure. Yeah.

313
00:49:43.580 --> 00:49:55.580
And my sense is that even our model iterations, where it already is smart, um, it-- between model versions, the, the changes tend to be even smaller than the string models calling them twice.

314
00:49:55.900 --> 00:50:02.080
But when, when it-- we go from, like, you know, jagged to, like, wow, um, that is where the big deltas are. Yeah.

315
00:50:02.720 --> 00:50:12.980
Uh, one, one thing, one thing that's beautiful about LTS-ing models is that actually you can also port them to other silicon. Mm-hmm. Uh, I don't know if you've thought about this. No comment. [laughs] Okay.

316
00:50:13.020 --> 00:50:19.900
So I care about intelligence per dollar. Yes. Right? Um- But speed. What? Speed as well. We'll see. Yeah. We'll see.

317
00:50:19.940 --> 00:50:24.860
Honestly- I mean, it's a whole part of the inference tech tree that is, like, I mean, exploding in the past year, right? Yeah.

318
00:50:24.880 --> 00:50:33.420
Like, that you can, you can move to, like, a Cerebras- Yep...um, an Etched or whatever and get, like, the hundred, hundred thousand times speed up. Yeah.

319
00:50:33.620 --> 00:50:43.700
Like, I-I think that intelligence per second is, like, a different metric, and we've even talked about, like, things like intelligence per dollar times second and, like, metrics like this.

320
00:50:44.420 --> 00:50:56.980
My guess on, like, Jevons paradox occurring, or at least the Jev series of models, and the thing I, like, hunt people down about internally is, like, I don't care how much smarter it is, it needs to be in the pre-different tier.

321
00:50:57.300 --> 00:51:08.380
So like, like that is what the brand of Jev is. It is the best thing at intelligence per dollar. For intelligence per second, we'll see. I, I think that it's an intriguing thing.

322
00:51:08.820 --> 00:51:20.900
I know that there's many industries that are, like, extremely dependent on real-time stuff, and they will-- like, like, intelligence per second means tons of dollars for them. But we'll see.

323
00:51:21.020 --> 00:51:31.440
I, I, I'm, I would love to, like, do both and, like, have the market correct me either which way. Yeah. You know? Yeah. Like, like, uh, I, I, I, I would love to be informed by people. Yeah, totally.

324
00:51:31.600 --> 00:51:41.120
Uh, and it's not, it's not just real- about real time, right? It's also about scale because, um, at, at scale, every m-microsecond is just multiplied by billions and trillions of times.

325
00:51:41.120 --> 00:51:50.160
It depends on how background it's running, right? Yeah. Like, if it's, like, a big background, like, database MapReduce query, the latency might not matter so much as, like, the cost- Yeah...to get intelligence from it.

326
00:51:50.440 --> 00:51:59.340
But, like, if it actually is, um, something more real-time, like user-facing, you have budgets, uh, like between one hundred milliseconds and one millisecond that are, like, totally magical.

327
00:51:59.740 --> 00:52:09.800
And actually, even if you were below a hundred milliseconds, if you could half that time, that means you can get double the intelligence or sequential intelligence calls to have like a, like a phenomenal experience.

328
00:52:10.040 --> 00:52:22.700
Yeah. So, uh, right, that is definitely happening right now. It is super-duper cool. I love the intelligence per second use cases, but I don't think that that will be Jev's niche. Okay. Yeah, fair enough. Yeah.

329
00:52:22.720 --> 00:52:33.220
When thinking about the promise of faster and cheaper- Mm-hmm... [clears throat] um, typically the other, uh, the, the trade-offs that other models are offering is faster but more expensive. Yep. Right?

330
00:52:33.480 --> 00:52:42.960
And so you're-- Like, one of the reasons I was thinking about why is Jev resonating so much is that you've done the faster but cheaper side of the quadrant- Yeah...which is very, very unoccupied.

331
00:52:43.220 --> 00:52:52.900
Uh- Yeah...while holding intelligence, like, somewhat constant. Uh, yes. It, uh, that, that, that's a very load-bearing statement while holding intelligence constant. That's the hard part, right?

332
00:52:53.240 --> 00:52:59.740
Like- Which unfortunately, like, so basically you, you refuse to have the, to like, like do any public benchmarks, or you don't like any public benchmarks about it.

333
00:52:59.760 --> 00:53:08.140
And I will-- I've actually pre-committed- But you need some internal sense. Say again? You need some internal sense of this in this- Oh, of course. Yeah. Of course. We have, we have our own internal evals, for sure.

334
00:53:08.200 --> 00:53:15.600
Yeah. But it takes a lot of discipline not to game those, and it needs to be, like, a top-level priority to not game them. Yeah. Of course we do that, right? Yeah.

335
00:53:15.620 --> 00:53:22.020
Like, how else can we make the guarantee that our models are in the pre-different tier of intelligence per dollar, right? Mm-hmm. Like, we're not flying blind in there, right?

336
00:53:22.380 --> 00:53:32.420
If we're doing, like, completely weird things with different costs or whatever else, you know, like, how do we compare them? We plot them and get, you know, we try to figure out, like, what is the best for the users.

337
00:53:32.680 --> 00:53:46.600
Yeah. So we for sure measure them. I'm not anti-measuring. Um, but it, it, it's extremely dangerous when you have, like, any alternative incentive, and this is the one thing that, like, I kind of rule with an iron will.

338
00:53:46.780 --> 00:53:57.160
Maybe my coworkers might think I rule many things with an iron fist. But to me, like, not shitting ourselves, uh, uh, um, about how smart our model is is one of the most important things there. Yeah.

339
00:53:57.200 --> 00:54:04.880
Like, we need to be truth-seeking. Yeah. Yeah. Uh, agree. Agreed. Um, okay, I wanted to go over some, uh, details on the, uh, API choices. Ooh.

340
00:54:04.960 --> 00:54:11.980
Um, mostly because this is the only podcast that will ask you these, these kinds of questions. Oh, hell yeah. Hell yeah. Um, so you have three primitives. Yeah. Um, choice, score, null.

341
00:54:12.240 --> 00:54:25.960
First of all, null, where is that from? Um- Is this, is this like a term in the, in the literature or what? I've never heard- Uh, now, now it is. Yeah. Um, we debated this a lot. We debated this a lot. Um, it is-

342
00:54:27.076 --> 00:54:39.976
You know, it is bool-ish, right? Like true/false. Um, it is- But it's continuous. Yes, exactly. Uh, so, so first, the origin of the name is Bernoulli. Ah. Yes.

343
00:54:40.136 --> 00:54:50.476
So it's, that's why it's even spelled that weird way, that is like a subset of the name Bernoulli from like a Bernoulli probability. Right. Which is actually what that is. Um, so that is the origin of it.

344
00:54:50.736 --> 00:55:04.536
We were debating this a lot. We liked Pbool, we liked Pool. We, we, we wanting to call it like a pool party, but then no one let me. Um, you know, we had like a bunch of like other arguments about that. And

345
00:55:06.076 --> 00:55:19.156
null, we figured, was like the best thing. The... Our, our rationale, and like this was actually the same thing with Jev too, um, is that we think that we are like an irreverent, insane bunch, and programmers don't care.

346
00:55:19.476 --> 00:55:28.616
You know, like if Jev is just gonna be a string... We didn't expect it to catch on or even have puns or anything like that, right? Actually, there was a lot of hate on the name internally.

347
00:55:28.936 --> 00:55:43.456
They've all apologized except for one person, um- Still holding strong. Yes, uh, our mutual friend. Um- Okay, okay. Yes, yes, yes, yes. Um- I respect her for that. Yeah, uh, yeah. Uh, she wanted Jev to be called Meow.

348
00:55:44.396 --> 00:55:51.436
She would, of course. Yes, of course. Okay. You, you in there, you in there, [laughs] But, but like, um, yeah, null is...

349
00:55:52.136 --> 00:56:00.596
We, we had to make a new concept for this thing, 'cause if it was a bool, it would be confusing to people. So actually, all three of these are actually new concepts. Okay.

350
00:56:00.616 --> 00:56:08.516
These are not types that exist in programming, and that was intentional because they map very closely to types, but they're not quite that.

351
00:56:08.896 --> 00:56:17.116
A score is not an int, so if you had like instructor or pedantic or whatever map ints or floats into scores- Mm-hmm... you'd get a little bit cooked, you know?

352
00:56:17.156 --> 00:56:25.756
And, and like we, we were really erring on the side of clarity over the side of like making people like easily understand what's going on. I mean, don't you worry about that?

353
00:56:25.836 --> 00:56:37.176
Don't you want things to integrate directly into things that people are already using? Yes. Yes, we do, and actually, I think that, you know- You have integrations with like other SDKs and stuff. Yeah. But you have...

354
00:56:37.276 --> 00:56:47.356
Sorry, you have your own SDKs. Yep. But typically, for example, as a developer relations person, I would be very obsessive, like, "Yes, here is how you use, uh, you know, Jev with Instructor." Yep. "Here's how..."

355
00:56:47.396 --> 00:56:55.876
You know, that kind of stuff. We might have that somewhere. [laughs] I'm so behind on everything that- Someone would do it for you in the community. Oh, yeah. Yeah. Now that, now that you're successful- Yep...

356
00:56:55.916 --> 00:57:04.696
people will be like, "Oh, that's cool." Cool. But like, you know. Anyway- I, I don't see that as binary either. Yeah. I actually see success as a score- Yeah... and there's always more to climb- Yeah, yeah...

357
00:57:04.776 --> 00:57:16.176
in like how much we can like be there for our community. J- just to be clear. And, um, I'm... This, this section is stressful because I didn't review the docs- [laughs]... and they're constantly changing. Okay.

358
00:57:16.256 --> 00:57:21.276
But, but- But to me, scores- Yes... do exist. So, so, so scores are similar to like LM judging. Mm-hmm. Right?

359
00:57:21.516 --> 00:57:30.956
So like if you want to call it like a judgment, I guess you could, but like that, that is like the, the, the way people con- already use this type of thing, right?

360
00:57:31.216 --> 00:57:48.856
Like maybe a null could be like a probability, but everything for us is a probability, and a choice is actually closest to a function call, but a function call is like an extremely disgusting thing that, um, if you want OpenAI juice, uh, sauce, uh, uh, tea, the, the...

361
00:57:49.156 --> 00:58:00.596
We should go back into that later. Like if, like, like a choice is just like the right way of explo- uh, of exposing like a switch match statement- Yeah... within code. So, so it like maps cleanly to an enum. Yep.

362
00:58:00.636 --> 00:58:06.196
And you can choose to hydrate it into a function if you want. Yes. And like in the enum, choice is the important part- Yes... of that.

363
00:58:06.516 --> 00:58:16.076
And, uh, like actually, I think these map all into like programming primitives where like choice maps into like a, like a switch statement on an enum. Oh. Um, nulls map to if statements. Yeah.

364
00:58:16.136 --> 00:58:28.576
And scores map to sorting or thresholding at a greater than or less than. Okay. And this has been always what the vision is. Like there will be more types, and they will map into programming primitives. Yeah.

365
00:58:29.476 --> 00:58:38.276
Um, any other... So, uh, any nuance you wanna go through for... L- literally, this is for the Jev people who are like deciding to really invest in Jev. You are the expert, right?

366
00:58:38.676 --> 00:58:53.216
Um, I'm just like wanting to provide more background for them on, uh, API choices, um, you know, how they should use some of these, these things like legends, confidence, um, how critical in your testing, um, you know, like how, how...

367
00:58:53.556 --> 00:59:01.716
Like just any sort of like pro tips that you like want to offer people- Hell yeah... when we're down at this level. Thank you. I love this. Um, no, no, no. This is- This is why we're here. Hell yeah.

368
00:59:01.776 --> 00:59:12.696
I didn't expect this, and actually I... No one has asked me this, um, in probably like months when I was like onboarding like our DevRel. Okay. Um, so sick. Um,

369
00:59:13.816 --> 00:59:19.876
so our model is designed for being like deep in the insides of computer programs in the future.

370
00:59:20.136 --> 00:59:29.996
We like unironically believe that this will be much more massive than anything people are even considering today, and our model might not be ready for that, but we are like continuously working for that future.

371
00:59:30.316 --> 00:59:37.736
It will never be good enough at these shallow tasks. Um, uh, uh, sorry. It'll never be g- like we're not just gonna cl- keep on climbing the shallow tasks.

372
00:59:38.056 --> 00:59:47.616
We want to be deep in the guts of programs because that's how you make software powerful. All the s- all the types inside of our, um... This is an actually an output. Mm-hmm.

373
00:59:47.816 --> 00:59:58.156
Um, but all the, um, all the parts, uh, of like the, the, the input, like the state, the instructions, the criteria, all of them can be structured JSON objects. Mm-hmm.

374
00:59:58.336 --> 01:00:06.296
That way, like programs can like insert them in the right spot, and you don't need to like put things- This is-... into templates. Exactly. So if ever...

375
01:00:06.536 --> 01:00:14.456
I, I think people don't read into this part enough, and they think it's all strings, and that's, that, that's fine. Um, but these are all meant...

376
01:00:14.856 --> 01:00:23.076
Like, like I would say that if you're using like a template, like turning it into like a system message or something, you are thinking in like the old way, you know?

377
01:00:23.416 --> 01:00:29.096
We should be making things as easy for computers to understand because that structure is truly there, right?

378
01:00:29.136 --> 01:00:37.032
Like it would be weird in like- Like a programming language to have, like, all of your numbers in, then you pass it into, like, you turn it into a string.

379
01:00:37.112 --> 01:00:47.031
Normally, you do that for printing when, when you have a human in the loop, right? But for, like, within the computer, you want to be passing, like, nested structure that is semantic all around.

380
01:00:47.331 --> 01:00:49.212
And we are really gonna be optimizing our model.

381
01:00:49.272 --> 01:00:56.892
It-- the model's pretty optimized for this, but the thing is, every different nested level of structure is harder to reason about, and we want-- we are really cooking hard in that direction.

382
01:00:56.972 --> 01:01:05.772
I think people should keep cooking in that direction because it makes the code, like, so much more legible and beautiful and, like, agnostic to, like, the, the implementation details.

383
01:01:05.812 --> 01:01:11.912
It's like, "Here is my state," you know? Like, "Here's my function state." Like, think of, think of it as, like, an AI function.

384
01:01:12.432 --> 01:01:16.992
Which subsets of my state, which is, like, all the variables you have available, should I pass in here?

385
01:01:17.352 --> 01:01:23.272
System messages are, like, disgusting global variables where you just put everything in there and you put all those instructions at once. Slop, yeah.

386
01:01:23.312 --> 01:01:32.692
And then y-you know, like, you s- you hope that every single instruction gets nailed instead of asking the questions in parallel. Okay. And, uh, also I would recommend...

387
01:01:33.112 --> 01:01:44.952
I, I-- and I truly say this not from, like, a, like, a it-makes-me-money perspective. Um, I, I truly recommend asking lots and lots of questions. Break them down, make them smaller, and, like, really decompose.

388
01:01:45.192 --> 01:01:57.312
Like, no matter if the models can do it today or not, I believe that the biggest, like, um, s-saving grace of, like, what's happening this week will be people's code bases, AI code bases, are gonna be so much better.

389
01:01:57.752 --> 01:02:04.592
You know, like, if you decompose problems into simple decisions, every single one of these things is extremely evaluable.

390
01:02:05.012 --> 01:02:14.792
Like, l-like, a AI beforehand is big system message, and then maybe you have, like, another big AI- Big output, yeah... to see, like, if it actually does this. That's nuts, you know? [laughs] And it's just kinda crazy.

391
01:02:14.832 --> 01:02:18.752
Like, it, it, it, it, it's-- that was our Stockholm syndrome, right? But, like, that's kinda crazy.

392
01:02:18.812 --> 01:02:30.452
Like, if you wanna say, like, "Hey, don't read this subdirectory," or, "Don't pass any API keys to DeepSeek or whatever else," like, that should be programmatically basically guaranteed.

393
01:02:30.732 --> 01:02:37.772
And you know, you'll never have guarantees of any machine learning model, but, like, by breaking it down, you can actually s- you can actually measure it, right?

394
01:02:37.772 --> 01:02:46.472
Like- Yeah, you can verify that it was actually called or- Yes, and, like, our model, our model, like, the, the interface itself is so verifiable. This should be like a sigh of relief. Mm. You know?

395
01:02:46.512 --> 01:02:52.272
Like, it's, it's, uh, it's just gonna lead to way better engineering. Yeah, I think I get that.

396
01:02:52.372 --> 01:02:59.572
Um, and, and so, you know, one of the reasons people didn't used to do this in the past is because they would just call a, a small LLM, right? Yep.

397
01:02:59.632 --> 01:03:04.272
And it's still too slow, it's still too expensive versus chunking everything that... I've done exactly this myself. Yep, yep, yep. Right?

398
01:03:04.292 --> 01:03:13.612
Like, I, I benchmarked, here's a pipeline that throws everything in system prompts and then just gets one big output versus break it down into 100 different things. It was slower, more expensive- Yep... not as good.

399
01:03:13.772 --> 01:03:23.372
Yep, yep, yep. Right? Uh, th-th-then that happens all, uh-- Yeah. It's-- a-and it's, like, super inconvenient. It's unwieldy. Why not just put it all together? You kind of end up repeating some stuff between questions.

400
01:03:23.392 --> 01:03:33.852
Yeah. So it's, like, maybe, like, you know, inefficient or something like that. But then it results in something that is very hard to rely on. Yeah. And software doesn't need to run in the background.

401
01:03:33.912 --> 01:03:46.512
It would break my heart if our stuff couldn't run in the background. Is there a way to break things down that you guys have found that works versus, uh, what you thought worked and doesn't work? Interesting.

402
01:03:46.732 --> 01:03:53.852
Because, like, people are just gonna be exploring this, you know, now that you've said it. Like, they would use this as a reference and be like, "Okay," like, "that's how I'm supposed to use Jeff." Yep.

403
01:03:53.952 --> 01:04:06.872
Um, then the question is, how do you break things down? [laughs] Interesting. I, I like to break things down into its, like, its smallest semantic unit. Yeah. Like, what is the lowest level thing? I try to never have...

404
01:04:07.332 --> 01:04:19.512
Uh, I've probably queried, uh, the, the model the most, uh, among anyone. Mm-hmm. And, like, I try to, number one, in my, in my queries, this is, this is a lot more like the way I prompt things.

405
01:04:19.812 --> 01:04:27.252
Like, I make it really, really structured and explicit, and in the questions, I always-- I like the back ticks, but, like, it works for all of them, you know?

406
01:04:27.612 --> 01:04:36.632
Uh, like, be really clear what I'm referring to because we wanted the model to be really literal because when you program, you want things that instruction follow really, really well.

407
01:04:36.952 --> 01:04:46.632
That is what the art of programming is and what AI does is expanding the things, the kinds of instructions that can be followed. So, uh, I'm a fan of doing that. I, I...

408
01:04:47.692 --> 01:04:58.092
Sometimes I'm a little lazy, and I, like, I have, like, more, like, hybrid things, but, like, I think that for, like, really big production things, you just want to, like, keep on adding more questions, and you wanna make it really easy to add more questions.

409
01:04:58.472 --> 01:05:12.572
You know, be really, really precise about all of that breakdown and then have the code to have the exact behavior you want. If I could give, like, a tiny little example of this, um, is, um, like refusals, right?

410
01:05:12.612 --> 01:05:16.732
Like, uh, I'm not gonna talk about why we don't refuse. I might have done that already. Yeah. It's, like, all a blur. You did already, yeah.

411
01:05:16.812 --> 01:05:23.532
Um, but, uh, like, for refusals, I don't think you should ask, "Should I refuse here?" You know? Uh, that's a really...

412
01:05:24.072 --> 01:05:28.592
It, it-- the, I think the answer will be pretty good because, like, that's a system one compatible task.

413
01:05:28.892 --> 01:05:40.752
But I think you're way better off, like, asking many different independent questions about, like, the different situations you can refuse about because instead of having to, like, just guess based on you, you know, you can actually specify what you want.

414
01:05:41.132 --> 01:05:48.712
And beautifully, you know, like, and I think this is, like, truly, really beautiful, if you find a situation where it's like, "Oh, it didn't refuse because of this reason.

415
01:05:48.792 --> 01:05:53.192
I didn't specify this part of the task," that is awesome. That's what software engineering is about.

416
01:05:53.252 --> 01:06:00.512
Like, you fix the bug by adding that question in, adding the threshold, maybe remembering that as a test case, and now it is just solved forever.

417
01:06:00.552 --> 01:06:08.192
Like, your software can't forget about that, like, in the prompt because of context rot. It is just there, and you can, like, just keep measuring that forever.

418
01:06:08.672 --> 01:06:16.752
You know, and if the models are not perfect at some of these things, you can choose what threshold you want for all of these factors based on real examples.

419
01:06:16.812 --> 01:06:23.492
It's like, it's like ML without the ML, and you can just do it for, uh, for anything. And, like, there might be some things the model's not good enough yet, right?

420
01:06:23.532 --> 01:06:31.612
Like, I would-- I'm a little bit afraid when I see people doing trading with the models- Mm... like, um, automated trading. Mm. Um, it, it looks cool.

421
01:06:32.072 --> 01:06:40.992
I th- I, I just think that people should leave it to the professionals. [laughs] Um, but, like, that's just a very hard, high-level task that maybe the models aren't good enough yet to figure out. Yeah.

422
01:06:41.072 --> 01:06:45.576
Uh, well, I, like, uh, even if they were, then they would It, it suddenly wouldn't be 'cause of efficient market.

423
01:06:45.976 --> 01:06:55.816
But, like, that's one of those things where, um, you can, like, break it down into things and just evaluate them, and you might be like, "It's not smart enough at this. Maybe we don't deploy it yet for this version."

424
01:06:56.076 --> 01:06:58.636
Yeah. Or we make a trade-off, or we err on the side of safety.

425
01:06:58.936 --> 01:07:10.996
Or like, hey, the models are not good enough at, you know, like, detecting, like, this weird combination of, like, sarcasm with a VIP customer, that this is when we escalate to a human, and that's what confidence estimates are about too.

426
01:07:11.616 --> 01:07:21.436
Okay. Uh, great, great answer. I think, uh, one thing I'll, I'll mention very quickly, which, uh, I don't expect that you have as too long of answer for- [laughs] You don't... is, uh, well, no, no, no, no.

427
01:07:21.676 --> 01:07:30.776
It's just- [laughs] Yeah... it's just, um, typically, like, you are still relying on thresholding as, like, the, the lever that the user can pull. Mm-hmm. But what if just the, the, the, the calibration is wrong, right?

428
01:07:30.856 --> 01:07:35.456
Like, you're s- saying your, your calibration's perfect. But- I didn't say that... I, I mean, like- I didn't say that...

429
01:07:35.516 --> 01:07:43.856
so is it ki- perfect cali- and then good calibration means, like, lower value is lower, like, sort of probability lower. Higher value is probably higher. But it could be wrong. Of course, of course.

430
01:07:43.876 --> 01:07:52.496
It could be locally misaligned. Yes. And so then I would want to fine-tune it or something, right, which you don't offer, but you could. Uh, I, I, I- We could... again, see, this is a short answer- Yeah. Oh...

431
01:07:52.516 --> 01:08:00.716
which is you, you don't have it right now. [laughs] Oh, do we want to offer fine-tuning, is it the question? That could be, that could be one version of it, or you could have a, a different knob. Yeah. Right?

432
01:08:00.776 --> 01:08:09.756
Uh, where, like, because, like, n- right now you're, all you're saying is, like, if something's wrong, uh, skill issue, you should, you should just change the prompt again or break it down even further, or you change the confidence.

433
01:08:09.836 --> 01:08:18.536
Yep. Those are my two options. Yep. Right? And that doesn't feel super satisfying if your model is just getting it wrong. Yep. And it will, it will get many things wrong, to be clear. Right.

434
01:08:18.576 --> 01:08:25.876
We have, like, a report issues button, complain to us in Discord. We want to make it a lot better. Every single model version will be, like, notably better. Yeah.

435
01:08:25.916 --> 01:08:35.856
We will stop shipping them quickly if they weren't getting big improvements. So number one, that is, like, totally reasonable. I think that that's simply pragmatic to admit that AI is imperfect at some stuff, right?

436
01:08:36.216 --> 01:08:42.036
I do think we'll find use cases that they are, like, good enough at, and good enough kind of depends on the use case, right? Like,

437
01:08:42.956 --> 01:08:56.736
human beings can do a lot of work despite being bad at that work because their EV is quite high, and presumably with the right thresholding and everything, there probably is, like, large amounts of work that could be done even if mistakes are being made.

438
01:08:57.416 --> 01:09:08.276
Um, on the question of fine-tuning, I could imagine, I could imagine it in the cards. I do have concerns because, like, uh, in the what people need versus what people want category, um,

439
01:09:09.676 --> 01:09:12.556
like, I think general models tend to be really...

440
01:09:12.796 --> 01:09:25.636
Like, like, again, there's, there's the je ne sais quoi of generality that making it good at, like, a million other tasks than this one narrow task might make it better at edge cases in that task, which I'm, I would be a little bit afraid of, you know?

441
01:09:25.856 --> 01:09:38.856
Yeah. Um, I, I, I could imagine it is, is my answer. Uh, I'm endlessly practical on these things. I want everything. Like, my vision of the world is j- I, I... There's, there's so much we want to be building. Yeah.

442
01:09:38.876 --> 01:09:46.356
But also, like, I would not want to ship something that is, like, a giant foot gun, uh, like some other AI companies, which, yeah.

443
01:09:46.376 --> 01:09:53.516
[laughs] Well, you know, so both OpenAI and Clau and, and I think even Gemini have rolled out fine-tuning and then took it back. Yep.

444
01:09:53.636 --> 01:10:05.396
Uh, which is an interesting, um, observation that pretty much fine-tuning is now in the domain of open source models. Uh. [laughs] Yes. Yes. Um, I do know about that.

445
01:10:05.476 --> 01:10:10.376
And, like, it was kind of crap, so, like, that's probably better- Yeah... that they took it down.

446
01:10:10.376 --> 01:10:15.076
Yeah, so it could just be a foot gun, and then telling people that fine-tuning it is probably the wrong way to go is great.

447
01:10:15.416 --> 01:10:23.156
Another interesting answer could be that, like, well, our model is so different, like, you know, j- the scene and the quantization doesn't, doesn't apply to us. Yeah. Output tokens doesn't apply to us.

448
01:10:23.616 --> 01:10:32.236
Fine-tuning also doesn't apply to us. [laughs] Well, a- actually, I, I, I'm super open to that possibility. Yeah. Like, uh, my, uh, this is not a promise. This is a desire.

449
01:10:32.416 --> 01:10:34.596
Just so to, to make it clear, I like to be really honest.

450
01:10:34.956 --> 01:10:46.816
Like, I think that, um, as intelligence per dollar gets cheaper, cheaper, cheaper, cheaper, I think that we could get really, like, small approximate things that hopefully are proxies for intelligence.

451
01:10:47.076 --> 01:10:59.436
Like, is there a world where people don't write regexes anymore because, like, you know, the intelligence per dollar that uses AI is cheaper than, like, the complexity of a regex? You know, that would be kind of sick.

452
01:10:59.616 --> 01:11:10.896
I would love that, you know? And it might require fine-tuning for some of those narrow use cases to really get past the threshold. We will see. My hope is calibration gets that. Calibration plus a ca- cascade of models.

453
01:11:10.936 --> 01:11:17.036
Like, if it's super confident, then maybe it's right. And, and if it's in the middle, then you do the next bigger model, and you chain off from there.

454
01:11:17.416 --> 01:11:27.996
I don't really know how that's gonna go, but, um, yeah, I, I, I could imagine it. And something that I could imagine too is, like, imagine you have, like, a, a series of, like, we own the en- entire Pareto frontier.

455
01:11:28.916 --> 01:11:38.136
Something that a business might want to do or, uh, I, I think a hacker would be okay with dealing with a Pareto frontier of models. Maybe a business wants something more dynamic.

456
01:11:38.536 --> 01:11:47.116
You could imagine, like, having, like, a s- different sizes of models and to dynamically pick which model based on how smart it is on different parts of your stack.

457
01:11:47.436 --> 01:11:57.256
And you could even imagine, because of how simple our thing is, you could imagine, like, some automatic fine-tuning on that. Yeah. Uh, not the promise in the slightest. I'm just, like, cooking on sci-fi.

458
01:11:57.556 --> 01:12:06.636
But you would consider different sizes of dev models, so to offer that gradient? Absolutely. Yeah, yeah, yeah. Yeah. Like, we, we... Like, how would I know how much intelligence people need? I don't know. Right?

459
01:12:06.756 --> 01:12:17.496
Uh, yeah, I don't know either. [laughs] Demand is, demand is, uh, unlimited. [laughs] Well, yeah, people are telling us not to ship things right now because we don't need to ship things because again- It's good enough.

460
01:12:17.756 --> 01:12:29.256
Yeah. Yeah, but that's kinda lame, and, uh, I really like the saying, "This is something that I hope people hold me to," because it'll be hard to, to- To come back... to walk back from. Yeah. Yeah. Like, the...

461
01:12:29.336 --> 01:12:39.636
I, I don't know if it exactly the saying that culture is what you do when the market doesn't reward it, and I really like that because I think that we are standing for something.

462
01:12:39.916 --> 01:12:53.740
Maybe in the future- What we're standing for is, like, so obvious that we're the equivalent of, like, boring, like, Visa or something like that, and, like, we're just, like, a utility that no one really thinks about, and I'll be wearing non-pink suits or whatever else.

463
01:12:53.760 --> 01:12:59.320
[laughs] But I really want to be, like, rallying the world to this, you know?

464
01:12:59.620 --> 01:13:10.240
Like, I want to keep doing cool stuff bec- not because we need to, but because I want, like, people to realize that this is just the beginning, you know? Like, that wasn't even meant to be the opening salvo.

465
01:13:10.280 --> 01:13:18.400
That was, like, kind of like a, a, you know, low-key research preview or whatever you wanna call it. Yeah. And there, there's a lot more we can do. Yeah. We've-- With...

466
01:13:18.440 --> 01:13:20.980
Like, machine native intelligence is gonna go wild.

467
01:13:21.260 --> 01:13:29.460
So not the only s- potentially not the only size, potentially not the only model that you guys launch, uh, you know, that you, you want to open people's minds- Absolutely not for any of those. Yeah.

468
01:13:29.640 --> 01:13:40.900
I want, I want to, like, meet whatever needs we can. Yeah. Right? Uh, like, uh, at-- But with, like, a giant caveat, I don't want to be like OpenAI's product teams that, like, like, throw stuff at the walls.

469
01:13:41.320 --> 01:13:51.080
Like, I want it to be, like, in a, under a unified vision. Like, if you go back to the manifesto, like, everything needs to be under one of these three, three things- Yeah... in my opinion. Um- Uh, I'm not- Yeah...

470
01:13:51.160 --> 01:13:58.700
I'm not prepared to do this, uh- Oh, I'm sorry. [laughs] I'm sorry for asking. I, I can just, uh, just talk about it. Like, we have, like, three steps in our stuff. Yes. It sounds like a tease.

471
01:13:59.000 --> 01:14:06.820
I want everything to go under one of these three things- Good, good... to keep pushing the boundaries and everything. Like, like, this is not-- these are not, like, checklists.

472
01:14:07.220 --> 01:14:11.460
These are, like, axes that we think build, like, the foundation of,

473
01:14:12.820 --> 01:14:22.840
you know, of, of, like, a new te-te-technological revolution, and I want all of the, all the bets we make to be somewhere in there, and we will be doing some weird, weird stuff model-wise.

474
01:14:22.840 --> 01:14:32.320
[laughs] So, um, because machine native, right? Like, uh, humans don't need to totally get it. It needs to just be valuable. Uh, you know, th- w-with, uh... Just give people a tease or hints.

475
01:14:32.380 --> 01:14:45.500
Like, what, what does weird look like? What is weird? I, I'll give people a hint. Yeah. Um, some people are trying to call them decision models. Okay. Um, th-the-- our primitives are decisions. Um,

476
01:14:46.740 --> 01:14:56.640
I, I wouldn't do that, uh, because I think there's other types that are machine native that are not decisions. Okay. We'll leave it at that- Is that a fun-... and let people guess... a fun hint? Yeah, yeah. Yeah, yeah.

477
01:14:56.700 --> 01:15:03.080
I think it's a, I think it's a pretty fun hint. Yeah, yeah. There's people... Look, there's, there's people saying, like, "I've done this before. I made a decision model a year ago."

478
01:15:03.120 --> 01:15:09.020
Like, "Geoff is not new and Geoff's not cool." Yeah. But, like, I, I think, uh, there's the categorical, like, here's what you're establishing is possible.

479
01:15:09.340 --> 01:15:19.160
There's the, uh, performance of, like, well, actually the-- for the benchmarks and the numbers that you're getting, you are still beating ev-- as far as I can tell, you're still beating every single clone of you out there.

480
01:15:19.320 --> 01:15:26.880
I, I don't care about the benchmarks- Exactly... just to be clear. Exactly. So, like, uh, like, even if we were w-winning or losing, I want to denounce them. You've established the category, right? Yep. Yeah. Yep, yep.

481
01:15:26.960 --> 01:15:37.180
Uh, but al- but also I think this, this nuance between decision models and system one, uh, I think is actually the thing that you're trying to- Yes, and I just wanna make software engineers super powered. Yeah. Right?

482
01:15:37.260 --> 01:15:52.640
Uh, l-l-like, with AI. Like, and, or, or, like, the tragic thing to me is, you know, in that AI winter direction, I think, like, it's, it's, it is so sad that AI was so powerful yet so underutilized. Like, um,

483
01:15:57.060 --> 01:16:09.660
um... It's a thing that gets me emotional, but, um, man. Like, I think that that is-- I, I don't want to, like, just be, like, pure techno-optimist, like all technology is good.

484
01:16:10.200 --> 01:16:20.740
I think what was happening now was, like, a travesty. Like, it's... A-and, like, there's... You know, I just want to, like, open up those possibilities for people. [laughs] Yeah, I'll stand it there.

485
01:16:20.940 --> 01:16:27.320
I, I, I, I, I've, I've cried too much these last few days- [laughs]... to, to, to want to do it on the record. Yeah, yeah, yeah. No, it, it- Yeah.

486
01:16:27.360 --> 01:16:33.420
I appreciate you sharing a little bit of that, and I, I think people can see that you're very authentic and- Yeah... passionate about this. Um, th...

487
01:16:33.600 --> 01:16:43.580
You know, you don't necessarily get that from a name like Typesafe AI, but, like, uh, I think once people immerse themself, themselves enough in, like, here's the genuinely different direction you want the world to go- Mm-hmm...

488
01:16:43.660 --> 01:16:51.480
and, like, actually you have done, like, the, the hard part about going from zero to one on the, on the thing, then, like, like, now let's all go to-go together in, like, the, the new direction, right? Yeah.

489
01:16:51.540 --> 01:17:05.380
Yeah, yeah, yeah. Yeah, yeah. But I, I don't kn- I am sure that I won't think... Maybe I will think that the hard part was done, perhaps. I think that there's going to be many more hard parts. Um, like, if, you know,

490
01:17:06.600 --> 01:17:16.180
all sorts of stuff gets automated and we finally see GDP growth and, like, you know, it's like, you know- Yeah... a jazz party every day, then maybe the hard part is done. But, like, I, I don't th-think so.

491
01:17:16.200 --> 01:17:27.980
And, like, I really, really think that people focus too much on speed and cost and not enough on reliability. Okay. Like, reliability is what makes it delightful. Like, reliability is what, like, allows you to trust it.

492
01:17:28.060 --> 01:17:38.280
You have this line, uh- [laughs] Yeah... TFP growth rating three percent in five years. Hell yeah. I've never seen- Hell yeah. Let's fucking go. [laughs] I've never seen- Yeah... a lab care about TF-TFP growth.

493
01:17:38.300 --> 01:17:47.160
But, like, that is what an economic revolution is, right? Like, it's actually extremely consistent with what the OpenAI charter used to stand for. You know, it was talking about, like...

494
01:17:47.600 --> 01:17:55.160
I think the charter is the same, but they've kind of tried to move definitions around to, like, you know, hundred billion in profit or something like that. Not that I hate on OpenAI. It, it wasn't, like, a...

495
01:17:55.240 --> 01:17:59.200
Yeah, it wasn't a well-defined term what AGI is, right? They tried to do it. Yeah. Yeah. Right?

496
01:17:59.220 --> 01:18:12.800
Like, doing majority of the world's economically valuable work, and they should have to answer the question: How can it do millennium price problems in math and zero of the world's economically valuable work, like rounding error?

497
01:18:12.820 --> 01:18:21.420
You know, like, I think that all models are roughly tied right now at zero. There's some chance that, like, we have started already. But, like, I would guess that it's not yet one percent.

498
01:18:21.960 --> 01:18:29.800
Um, and I think that that will show up in... Like, when it does happen, it will show up in the economic statistics. Mm-hmm. It's gonna be fucking awesome.

499
01:18:30.060 --> 01:18:37.060
It will not cause mass unemployment, but it will cause, like, a whole bunch of awesome shifts, and wor- the world will be a lot better.

500
01:18:37.360 --> 01:18:48.432
And also, like, I'm really tired of AI always being the foreground character, uh, of things. Like, I think that- The world should just be more delightful, and AI should just help with that. Mm.

501
01:18:48.552 --> 01:18:53.112
You know, and I- Just, like, disappear into the background. Exactly. Yeah. You know, like, like, uh, the-- I, I say this in my talks.

502
01:18:53.392 --> 01:19:02.072
Like, how can it be that twenty nineteen software, like software, SaaS, whatever, uh, super-duper valuable, right? It's twenty twenty-six now.

503
01:19:02.432 --> 01:19:12.092
How is the software basically exactly the same despite AI being so freaking awesome, other than sometimes having a chat box on the side, right?

504
01:19:12.332 --> 01:19:21.612
Uh, that, that, like, that kind of works but doesn't allow you to make decisions that the companies have stakes in because they, they can't be trusted to make decisions. That to me is nuts.

505
01:19:21.992 --> 01:19:31.312
You know, there, there's so much economic incentive for this, and I, I think it's going to be, uh, like a, like an inverse SaaSpocalypse. I think SaaS is going to be supercharged by this.

506
01:19:31.592 --> 01:19:38.792
They are the ones who are, like, most in the know of what things are valuable to automate, and it's gonna be, like, a crazy time. Yeah. I, I think, I think so too.

507
01:19:38.832 --> 01:19:41.692
It's a s- it's a beautiful thing that you've unlocked, you know? Yeah.

508
01:19:41.772 --> 01:19:52.012
Um, you mentioned one thing here, which I don't know if it's, uh, like directly here, which is what is a system one problem and what is not-- what, what is a system two problem? Like, you know. Fuck. Um, [laughs].

509
01:19:52.052 --> 01:19:59.792
That's a hard one. [laughs] That's a hard one, my friend. Um- 'Cause people now are just trying to Jev everything, right? [laughs] Which, like, probably- Jev everything... is gonna fail, right?

510
01:19:59.872 --> 01:20:08.852
[laughs] Um, but, like, some things are gonna be good. J-Jev everything is pretty funny. Yeah. Uh, it's a pretty funny way of doing it, saying it. Um, the-- so I'll tell you the truth. Yeah.

511
01:20:09.152 --> 01:20:21.532
Um, the truth is that this is an empirical problem, just like scaling laws are, are an empirical thing. You know, like, why doesn't, like, robotics really work right now despite all the money being spent on it? Mm.

512
01:20:21.592 --> 01:20:27.032
I don't think it's about, like, spending more money necessarily. The, the empirical results m- just might not be there, right?

513
01:20:27.532 --> 01:20:39.512
Um, so empirically, I believe that these, like, pre-trained super condensations of intelligence are fundamentally system one thinkers. I think that they truly...

514
01:20:39.792 --> 01:20:50.512
L-like, system one is the closest thing to describe what LLMs are strong at. RLVR has done incredible things for system two thinking. I am at awe. It is super freaking cool.

515
01:20:50.772 --> 01:20:57.812
Like, I don't think that it's going to result in AI doom in the slightest. Um, not, not zero percent, of course, 'cause I think zero percent is miscalibrated.

516
01:20:58.112 --> 01:21:04.892
But, like, it, it's, it's really cool what they've done, and they've really pushed it to the limits. Well, maybe they don't think so, not the limits limits.

517
01:21:05.232 --> 01:21:19.152
But, um, like, it is, it is a weird thing for models to do, and they are very fragile at this. Like, think about how people used to talk about AI back in the ChatGPTJs. Like, wow, it's really general.

518
01:21:19.472 --> 01:21:32.412
It can do a lot of general things. And, and, and then, but it's bad at math problems and, like, GSM8 grade school math. Um, and then now look at how people talk about RLVR. It's so fragile. It's so jagged. You know?

519
01:21:32.452 --> 01:21:38.452
Like, it can-- why can it do this, like, really weird thing? And actually, you know, math is not just spiky, it's fractal, right?

520
01:21:38.972 --> 01:21:51.192
This is because RLVR is, you know, like, if we talk about, like, what is the North Star for each thing, RLHF is please humans, right? That is what the human feedback is. RLVR is optimize benchmarks.

521
01:21:51.592 --> 01:22:00.472
You know, that l- everything that goes into the RLVR category literally is a benchmark by definition 'cause a benchmark is programmatically verifiable, simple outputs that can, like, do well.

522
01:22:00.932 --> 01:22:15.292
And, you know, RLCD is make it reliable for, you know, programmatic use. And, um, yeah. That, that, I'll- Yeah... yeah. This, this, uh, maybe I'll, I'll offer some thoughts, and then you can sort of, um- Mm-hmm...

523
01:22:15.312 --> 01:22:24.112
correct me if I'm wrong. Um, one, for example, one thing that I've been thinking about is also... Uh, so I, I threw Jev at a bunch of things when you gave me access- Ooh. Yeah, yeah... uh, on day one.

524
01:22:24.572 --> 01:22:32.992
Um, and, uh, multi-hop reasoning, right? Like- Yep... uh, s-so single hop, fantastic. Mm-hmm. Like, state, state of the art. Uh, you should never use anything other than Jev for single hop. Yep, yep, yep.

525
01:22:33.032 --> 01:22:41.032
Multi-hop, it's gonna start to falls down. Yep. And it's, like, kind of mon- monotonically increasing as you increase the hops. Yep, yep, yep. Right? So, oh, yes, back to that empirical question.

526
01:22:41.352 --> 01:22:50.312
It depends on what we can, like, pull out of the models. Yeah. Right? So we want everything. Like, we want to unearth as much intelligence as possible, period.

527
01:22:50.812 --> 01:23:00.112
Um, the models-- like, I see us as, like, unlocking and smoothing and sculpting the intelligence while, like, adding new capabilities and, like, you know, like, like filling in gaps in it.

528
01:23:00.592 --> 01:23:10.912
Um, and we will be filling in, like, you know, more and more and more and more of these gaps over time. But the reality is that we are in the business of unearthing properties.

529
01:23:11.292 --> 01:23:21.512
Those properties are actually a function of what is available from, like, these, like, you know, these conden- condensed cores and, like, Frankensteining them all together to have all of the properties of everything.

530
01:23:21.692 --> 01:23:30.332
Yeah. You know? Um, but the reality is we are in the business of unearthing as many capabilities possi- as possible, and system one just happens to be the description of what works.

531
01:23:30.652 --> 01:23:43.392
And everything that works in that paradigm will be system one-ish. You know, like, I, I am... Like, like there is a reason why we don't do what's called latent reasoning, reason-reasoning in strings. Yeah.

532
01:23:43.452 --> 01:23:51.032
I think the reasoning... Like, the, what models do really well is reasoning within the models. It's not totally complete. It doesn't do great at all. Wait, la-latent reasoning is reasoning in strings?

533
01:23:51.072 --> 01:23:58.672
I thought latent reasoning is reasoning in, in, inside the model weights. The, the re- That I think that's- I don't know... people used to call that- I just wanted to clarify... continuous reasoning. Okay.

534
01:23:58.692 --> 01:24:07.572
I'm not entirely sure. Okay. It was called latent reasoning because, like, it used to be that the reasoning traces were secret. Mm-hmm. So they're kind of like a latent variable for the answer. Ha. Yeah.

535
01:24:07.812 --> 01:24:17.012
So what's secret is now shifted. They, well- Yeah, yeah... it, it's still secret for OpenAI Anthropic, right? So no reasoning Jev- Yes... as far as you will ever do it, right?

536
01:24:17.072 --> 01:24:28.232
Because that, that, like, violates the whole promise of system one. I-- my promise is to do whatever necessary- [laughs]... for machine native stuff. Yeah.

537
01:24:28.432 --> 01:24:38.232
I, I could imagine there are, like, there are some forms of reasoning that are less slow, inefficient, and fragile that I, that are, like, totally on the cards, just to be clear.

538
01:24:38.532 --> 01:24:48.152
So pragmatic person, I'm not making promises on, like, you know, like, methods. I'm making promises on, like, the, what my ROI North Star is, and I'm going to fight for that.

539
01:24:48.212 --> 01:24:54.592
Like, like, you know, like, like this launch didn't happen, and we are still, like, hungry for our place in the world. That's great. Yeah. Yeah.

540
01:24:55.012 --> 01:25:03.972
Um, I, I think the other thing that, uh, vision is another one that it's like a, a big, like, you know, capa-capability that you don't have. But maybe it doesn't ever belong in sy-system one.

541
01:25:04.052 --> 01:25:10.520
I, I think I have a pretty good vision. Uh, what? Sorry? I think I have a good vision. No, no, no. Sorry- I am kidding. I'm kidding. [laughs] Oh my God. Yeah, yeah, yeah.

542
01:25:11.300 --> 01:25:22.240
Um- Because people obviously, the first thing they want is vision because of the Doom demo, but also just, like, everything, you know, o- other than text, is vision. Everything is in the cards- Yeah... in my mind. Okay.

543
01:25:22.360 --> 01:25:29.620
Like, like, um, a- and actually, this is, like, a debate we have. This... Man, your audience is probably, like, the great one to have in this debate.

544
01:25:29.960 --> 01:25:42.960
There's a question about, like, how much do we try to, like, give people what they think they want, which is what we did in stealth for two years, we just knew that this is obviously going to be valuable, versus give them what they say they want, right?

545
01:25:43.240 --> 01:25:51.440
A- and, like, uh, there's a lot of dimensions of this, right? And y- you know, like, context length is an example of this, right?

546
01:25:51.780 --> 01:26:01.480
Um, every single model, including ours, I actually think as far as I can tell, ours is, like, by far the best- The longest context. Yeah... at not degrading in long context. Yep.

547
01:26:01.520 --> 01:26:09.920
Um, but, like, the other providers are just like, "Whatever people wanted, let's just give them the stupid thing." And, um, like, we need to figure out a balance for this because- Yeah...

548
01:26:09.940 --> 01:26:19.520
you know, like, if you take the, the, the former side too far, give people what they want, you end up with, like, anthropic nanny state-style thinking, which is very, like, anti-developer.

549
01:26:20.120 --> 01:26:28.600
While, like, the pro-developer route would be like, give them what they want, but developers are... Like, we don't want to put the burden on them to figure out the je ne sais quoi of intelligence.

550
01:26:28.960 --> 01:26:41.680
So we are trying to, like, figure out this navigation of, like, how quickly to release things, uh, to still, like, have our, like, brand of trust, and also, like, teach our... treat our users like adults- Mm-hmm...

551
01:26:41.700 --> 01:26:50.000
that can make informed decisions that, uh, don't need, like, nanny stating on top of this stuff. Yeah. I think that's fair. Uh, and- Yeah. And, and we don't know the answer, to be honest.

552
01:26:50.340 --> 01:27:00.380
Like, uh, we'll, we'll have to figure it out. It's gonna be... That's probably going to be, like, one of my biggest debates over the next couple of days. Yeah. Because, like, we have a lot of stuff.

553
01:27:00.740 --> 01:27:07.900
Again, we didn't expect it to pop off, so we were like, "We need some follow-up launches." So yeah. Yeah. I don't know, I don't know if you didn't expect it to pop off. Like, I... You put...

554
01:27:08.100 --> 01:27:16.160
I, I saw the work that you put in. Like, I have never seen you lock in so hard as- [laughs] It was like the last two months, basically, right? Like- Well, well- Like-... that's also because- [laughs]...

555
01:27:16.180 --> 01:27:26.260
my chief of staff made me lock in. [laughs] Yeah. So- Like, it's like, uh, I have never... I thought I worked hard before. Yeah. And- No, but, like- Yeah...

556
01:27:26.280 --> 01:27:35.020
you were showing up at our writing workshops, and I was like, "What are you doing here?" And, and, like, oh, you- It was useful. It was great... you, you clearly, like, were very intentional about your launch. Yep.

557
01:27:35.060 --> 01:27:43.740
And the work showed, and like- Yeah... congrats, like, you know- Thank you. Thank you... all the kudos. Like- Um, I hope to keep locking in- Yeah... is my, is my sense. Yeah. I, I want to...

558
01:27:45.120 --> 01:27:53.280
Like w- like, I think that we've passed many great filters for the, the, the tech world, what we're wanting, but, like, there's still gonna be a bunch more. Yeah.

559
01:27:53.360 --> 01:28:01.560
And, like, holy smokes, am I excited to fight the good fight. Yeah. It's exciting. Um, before we broaden out to, uh, uh, topics outside of TypeSafe- Ooh...

560
01:28:01.620 --> 01:28:12.840
I just wanted to offer, uh, any other things that you think, like, underrated or misunderstood about what you have launched. Underrated or misunderstood? Yeah. You have pan outs...

561
01:28:13.020 --> 01:28:27.280
Uh, sorry, patterns here, uh, maybe, maybe you wanna go into that, um, model jaggedness, any- anything. Yeah, you know. Give me one- Yeah... noodling of it. Oh, man. I, I would rant about all of these.

562
01:28:27.780 --> 01:28:36.020
I, I really shouldn't. I really shouldn't. Okay. Um- And, you know, people can come go to your Discord if they want. Yeah. Uh, uh, people put a lot of love into the cookbooks- Yeah... is what I will say.

563
01:28:36.360 --> 01:28:45.940
The cookbooks have, like, some fire stuff. We had considered putting a bunch of these things, like, in the main launch blog post, but it got kind of long and unwieldy and, like, very power usery.

564
01:28:46.300 --> 01:28:59.440
But, like, we really, really, um... I, I'll be frank, like, before the launch, um, every... Like, like what we're saying sounds, like, sounds like, like this weird alien tool, why would anyone need this?

565
01:28:59.860 --> 01:29:04.420
You know, it was a very weird thing. We were very worried about teaching people about, like, this new frontier.

566
01:29:04.840 --> 01:29:12.800
It obviously succeeded, but, like, we put a lot of work because we thought the education would be, like, a gigantic bottleneck for us. Um, I

567
01:29:14.100 --> 01:29:22.140
pr- it probably worries and it's no lo- probably no longer a problem because people are doing things, like, well beyond what you could ever expect. They'll show you what- Yeah... how to use your model. But... Exactly.

568
01:29:22.179 --> 01:29:25.260
But, like, they, yeah, and their use cases are, like, kinda cooler than ours.

569
01:29:25.299 --> 01:29:34.900
Like- [laughs] Like, there's a bunch of stuff where I'm like, "Man, if that was our demo, holy shit, that is way cooler than what we were showing." Um, like the, the computer use stuff, holy smokes, is it cool.

570
01:29:35.320 --> 01:29:46.560
Um, but like, like, we put a lot of love into this. This is not like AI-generated trash, as far as I know. Yeah. We put a lot, like, it's, like, a lot of love in here. Yeah. Fair enough. And, like, each of these are...

571
01:29:46.860 --> 01:29:59.600
Like, like, there's real alpha there. Okay. Like, these are inspired by solving real customer problems that existed, and we went through the work of, like, helping them do cool-ass stuff. Yeah. How much, uh...

572
01:29:59.820 --> 01:30:08.480
While you're talking about this, right, how much validation did you do before launch? Like, what, you know, what was that process like? What was that process like?

573
01:30:08.780 --> 01:30:14.840
Like, clearly you did some, but obviously, you're not getting in touch with as many people as you are today. Yes, of course.

574
01:30:15.120 --> 01:30:26.300
But, like- Um, I actually think that the reception was pretty bad, and, like, actually for the non-technical people in the team, they were really worried. Yeah. You know, like, there was a lot of fear.

575
01:30:26.320 --> 01:30:33.160
It's like no one really gets this. And, like, you know, they don't want it. We're, like, selling, like, a vitamin and not, like, a painkiller.

576
01:30:33.460 --> 01:30:44.600
Like, should we have FDEs to, like, write the software around solving that problem? Yeah. We had almost no revenue before launch. Um, it was kind of like, uh, like we...

577
01:30:44.900 --> 01:30:54.600
Like, the technical people were, like, obviously true believers, right? Like, we knew that this was sick. Its pro- computational properties are, like, off the charts on, like- [laughs]...

578
01:30:54.720 --> 01:31:02.420
so many axes that we're like, "Yeah, obviously it's gonna be huge." I was definitely super afraid, which is why I locked in super hard.

579
01:31:02.820 --> 01:31:10.580
But, like, the most common thing was the, like, I would say, like, more than half the people we had play with it just did not get it.

580
01:31:11.120 --> 01:31:17.960
And, like, the people who did, uh, like, were like, "Man, this is really cool, but how do we get this through procurement and stuff like that?"

581
01:31:18.320 --> 01:31:31.220
You know, it was like, it was, like, quite a, uh, quite a battle, and we just knew, like, okay- Our target market is gonna be developers. People will find the use cases, and that way everyone is gonna FOMO in and, like,

582
01:31:32.240 --> 01:31:41.640
I, I don't want to rub in people, like, changing their minds with the facts changing. Mm-hmm. I do want to p- call into question, like, the concept of product market fit, you know?

583
01:31:41.980 --> 01:31:50.740
Yeah, but, like, because, like, there was a product, there was a market. Like, we were like, "Hey, do you want to use this?" And people are like, "I don't know, really know if it solves our problems."

584
01:31:51.360 --> 01:31:59.320
It explodes and everyone's like, "We need as much rate limits as we can. Can we literally give you GPUs because we are constrained right now?" Yeah.

585
01:31:59.380 --> 01:32:04.480
So of course, uh, you know, like marketing is an element of it- [laughs]... of course, but I don't even think it's about marketing.

586
01:32:04.860 --> 01:32:14.900
I think it's about, like, passionate developers who've, like, you know, our souls basically resonated at the same frequently, and that frequency, and that got, got everyone else excited too. Yeah.

587
01:32:14.980 --> 01:32:23.220
And I'm hoping as well that, like, we as a company will be eternally, eternally, eternally grateful to those developers.

588
01:32:23.440 --> 01:32:31.860
Like, not, and not just like, you know, like, um, the, the companies that, like, are, like, start off with developers and, like, go to enterprises. Big oil market, yeah. Exactly. Yeah.

589
01:32:31.880 --> 01:32:41.620
And like, I'm, like, even thinking about, like, how can we launch things that are better for... Oh, man, I don't know if I should say this- [laughs]... but I will. Better for developers than enterprises. Exactly. Okay.

590
01:32:41.640 --> 01:32:55.180
How do we do that? Like, how do we empower them? Um, and I have cooks, I have cooks. But, um, it's a, it's a very weird thing to do and, like, I don't know how else I can show my thanks and loyalty to that.

591
01:32:55.560 --> 01:33:00.220
You know, like, a- and, and that's why I did, like, the, the dyeing my hair yesterday. It was like- [laughs]...

592
01:33:00.260 --> 01:33:10.160
it's like I wanted to talk to them 'cause it felt dirty to me during our company's, like, most important times not to keep talking to them. Good. Uh, well, I mean, that's why- one of the reasons- Yeah... you're here.

593
01:33:10.160 --> 01:33:18.320
Hold me to that, please. Yeah, yeah. We will, we will. I, I try to be principled. Yeah. Quote me on this. [laughs] Uh, call me out. You know, have the pitchforks out if I change.

594
01:33:18.760 --> 01:33:28.640
Uh, I was just gonna briefly show the computer use stuff. Whoa. Is this, is this what you're referencing? I've never seen... I haven't... I- I've seen... [inhales] I saw, like, a airline browser use thing.

595
01:33:29.720 --> 01:33:41.960
And, um, inside this new note, let's make the title say hello. Wow. Great. Great. Okay. Um, let's move on, and can you open the Arc Browser? And once you're there, can you Google search Norbert Wiener?

596
01:33:43.800 --> 01:33:55.240
Um, now can you open up x.com? Is this this kind of use case? Oh, the voice use cases. This is actually the first one I've seen. This is- Oh, okay... open up the photo booth. Wow. [laughs] Oh, wait, wait, wait, wait.

597
01:33:55.260 --> 01:34:05.400
Hold on. Can you pause- can you go back a second? Can you go back a second? Um- Rumors claim Anthropic engineers wor- worship Claude as god. Wow. [laughs] Wow. Dang. That's pretty funny.

598
01:34:05.820 --> 01:34:18.120
Um- And here you are building Prod. [laughs] Abso- wow, this is sick. Uh, yeah, so, so c- clearly you can operate the whole computer with voice, with Jev as a decision model. So

599
01:34:19.180 --> 01:34:29.780
just like I'm anti-benchmaxing, I'm also anti-demos. I want to make sure that it works reliably. I love people are playing with it. Mm-hmm. This is super fucking sick, have no doubt. I want to see this.

600
01:34:30.160 --> 01:34:39.060
I wanna see it be used. I want our team to play with it. I wanna find the weaknesses, and I wanna solve that. Yeah. And I would lo... Man, that looked really cool. [laughs] That looked really cool. I want that.

601
01:34:39.380 --> 01:34:47.520
Uh- I want that. Like, when my, when my wrists are sore, I, like, just whip- Whisper flow everything. That'd be sick. Well, as, uh, well, you know, just, just to round out the use cases side- Yeah...

602
01:34:47.600 --> 01:34:55.640
'cause I, I do have to let you go. Um- [laughs] Uh, who's, uh, who are the, who are the bigger companies that have reached out and have surprised you with what they want to do?

603
01:34:56.820 --> 01:35:05.260
Just, you know- I am so out of touch for that. Okay. People have shown me screenshots of companies, and from what I've seen, it's all of them. Um- Yeah.

604
01:35:05.300 --> 01:35:13.400
Mo- mostly, like, you know, for, for those people who work at larger companies and they're not doing this kind of work, uh, I just wanna give people examples of, like, you should go look that up, look that up, look that up.

605
01:35:13.600 --> 01:35:25.620
Oh. So, uh, like, I think demos are super-duper sick. Obviously, the coding agents are, like, gigantic use cases. Mm-hmm. Like, they are, like, also super sick. Oh, Cog is all about Jev right now. Oh, hell yeah.

606
01:35:25.620 --> 01:35:33.540
[laughs] Can I, uh, can I give a little bit of a, a tangent about coding agents, if that's okay? I- Yes, please... uh, uh, let me- We love coding agents here... give me a second. Give me a second.

607
01:35:33.580 --> 01:35:41.600
Okay, actually, r- I'll come back to coding agents. Let me describe, like, the big families of use cases. Yes. Like, we've mapped this out from first principles, like, long before release.

608
01:35:42.019 --> 01:35:50.700
They are what we call dark data. Like, people hoarded big data, but they would not throw a LLM at it 'cause it was too expensive. So large companies l- adore this.

609
01:35:50.960 --> 01:35:58.240
They have, like, piles of data that they wish they could analyze, and this is, like, a data scientist's wet dream. So this is, like, this is a giant one.

610
01:35:58.500 --> 01:36:08.780
Like, I think this plus, um, coding agents are the m- big money makers because that's what the all-- where all the volume is, right? Um, there's the real-time stuff, you know.

611
01:36:08.820 --> 01:36:23.260
Like, people who need, like, intelligence in the loop. They-- Like, I would guess that every CEO, if not CTO at those companies, um, knows how much better their product gets with every, like, ten milliseconds shaved.

612
01:36:23.460 --> 01:36:32.320
Yes. Uh, and, like, to- Especially e-commerce, yeah. Yeah. Oh, o- or, like, assistant-y things, you know. There's many AI assistant-y things and, like, as far as I can tell, they really love it.

613
01:36:32.740 --> 01:36:40.740
Um, again, I'm not in the front lines of customers right now, so I just get-- know what my team tells me. But, like, this, I'm so excited for this. I'm really excited for this for games.

614
01:36:41.100 --> 01:36:49.900
I really wanna play, like, sick-ass auto battlers, where you're, like, commanding your team or, like, semi-auto battlers. I, I think that'd be so cool, but don't make it too good while I still have a job.

615
01:36:50.340 --> 01:37:00.520
Um, [laughs] and, um, the, you know, like, there's the, uh, the, the, what we call, like, verify everything. You know, like, verifying all LLM calls, kind of like observability.

616
01:37:00.880 --> 01:37:09.580
Um, I think, uh, actually, on the note of docs, what people should be doing is, like, the parallel questions are very cheap. So if you have, like, big states you wanna ask many questions on- This right here? Yeah...

617
01:37:09.620 --> 01:37:16.220
put IDs on every, like, message and then ask a question about each ID. So, like, when you have, like, a long state. Uh-huh.

618
01:37:16.400 --> 01:37:21.540
So that way you can, like, pay for the state once and ask lots and lots of questions about each message within it.

619
01:37:21.880 --> 01:37:36.180
I think that is, like, a, like, a great way that, like, saves money and is- Which, by the way, I always think, like, it's interesting framing system one, system two, because it, it basically makes the case that you should always make one or ten or a hundred Jev calls for every one reasoning call that you make

620
01:37:37.240 --> 01:37:42.480
Well, uh, maybe. Right. Well, I, I mean, I don't-- I would like people to spend less. Yeah.

621
01:37:42.600 --> 01:37:53.020
You know, maybe you do like, you know, one half the reasoning calls and, like, ten Jev calls each or something like that, or whatever solves the problem that, like, couldn't have existed otherwise.

622
01:37:53.480 --> 01:38:04.860
Um, okay, uh, wait, uh, number four use case was, uh, what I describe as, like, smart software. Like, software that's intrinsically composable and, like, does, like, weird, fun stuff that could never happen before.

623
01:38:05.260 --> 01:38:09.720
You know, like the programming language as Jev thing. I don't know if you've seen that. That is so cool.

624
01:38:10.100 --> 01:38:16.260
Man, if we knew how to give out credits, because, like, we're really early in our infra days, I would wanna give all these projects credits. Mm.

625
01:38:16.580 --> 01:38:29.460
Um, and I, I think that those are, like, how we've mapped out, like, the main use cases. Um, computer use has also come in kind of like the real-time direction as well, and, like, that's really, really cool.

626
01:38:29.860 --> 01:38:40.780
If it is reliable, I am super jazzed about that. I suspect we can make the model a lot better at these use cases, 'cause, like, that came out of left field a little bit, so that's, that's really cool. Um,

627
01:38:41.699 --> 01:38:54.360
um, um, on the coding agent thing, and this is, like, a really surprising thing that is happening right now. Okay. Um, Claude Code and Codex are, I believe, the, the winner, like, the, the number one and two.

628
01:38:54.420 --> 01:39:04.180
I'm, I'm not entirely sure. I don't follow closely- Roughly... but, like, it's roughly that. But they're built around a single model world. You know, like, and, and that makes a lot of sense for them, right?

629
01:39:04.520 --> 01:39:09.740
Because, like, it has been a one-model game where it's, like, kind of like the same model but different intelligence that you're shopping. Mm-hmm.

630
01:39:09.800 --> 01:39:19.100
But all the open coding agents are, like, fucking jazzed right now because they're, like, getting their Jev on and- [laughs]... like, the thing is there's-- I'm sure they're trying a lot of weird stuff. Mm-hmm.

631
01:39:19.680 --> 01:39:26.780
But all the coding agents are kind of roughly at, like, approximate parity, right? Because, like, there's not so much you can do with a while loop.

632
01:39:27.220 --> 01:39:37.740
But the moment one person finds one killer use case that, you know, you can only do with that coding agent, everyone will flock to it because they have, like, a monopoly on that thing.

633
01:39:38.200 --> 01:39:48.600
But all the open coding agents will be able to copy that, right? The end-- But I don't know what the Claude Codes and Codexes will do because they are built around that one model world. Single model, yeah.

634
01:39:48.640 --> 01:39:56.980
And, like, I think that's gonna be, like, a really interesting thing. You know, like, I would love to be able to integrate with them personally. Like, I want to integrate with everyone.

635
01:39:57.260 --> 01:40:07.580
Like, uh, they, they might make competitors eventually, I don't know. But, like, it is not me, my job as Sonfire Infrastructure to be opinionated on that, right? Like, I want to just serve the world.

636
01:40:08.120 --> 01:40:16.080
Um, but I, I don't know if they would do that. And, like, yeah, I think it'll make the coding agent game super weird. You know, like, I'm so excited for that.

637
01:40:16.400 --> 01:40:27.920
Uh, and, like, I'm sure, uh, I'm getting my team to review right now an internal document I made on design patterns I suspect will be useful for coding agents, so hopefully I can share it, like, right after I walk home.

638
01:40:28.440 --> 01:40:40.560
Um, uh, but, like, I think that there's just, like, such ripe area for exploration out in the world and, like, it's, it's-- Man, if I did not have this, I would love to experiment with coding agents right now.

639
01:40:41.200 --> 01:40:45.380
Yeah, I mean, and, and I'm sure the coding agent companies would love to work with you as well to, to figure that out.

640
01:40:45.740 --> 01:40:51.980
Um, yeah, I, I do think that there's still use cases for Claude Code and Codex with you guys- Of course... which, um, uh, it's, it's easy to explore there.

641
01:40:52.460 --> 01:41:00.620
Um, okay, I mean, you know, we've, you've, you've been very, um, uh, obliging and sort of indulging in all these, all these things. I just wanna take you out of TypeSafe- Hell yeah...

642
01:41:00.640 --> 01:41:11.500
just generally about-- And you've, you've made very clear your position on, on the state of AI. Uh, give you m-more room on the alignment safety side of things. Oh, uh, did I not talk about safety alignment at all?

643
01:41:11.860 --> 01:41:20.100
Oh, you did. You did. I think I didn't. I think that maybe I didn't. You did. Oh. I, I just like, you know, I, I think that there's, there's a lot of, uh, you have a lot of researcher discussions.

644
01:41:20.160 --> 01:41:23.380
We have this every- Of course... every NeurIPS. Yeah. What are people talking about?

645
01:41:23.500 --> 01:41:34.800
You know, like, uh, so for example, um, I, uh, recently was at, uh, one of these researcher gatherings, and people are genuinely worried about the pacing, right?

646
01:41:34.840 --> 01:41:45.140
Like this, this whole topic about, like, we should slow down because- Mm-hmm... uh, the public is, like, clearly not ready. Um, and I, I'm sure you have strong feelings. Um,

647
01:41:47.640 --> 01:41:58.779
I feel like this is the kind of thing that is a dangerous topic- Okay... to talk about. Um, I'm happy to talk about it. I, I live for danger. [both laughing] All right. Our company brand is chaos. It's not Jev.

648
01:41:59.000 --> 01:42:08.660
It is irreverence and chaos. And you know, yeah, and, like, you were at OpenAI during, like, the, one of the very first, like, very visible incidences, the blip, right?

649
01:42:08.700 --> 01:42:20.600
Like- Oh, the- Which, like, and, like, you-- the dominoes have gone down now- Mm-hmm... to now every frontier lab has co-signed a document saying that they want to pace. Um, interesting. I--

650
01:42:21.780 --> 01:42:31.760
So comp-- It's a very complicated, nuanced thing. I actually do want to write a response to this more formally. I do have, like, a little bit of a short version of my response- Yeah...

651
01:42:31.820 --> 01:42:45.400
which is that, um, as you RLVR more, like RLVR is like, so RLVR is not actually about verifiable rewards. Like, that has been failing since before the reasoning revolution.

652
01:42:45.600 --> 01:42:50.520
Like, like, like, and that's the weird part about tasks, right? Like back when... Oh, fun history.

653
01:42:50.800 --> 01:43:02.160
Back when RLHF was becoming a thring, thing, there were three different things that, like, are now called post-training, different efforts. And instruction following was by far the, like, the, the bastard child.

654
01:43:02.460 --> 01:43:09.740
Like, people didn't like it. They didn't want to take it into account. It was annoying. You know, like, I talked to the pre-training team, and I'm like, "Guys, this is the magic."

655
01:43:10.420 --> 01:43:16.460
And they're like, "We'd run so many model sweeps. You want us to wait for human evals to figure out which models to use?"

656
01:43:17.020 --> 01:43:29.800
And, like, everyone is like, you know, giving tons of, like, resources to, like, the code gen team, which, like, they did s- have some successes, but they were trying really hard to do RL on co, uh, like unit tests.

657
01:43:30.300 --> 01:43:38.880
And it didn't work, obviously, right? Like, you needed reasoning for that. So, uh, re-- So just to be clear, RLVR is not purely about the reward. It's about, like, the shape of everything too.

658
01:43:39.220 --> 01:43:43.480
And part of it is that reasoning is included in here, like this latent variable that you're doing things.

659
01:43:43.900 --> 01:43:52.380
And when you're doing things, you're just letting the models do whatever they want in order to make them be as powerful as you can to answer the hardest problems.

660
01:43:53.296 --> 01:44:03.056
And this whole pace the frontier discussion I think is, like, a very narrow focus because it assumes that everyone needs to do more RLVR, right?

661
01:44:03.336 --> 01:44:16.196
Which, um, like I obviously don't think I need to do more RLVR on our models. Mm-hmm. You know, I think zero is the optimal amount- [laughs]... for our shape. Hey, right? Like, come on. Yeah. Uh, you know? So, um,

662
01:44:17.596 --> 01:44:31.336
it, it's really, I think, a bit of a sleight of hand where they are saying that we actually want to keep doing the thing that looks dangerous because it does dangerous things.

663
01:44:31.636 --> 01:44:40.336
You know, like people say like, "Oh, maybe the sandboxing was a problem," or whatever else. Um, yeah, I mean, obviously it is, and they could've easily solved that, right?

664
01:44:40.636 --> 01:44:48.336
But they chose not to because the more things you let the models do in this do anything category, the more powerful it is, right?

665
01:44:48.656 --> 01:45:00.316
So, uh, like n- they're-- I think there's some, like, dissolution of responsibility there on, like, things that by design or non-design they're trying to make is just an assumption.

666
01:45:00.656 --> 01:45:23.056
You know, we must do RLVR and not just we must do it, we must do more and more and more, um, with giving the models, like, the power to do powerful, you know, do anything they want in the middle 'cause that teaches them to be powerful outside of it, and we don't want to limit those things well because it'll make it slightly less powerful on those things.

667
01:45:23.556 --> 01:45:31.696
Um, so like if you assume all of that, they're like, "Oh yeah- That's the logical conclusion... we're heading into a dangerous world, guys." Right.

668
01:45:31.716 --> 01:45:35.796
Like, everyone is gonna be doing this, and this, this is the only way to make AI sick.

669
01:45:36.376 --> 01:45:46.016
So, um- So basically it's like, it's like i- these are all internally consistent but actually starts from a premise that has alternatives if you- Of course... think about it. Of course.

670
01:45:46.236 --> 01:45:56.896
I, I think there's-- like, like, on the bitterest lesson direction, I think that there's very few people who've, like, made right tasks. You know, like new directions of AI. That is-- or new, new north stars.

671
01:45:57.276 --> 01:46:07.596
That is rare. Again, like, I think 2.2 times or something for LLMs itself, like RLHF and then RLCD. Oh. R- RLVR is, like, a.2 in my opinion, and I think that's generous.

672
01:46:08.136 --> 01:46:12.116
Um, but or a.5, or, like, it could be one whole one. I don't really care.

673
01:46:12.516 --> 01:46:26.296
Um, but I do think that people are thinking very closed-mindedly about this type of thing, and, um, this-- the, the only people who are at fault here are the researchers because it's definitely not the populace.

674
01:46:26.676 --> 01:46:36.736
You know, like, they just assume that OpenAI and Anthropic are just doing the best they can, and they are not the experts who are aware of the true optionality available. Yeah. Uh, and that's fair. And, and like- Yeah...

675
01:46:36.736 --> 01:46:45.196
you're, you're also doing your part in waking them up. Yes. Yeah, well, I, I'm doing my best. But, like, my goal is not, like, convince labs that there's, like, other directions- Yeah... to, to go down.

676
01:46:45.516 --> 01:46:53.736
My goal is have, you know, it's like spark hope in software engineers to start, like, actually automating things they've always wanted automated.

677
01:46:54.136 --> 01:47:04.216
Um, I had this, like, article that I wrote that my team didn't let me write, uh, they didn't let me publish, about, like, the, the, the future I want of AI. And, like, there's, like, a lot of, like, little things.

678
01:47:04.256 --> 01:47:09.956
Like, remember do what I mean? Imagine if everything could do what I mean, 'cause, like, that, that, all-- oh, that demo was do what I mean. Mm.

679
01:47:10.076 --> 01:47:14.796
Like, like you, like, like there's levels- Don't-- Yeah, don't do what I say. Hmm? Yeah. Don't do what I say, do what I mean. Yeah.

680
01:47:14.836 --> 01:47:21.316
And, like, we couldn't do what I mean yet because, like, computers are so basic and literal. But that computer use one was just that.

681
01:47:21.676 --> 01:47:29.256
And I think that there's, like, levels of smoothness that'll happen in the world that people just don't understand, and, like, the promise- [laughs]...

682
01:47:29.256 --> 01:47:40.296
of, like, smarts all around are-- it's, it, it's, it, it-- I don't wanna overpromise. Uh, I don't think it's going to happen right now, but, like, we are gonna do whatever the fuck we can to make that happen. Yeah.

683
01:47:40.816 --> 01:47:49.876
Um, any other things on the sort of general shape of post-training? You know, um, you obviously, you have been very intimately involved. Uh, mid-training, is that, uh, something that you do have comments on?

684
01:47:49.916 --> 01:48:03.156
I don't think we've ever talked about it. Mid-training. Um, I mean, it's all a spectrum. Yeah. Right? Like, am I- This is a curriculum, but, like, fancier. Yeah.

685
01:48:03.176 --> 01:48:09.716
[laughs] I mean, like, it's, it's, it's like, you know, it's a cost-saving thing. Yeah. You know, instead of, like, having to pre-train again. Like, there's intriguing stuff.

686
01:48:09.836 --> 01:48:17.836
I actually think that, like, intelligence has a je ne sais quoi at every single level, and it's always super-duper fascinating.

687
01:48:18.156 --> 01:48:23.716
Like, I am a shape rotator, so I don't like finding that, but I love it when people find it and teach me about it.

688
01:48:24.176 --> 01:48:34.696
Um, but, and, you know, looking at the data, this thing that, uh, our data team is so good at that I'm not, um- Hmm... uh, it's, I find it really, really fascinating.

689
01:48:34.776 --> 01:48:42.536
I love actually thinking about, like, how capabilities are, like, put into the model, like, over, like, the short term.

690
01:48:42.796 --> 01:48:49.376
You know, like, there's, like, the, the, the, the really rapid alignment of fine-tuning and over the long term after seeing it over and over and over again.

691
01:48:49.416 --> 01:48:54.036
Like, this stuff gets baked deeper and deeper and deeper and deeper into the model until it gets robust.

692
01:48:54.396 --> 01:49:02.476
You know, and that is, like, the, the north star to surface, and, like, the system-run stuff is the stuff that ends up getting robust. So I find mid-training to be, like, a fascinating thing.

693
01:49:02.876 --> 01:49:19.076
Um, I'm a fan of all forms of training. Uh, I'm a fan of all forms of, like, surfacing new types of intelligence. I wouldn't do it all myself because it's expensive. Um, and I have said privately and also...

694
01:49:20.696 --> 01:49:22.796
Should I say this? [laughs] Huh. Huh.

695
01:49:23.536 --> 01:49:33.516
You know, like, my philosophy is anything I sh- I, I should say in, like, private with, like, an investor, I should say in public with the people because that is, like- Power to the people... my thing. Yeah. Yes.

696
01:49:33.936 --> 01:49:43.096
So m- the thing I've said i- before is if you gave me a billion dollars, I wouldn't pre-train. Um, I still believe that to be true.

697
01:49:43.556 --> 01:49:51.996
It is a very expensive thing when if you are, like, like, if you're an AI engineer, you can, like, slice and dice and do all sorts of stuff.

698
01:49:52.056 --> 01:50:02.776
You know, like, Frankensteining is not the most elegant, beautiful thing, but it solves problems, baby. Hmm. So, um- Anything except pre-training Mm, mm. Yeah. Uh, amazing.

699
01:50:03.116 --> 01:50:18.576
Uh, I, I think one, one direction that I do think that is interesting, uh, just like synthesizing all your, all your commentary about these model things, is like, do we have a, a, a, a, a super model, uh, that has all these capabilities involved or do we break them out, uh, in, in fur- further, right?

700
01:50:18.676 --> 01:50:27.696
So like, um, uh, one way to put this is that OpenAI was trending in the direction of the omni model. Mm-hmm. Right? Um, 4o was, was one of those.

701
01:50:28.016 --> 01:50:35.416
Uh, then for, for, for a brief period of time, there was always like, there was like a kind of a main branch of like this is the chat tune model, and this is the coding tune model.

702
01:50:35.936 --> 01:50:43.516
Those are t- completely different things. Those are extremely different concepts. I will, like, break that down a little bit. So multimodality is a little bit different- Mm...

703
01:50:44.036 --> 01:50:48.056
um, because sometimes the other modalities help, sometimes they hurt. Yes.

704
01:50:48.096 --> 01:50:58.556
Like, you know, people are moving-- They seem to be moving away from speech, which is different than audio, um, because it seems to not generalize well to the other stuff. Mm. This might get solved.

705
01:50:58.896 --> 01:51:10.396
I'm a fan of all of this. But these are, like, empirical real questions. Like, scaling laws are not about just throw money at it, and it gets good. Scaling laws are pragmatically how good is a thing, you know?

706
01:51:10.696 --> 01:51:21.036
Like, like there are worlds where, like, no matter what you scale, it may not be good enough. So, um, y- y- you know, like computer use is not currently solved is my understanding.

707
01:51:21.296 --> 01:51:32.716
Like, I'm hoping that we can be a, like play a part in solving that. But, like, there, there might be no amount of data we collect that will solve that. We might need better methods or something else like that. So, um,

708
01:51:34.016 --> 01:51:38.456
like we-- you need to be like really practical in all of this. Am I a fan of omni models?

709
01:51:38.576 --> 01:51:48.576
I'm a fan of all forms of intelligence, but I will go straight into one thing you talked about, which is different from pre-training, which is post-training. Mm. Because I hate fracturing intelligence.

710
01:51:48.896 --> 01:51:57.296
That is like the bad thing to me, and this whole, like, chat first reasoning mode is because, um, it forces the intelligence to be fractured.

711
01:51:57.636 --> 01:52:08.556
Like, when you're optimizing for chat, this tends to be like pure RLHF, and it's quite intrinsic in RLHF to do the stuff people, like, naturally complain about, right? Like, oh, I'm gonna- You're absolutely right.

712
01:52:08.656 --> 01:52:15.176
And- Yeah... you know. Sycophancy, psychophancy- Yeah... whatever word- Yeah, yeah... how- however how to pronounce that, overconfidence, hallucination.

713
01:52:15.596 --> 01:52:23.116
Uh, like even the kind of style that excels in LM arena, bold, italicized, emojis. You know, like it doesn't answer the question simply.

714
01:52:23.136 --> 01:52:34.296
It gives like a long writeup, and then it asks you a follow-up question, so it feels more like a human talking to you. All of these things, um, come because strings are super weird. You know?

715
01:52:34.576 --> 01:52:40.096
They are like weird-ass things, and you need to be miscalibrated. You need to like mode drop.

716
01:52:40.156 --> 01:52:46.936
You need to be hyper-confident in order to not go off the rails because the, the reward model will punish you so hard when it happens because it's obvious.

717
01:52:47.496 --> 01:52:55.836
Y- um, and then this, like, warps the probability space entirely, and it interacts with that of the reasoning models, right?

718
01:52:55.896 --> 01:53:05.496
Because, like, it, you know, the, the models are like these sin- simple linear things that tend to cheat a bit. So I think that that's very different than exposing intelligence is, is my guess.

719
01:53:05.996 --> 01:53:19.196
And a lot of the art to intelligence is studying the subtlety that I think that, um, at least when I was in OpenAI, people were not really studying that because, like, they were just like, "Chat, chat, chat, chat," just like people are on with Jeff right now.

720
01:53:19.316 --> 01:53:22.856
Yeah, you give me an ultima- you know, you give me an objective, I will just go optimize for that, right? Yes.

721
01:53:22.856 --> 01:53:28.876
Like, and it, it- But, but if you try, and there-- And, you know, like the saying is like you could have like two objectives, and you could just like optimize for both.

722
01:53:28.916 --> 01:53:33.476
But then that is literally the act of fracturing, right? Um, so yeah.

723
01:53:33.776 --> 01:53:42.116
So, I mean, in some ways you ha- you are also factoring intelligence into system one, system two, but you just don't agree with the other people's fractur- fracturing, which is fine. Well, it, it- Which is fine...

724
01:53:42.196 --> 01:53:52.836
it's a little different. No, no, no. [laughs] If I could, if I could add-- If I could defend- Yeah... um, the system two tasks, um, number one, like we don't toss out the system two tasks, right?

725
01:53:52.896 --> 01:53:58.816
Like you can try to make Jeff work on it, and there actually is an intelligent answer for that, which is unknown.

726
01:53:59.496 --> 01:54:07.416
You know, like my, like there, like there is better and worse behavior in the system two tasks, which would be like really low confidence, lots of uncertainty.

727
01:54:07.796 --> 01:54:17.775
Maybe some heuristics can like move the needle here and there. But we care about them too, just to be clear. I just think that that is not what the, what is the intelligence is native to.

728
01:54:18.316 --> 01:54:24.256
So we're not trying to fracture anything like that. Um, and all fracturing makes the model dumb.

729
01:54:24.496 --> 01:54:33.176
You know, like if people, like get the model to say like it is OpenAI or Quinn or, uh, d- t- you know, like Claude or whatever else, I don't really know what it says this, these days.

730
01:54:33.636 --> 01:54:47.076
I am not going to put into the models that you are Jeff from Typesafe. That fractures it, right? Like, it, it's, like, like I, I don't want that. Like represent what do the internet thinks, right? Like be correct.

731
01:54:47.316 --> 01:54:54.276
That is what I want because that's how you get the smooth, predictable intelligence. I, I mean, identity is a thing, I guess, that- And for an API-...

732
01:54:54.316 --> 01:54:59.756
it may be somewhat of a special- For a, for a first party product, yes. Yeah. But, like, for an API, I don't think so. Yeah. Okay. You know? Yeah, that's good.

733
01:54:59.776 --> 01:55:09.676
Like, I, I don't, I-- Like people don't want, if they're making a chatbot with, uh, you know, ChatGPT, they don't want it to say it's ChatGPT. They want to say it's like Chipotlei or whatever, right?

734
01:55:09.796 --> 01:55:16.476
Well, you know, so the, the way that you also have to make up for it is you have the skill, right? Yeah. The, the, the Jeff skill. Yeah. Which, which is for coding agents to- Yeah, they do... work with Jeff.

735
01:55:16.976 --> 01:55:25.496
Um, okay, a couple closing questions- Hell yeah... because I, I, I do want to, uh, g- get you out. Uh, one is like just reflecting on your two-year journey. It's roughly two years? Two point something?

736
01:55:25.876 --> 01:55:30.636
Um, with the company- Yeah... I think that this is like more like a four-year journey. Yeah. But- Well, I- Yeah...

737
01:55:30.816 --> 01:55:35.836
yeah, actually, like, uh, I was thinking, remembering that, like you had this like hero run around Thanksgiving.

738
01:55:35.896 --> 01:55:44.616
You were like, you were canceling everything because, uh, you were like, "Guys, like the, everyone's on holiday. I'm gonna take all the OpenAI GPUs and g- go do this thing." Yeah. [laughs] That was a good time.

739
01:55:44.656 --> 01:55:52.276
And that was like the, the pre-Typesafe- Yeah, yeah, yeah... moment, right? I-- That might have been... Was that when the coup was happening? I don't really know. Yes, actually. [laughs] Yeah, yeah, yeah.

740
01:55:52.296 --> 01:56:03.320
That, that sounds right. Yeah. I remember. Oh, my God. I don't wanna- [laughs] I, I'm not, I, I don't think I have the time to spill the tea about the coup right now, but, um- That was really annoying.

741
01:56:03.780 --> 01:56:13.880
Um- The, the coup was annoying or the, the run was annoying? The, the coup was annoying. The coup. Okay. Yeah, yeah, yeah. Yeah. Uh, it, uh, I will- Safetyists took over the company... Yeah. Anyway.

742
01:56:14.820 --> 01:56:28.120
Maybe next time we chat- Okay. All right. All right... I'll, I'll dump tea about- Tea... a, a tea about the coup. Um, yeah. It actually, this problem was one that, like, was in my mind since before ChatGPT even launched.

743
01:56:28.520 --> 01:56:33.700
I was like, "Holy shit, the ChatGPT team is cooking. They are doing the right task.

744
01:56:33.980 --> 01:56:43.340
They are doing the thing that AI researchers are bad at, but successful product people are good at, which is giving a lot of fucks about the experience." You know, it's, it's, it's very rare.

745
01:56:43.580 --> 01:56:54.680
They, like, there's very few people like that at OpenAI. Um, and those guys were cooking on it really, really well. And to be clear, this is the whole journey from GPT-3 to 3.5, which included AI Dungeon- Yeah...

746
01:56:54.700 --> 01:57:02.700
which you've talked about as like, yeah, well, that's, that's an example of a, a use case that we never predicted. Yes, exactly. That's right. Yeah. Well, uh, oh, yeah, that is, uh...

747
01:57:03.540 --> 01:57:07.420
Also, I had fought very, very hard to deploy InstructGPT.

748
01:57:07.920 --> 01:57:18.040
Um, like actually, the early versions of it were even trained with, like, an algorithm we didn't publish that I made myself because it was too slow to clean the, the PPO data, and I was like, "Fuck it.

749
01:57:18.080 --> 01:57:21.100
This is so fucking good. We need to get it in the hands of users." Mm.

750
01:57:21.140 --> 01:57:32.600
And, like, basically, immediately, it took 50% of the market share of LLMs at the time, and, but-- And we thought it- I, I made-- I went through great effort to make sure everything in our launch video is true.

751
01:57:32.960 --> 01:57:38.860
Um, you know, we-- I truly was thinking, like, "Is this AGI because it's superhuman at instruction in, instruction out?"

752
01:57:39.380 --> 01:57:46.260
Y- y- obviously, it's not, but, like, everyone, I think, should have an answer to why that was not AGI, 'cause it looks very smart.

753
01:57:46.860 --> 01:57:56.620
And my answer to that ended up, uh, you know, like, ended up only being used for copywriting. You know, Jasper AI, Copy.ai, like, writing, like, you know, what is now called slop on web pages.

754
01:57:57.100 --> 01:58:07.720
Um, and we were worried we made the internet a worse place, right? And I went back to the drawing board, and I was like, "What's missing? We are smart, clearly. Something is missing from it, like, creating value.

755
01:58:08.000 --> 01:58:15.620
What is it?" Like, I actually was doing more philosophy at the time of, uh, like, you know, like, "What is going on?" And the answer was, oh, machines.

756
01:58:15.940 --> 01:58:25.640
You know, the question I asked myself is like, let's work backwards from an AI-based economic revolution. When that happens, what will c- be... wh- what'll be calling the AI if AI is an API?

757
01:58:26.000 --> 01:58:38.140
Will it be humans, or it'll be code? And I figured it was many nines of code, and, but, like, all the optimization was going into the humans part. And then it clicked for me, and like, holy shit, this is the North Star.

758
01:58:38.560 --> 01:58:48.520
I think, like, I wrote a document. I was, like, talking to Sam about this. Sam was like, "This is so fucking good. You should go work on it." And we're like, "Yeah, yeah, yeah, Sam, I have a job."

759
01:58:48.820 --> 01:58:59.860
You know, like, I'm, you know, I, I was working on- Sam just told you to do it. Do, go do it. Like [laughs] Yeah. But, like, my, my guess at the time is, like, this is super obvious. Like, it's so unbelievably obvious.

760
01:59:00.220 --> 01:59:07.920
Anthropic must be working on this already. You know? And, like, we're already cooked, and, like, actually, OpenAI does better at, like, catching up than it does at, like, actually innovating.

761
01:59:08.340 --> 01:59:15.100
So, like, uh, uh, ChatGPT was a copy of Claude, right? Um, like, they had an internal thing. They- Yes... just didn't ship it. Yeah. Um- Claude, Claude in Slack.

762
01:59:15.140 --> 01:59:24.340
But, uh, but, you know, reasoning, I, I would say first-ish. Yeah. Yeah. But debatable how good of a product that is. Yeah. Um, great research, though. Super great research.

763
01:59:24.380 --> 01:59:35.220
I'm just not sure if people had that product need. Um, and, you know, Claude did the coding agent stuff too. So, Sam says that, and, you know, I just go back to my job for a while.

764
01:59:35.580 --> 01:59:44.380
Eventually, like, you know, the instruction following team just says, "We won. We, we've solved instruction following. We don't need to do stuff anymore." I'm, like, trying to think about what I do next.

765
01:59:44.440 --> 01:59:54.980
I was like, "You know, maybe I'll just, like, start playing around with this." Um, I, you know, do more philosophy and design and thinking. I thought it would end up taking a week, uh, when I started training models.

766
01:59:55.360 --> 02:00:08.680
It ended up taking, um, many years. At, at some point, I was like, "Holy shit, you know, there's signs of life here. This..." It, it obviously didn't work, right? Otherwise, we would have deployed it.

767
02:00:09.000 --> 02:00:20.340
But, like, I wanna explore what it would be like research-wise to go all in on this. You know, like, I wanna really see, like, like, what it would be like if you went, like, absolutely, insanely all in in this direction.

768
02:00:20.760 --> 02:00:27.400
Uh, because of what I said, you know, like, if an AI winter happened, would I-- how would I feel? I would consider myself personally responsible.

769
02:00:27.840 --> 02:00:39.560
I talked to other companies at the time, and I was like, "Hey, I want to start a lab on this direction." And, you know, like, there was interest. And I just talked to them like, "How fast-- What would be faster?

770
02:00:39.620 --> 02:00:49.380
This or a startup?" And they're like, "Startup." And I'm like, "Fuck it, man. We ball." Yeah. "I guess we're doing some crazy shit." [laughs] And- And you called Eric and Sacha and- Yeah. Well- Yeah...

771
02:00:49.460 --> 02:00:59.700
I, I call Eric first. Um, with Sacha, I actually didn't try to recruit her. I, I tried to be good, and I was just like- Mm... "Hey, am I crazy? Is something missing here?

772
02:01:00.060 --> 02:01:09.300
You know, isn't there, like, uh, l- like, like, am I too much in the OpenAI bubble that I didn't realize there must be a solution to this?" And then Sacha was like, "I'm in."

773
02:01:09.600 --> 02:01:17.680
And I'm like, "Sacha, you're working at a startup." And she's like, "I'm folding it right now." And I'm like, "Do you wanna think about that?" She's like, "Oh, yeah, good point. Let me think about it."

774
02:01:17.680 --> 02:01:27.400
[both laughing] And then she joined. Yeah. And then, you know, within two weeks, we had funding. We, we had, like, people move into my apartment. It was the worst 'cause I'm a neat freak.

775
02:01:27.880 --> 02:01:38.300
Um, and we just kept on cooking, and eventually, we got the research that, that, that showed the signs of life. Yeah. You know? It was, it was a crazy time.

776
02:01:38.640 --> 02:01:43.760
So the, the, the qu- the question is, that was all long context. Oh, yeah. And then another question is, someone like you- Yeah...

777
02:01:43.860 --> 02:01:53.000
is in the Frontier Lab right now, who is frustrated not getting the, the funding or the, the resources, whatever, uh, the attention. Uh, what's your advice to them? You know, do, should they do what you did?

778
02:01:54.820 --> 02:02:07.160
Should they do it? Ooh, that's a fascinating question. Um... Ooh, man, how do I do this without burning bridges? [laughs] Um,

779
02:02:08.420 --> 02:02:22.340
I, my sense is that most, n- unless there's some level of economics I don't really understand, um, I think most neo-labs are crap. Um, I, I don't want to see myself with that as peers.

780
02:02:22.920 --> 02:02:33.680
Um, like I, I don't really understand what's going on there. Like, is it because... Like, number one, I don't really value researchers. Um, I value people who, like, like, look at my Bitters lesson, right?

781
02:02:33.760 --> 02:02:41.860
The data, the task. I, I, I want... Well, not just that. Yeah. I, like, we need researchers, but we need them to give a lot of fucks about the right task, and that's the important thing, right?

782
02:02:42.080 --> 02:02:48.280
So it's actually, like, the, it's, it's kind of backwards when people value pure research pedigree, 'cause that generally doesn't create value.

783
02:02:48.840 --> 02:03:04.640
Um, so it, like, like, number one, I, the, I believe in North Star tasks and doing cool, really useful stuff. Um, number two, um, because I don't value researchers, I don't, I don't recommend going the...

784
02:03:05.340 --> 02:03:17.360
Well, it clearly is profitable for someone, or it might be in this environment. So, like, from a purely pragmatic perspective, I don't see creating neo-labs as want... something that creates value.

785
02:03:17.900 --> 02:03:25.860
It seems to destroy value because, like, they are, like, redoing work from scratch with, like, low probability of actually moving the frontier.

786
02:03:26.320 --> 02:03:36.220
And as far as I've talked to most neo-labs, they don't really have a direction. They tend to want money to play around with their experiments. Um, if they have a direction, I'm super in favor of it, to be clear.

787
02:03:36.440 --> 02:03:51.340
So my advice for someone is it really depends on why you're doing it, you know? If you are a researcher who wants to play a- around with research, um, probably the labs are the best place to do that, TBH.

788
02:03:51.600 --> 02:04:00.520
Like, there might be other places. I, I don't really keep track of that politics. But I would just recommend not being that way personally, you know?

789
02:04:00.580 --> 02:04:09.620
Like, I think it's better for the world with people being driven to solve real problems. Uh, and those problems may be exploratory. That's fine. But, like, ideally have principles that you stand behind.

790
02:04:10.120 --> 02:04:22.560
But if you think that you wanna do the right task, like, abso-fucking-lutely, like, please do. Like, please break this, like, unimind, you know, uni- unimodal, like- Hive mind. Yeah. Yeah, exactly.

791
02:04:22.880 --> 02:04:37.340
Like, like, ev- you know, like, again, this pacing the frontier is coming from, like, this one view of AI that looks like, you know, AI super genius that is incredibly jagged. And that is, um- Solvable...

792
02:04:37.540 --> 02:04:45.900
you know, it's solvable, and it's weird, and it's, like, not matching reality. And, um, it, it's, like, it's tragic, right?

793
02:04:46.020 --> 02:04:51.260
Like, like I, I think, like, all of these, like the, like, really unearthing technology I think is, like, just good. Yeah.

794
02:04:51.620 --> 02:05:00.000
For what it's worth, uh, you know, a- again, I, I'm trying to rep- accurately represent the position of the, uh, anthropic Open AI folks I was talking to, SpaceX as well, by the way. Oh.

795
02:05:00.000 --> 02:05:10.620
Uh, is that, uh, it is, this is a political thing much more so than a pure xRisk thing. Yep. Uh, so yeah. Of course. Political positioning is- And, and that, and that's beyond my pay grade. Exactly, yeah.

796
02:05:10.660 --> 02:05:21.200
That's well beyond my pay grade. Once they sh- once they told me that, I was like, "I get it. Uh, this is about the 2028, uh, election." [laughs] Oh, no. Yeah. [laughs] Oh, I wish I didn't hear that.

797
02:05:21.240 --> 02:05:28.420
That's such a bad vibe. [laughs] No, no, no, no. And so- This is not the whole company. Yeah. This is just that, that room's discussion. No, no, no, that, that- Yeah... makes sense. Yeah.

798
02:05:28.440 --> 02:05:35.220
That, that, that makes me lose faith in humanity a bit, but maybe I'm just a naive technologist. Uh, it's really starting to matter- Yeah...

799
02:05:35.260 --> 02:05:44.140
who's, who's in charge of the governments, uh, that, that will help to regulate, uh, these things as they emerge. And, like, as a lab- I, I totally-... you should probably think that through. No. Uh, no.

800
02:05:44.140 --> 02:05:58.980
[laughs] I, I totally agree with that, to be clear. Like, I, I think being opinionated on that matters a lot. I personally am afraid of trying to mislead people because I think that bites people in the ass a lot.

801
02:05:59.480 --> 02:06:11.360
You know, like, um, I think that, like, people trying to be overconfident, like, uh, I, I, I, I, I, I obviously t- I, I'm not actually gonna talk about politics.

802
02:06:11.740 --> 02:06:23.420
I think what happened in COVID is, like, people leaned too much in, like, appeals to authority and being overconfident to try to get people to behave in certain ways, and, like, obviously our response was extremely suboptimal.

803
02:06:23.760 --> 02:06:32.380
And that had, like, like, ripples of downstream ramifications that are now, I think, extremely bad for the world. Like, maybe I'm naive.

804
02:06:32.600 --> 02:06:42.220
I think that misleading people, even for the greater good or what it- they think is the greater good, is just, uh, it's just, I, I'm not a fan. Yeah. I'd rather, I'd rather not do it.

805
02:06:42.260 --> 02:06:49.180
For what it's worth, I, I, it's not a- Yeah... I don't think it's misleading. It is just, like, this is why now. Yeah. Uh, why, yeah. Like, you know.

806
02:06:49.220 --> 02:06:53.380
That, I, I think that if it, if it- Like, Dario said in May, like, how come we're doing it now? I think if that is why now- Yeah...

807
02:06:53.400 --> 02:07:02.520
that is a little bit of misleading about, like, the risks versus, like, the, the objective. It, there is, like- Yeah... some level of, like, sneakiness latent in it- Yeah, yeah...

808
02:07:02.580 --> 02:07:11.760
that, um, is worth calling out and I think owning up to. Well, ob- obviously they want to, if, if they want to manipulate, then they shouldn't own up to that. That seems like a bad strategy. No.

809
02:07:11.820 --> 02:07:23.740
But, like, that to me is just, just sad for the world. Yeah. Hopefully, I'm never... I've- [laughs] Yeah. Uh, hopefully, like, we are never involved in anything- Yeah... like that. It might be inevitable as we get big.

810
02:07:24.260 --> 02:07:36.060
Um, but I want to, I wanna stay, like, pure technologist to my roots as much as I can. I mean, Jeff for president, why not? Uh, yeah. I, I, I can, I, you know, I would trust Jeff's decisions over, um, my, my own.

811
02:07:36.500 --> 02:07:42.680
Um, okay, so, so, uh, less shitposting, uh, more about- [laughs] That's shitposting. No, no, no, no. You're just cr- For, for me. I'm sh- I'm shitposting. Oh, okay.

812
02:07:42.760 --> 02:07:47.860
You're just crushing my hopes about, uh- No, no, no, I'm not calling you shitposting... like America and the world right now. [laughs] Oh, my Lord. Yeah.

813
02:07:47.920 --> 02:07:55.780
I mean, like, there's, uh, I, I think I watch too much TV about, like, conspiracies to think about the presidency. Oh, no. Um, the, uh, you have chosen your North Star.

814
02:07:55.940 --> 02:08:00.640
You have chosen reliability, y- and programmable and composable AI. And cheap. Uh, and cheap. Yeah.

815
02:08:00.880 --> 02:08:09.380
What is a second or third one that you wanna throw as a bone to someone else that you're not, that you want someone else to work on that you're not gonna work on? Ooh. Like, uh, just basically give people tasks.

816
02:08:10.344 --> 02:08:17.904
Give people tasks? Yeah, like that your tasks- Oh, there's so many I want. Oh, what? You have picked your tasks, right? You know what I mean? What? Wait, I-- that's such a good question. Holy crap.

817
02:08:18.344 --> 02:08:23.744
Oh man, I'm so excited by that. 'Cause you're, you're gonna be, you're b-- f-- so the next, like fifty years you're gonna be busy doing your thing. Hell yeah.

818
02:08:23.804 --> 02:08:38.784
Okay, so let me give like a fun one and a not fun, uh, l-l-- and like maybe a valuable one that's also fun. Um, my, my fun one is I think games could be so freaking cool if they were intelligent.

819
02:08:39.104 --> 02:08:49.224
Like when I see people play around with like, like Allie's Doom demo wh-where like you can like get NPCs to control stuff, like, you know, l-like that was just really like the- Yeah... like a proof of concept.

820
02:08:49.264 --> 02:09:01.044
I think some really cool stuff could be made. It looks really, really cool. You know, like, um, like I, I'm a big Stardew Valley fan, you know, and like it's, it's really static and it's still compelling.

821
02:09:01.084 --> 02:09:06.804
Like I feel like there's a lot of cool story that could happen. You don't need to call like Jev in the game loop. It's probably too expensive for that.

822
02:09:07.064 --> 02:09:18.084
But even like simple like state machines for NPCs, I think you could make like such a compelling world. Oh, man. Mm. Um, and man, a little sad that I can't work on these types of things. Yeah.

823
02:09:18.144 --> 02:09:25.304
My, my life path is a little bit set right now, and I'm, um, yeah. Yeah, but you can call someone else to work on it. Yeah, yeah. That's cool. And then, and then you can like- Yeah... feedback on it.

824
02:09:25.584 --> 02:09:34.484
And, um, the thing that I would really, really like to explore is like coding agents free from the tyranny of the KV cache.

825
02:09:34.864 --> 02:09:46.244
L-like it might not be as good as true coding agents are, but I think there's just so many weird things to think about. Tha-th-that's why I wrote the article, "KV Cache Rules Everything Around Me." Mm.

826
02:09:46.904 --> 02:09:58.324
Um, uh, believe it or not, I don't think anyone has used the phrase on the internet, "cache rules everything around me," C-A-C-H-E, um, when I, when I Googled it. Um, so- Okay...

827
02:09:59.044 --> 02:10:12.624
uh, like I wrote this 'cause I wanted to tell people about like this is how coding agents w- agents work and how the KV cache- Right... works and everything. And, um, I think, uh, oh, um, yeah, um,

828
02:10:13.744 --> 02:10:25.184
like, uh, it explains a lot of stuff, like why routing is really hard, why sub-agents don't seem to work, like why compaction is such a hard problem. And I, I'm going to try to release a document.

829
02:10:25.224 --> 02:10:35.924
My team might veto me because believe it or not, I'm not in charge. [laughs] Um, you know, but, uh, I, I wish. Um, but, um, I want to release a document of like here are my thoughts.

830
02:10:36.244 --> 02:10:46.324
Please play with it, and please figure out all the ways that we can do things with coding agents, like once you're freed from that, you know, that, that KV cache tyranny.

831
02:10:46.644 --> 02:10:56.144
Which is it locks you in, and- Well, not-- it lo- it, it locks you in into one model, right? And in order to do it efficiently, you need to like keep on appending to it. Mm-hmm.

832
02:10:56.204 --> 02:11:08.124
So now you're not doing best software practices like state management, abstraction, decomposition. Why can't you give an easier task some... Y-y-- why, why can't you give a sub-agent an easier task?

833
02:11:08.444 --> 02:11:12.244
Because of the state that you're passing around. Oh, I touched this. Uh, because of the state you're passing around.

834
02:11:12.644 --> 02:11:21.004
You nee- would need intelligence that is way cheaper than the intelligence using to read this in order to pass this state around. Why can't you be smart about it, right?

835
02:11:21.184 --> 02:11:30.664
And I think there's just like tons of really cool, fun research to be had there on like different programming patterns. You know, kind of like how people are playing around like with like recursive language models.

836
02:11:30.984 --> 02:11:39.004
Like I feel like there's like just lots of cool stuff in here when you think about like, oh, I want to explicitly label the state of everything. Or imagine you have like a sub-task.

837
02:11:39.264 --> 02:11:49.884
Like coding agents, I, I think it's fair to say they work on sub-tasks at a time as from a decomposition perspective. Why do you need to pass all of that state back into the parent task? Yeah.

838
02:11:50.184 --> 02:12:01.164
Why couldn't you do smart things about it? A-and also, if you had a hierarchy of labeled sub-tasks, why can't you do a search through that sub-task tree for the relevant context when you need it in, right?

839
02:12:01.484 --> 02:12:11.144
And then, you know, another thing that you can do... Oh, man, I forgot to write something about this. Um, uh, the, I have like some cooks in here that are really, really cool. Um, hope to publish it.

840
02:12:11.204 --> 02:12:13.724
I'm down to jam about it, but like it's gonna be a long document.

841
02:12:14.204 --> 02:12:24.044
And, and, like if that becomes the case where context becomes cheap, like why can't you do cool patterns like looking at your historical context very cheaply?

842
02:12:24.424 --> 02:12:30.084
Is it kind of weird that you start from scratch every time and you need to solve a problem called continuous learning?

843
02:12:30.444 --> 02:12:39.004
That's a, that's actually like a memory management problem because you don't have a smart way of looking up the memory, right? But what if you could? What if you could do that all the time?

844
02:12:39.364 --> 02:12:58.544
Or what if when you have parallel sub-agents, they can like read each other's states because you have all of that in like your computer memory, and you can be smart about what's reading and writing at the same time, and your coding agent swarm or whatever has like locks around things and can coordinate intelligently, not with like basic-ass locks.

845
02:12:58.584 --> 02:13:06.904
Like what are you doing? What am I doing? You know, Jev, who should write first? Blah, blah, blah. And l-like I feel like the future there is- Oh my God... nuts. Jev to solve locks.

846
02:13:07.324 --> 02:13:13.924
I mean, it could be so cool for like multiple agents working together. Or like if you think about state- Yeah... like when you have- Agent swarms stuff. Yeah. Yeah.

847
02:13:13.944 --> 02:13:22.904
And you know, some things, for example, are read-only processes. You know, some people like getting like summaries of what the agents are doing. Mm-hmm. Yeah. Why can't they share state easily?

848
02:13:23.224 --> 02:13:33.204
Because like a read-only agent needs to like, you know, read parts of the context and figure out what's relevant to say, well, like what's actually being written, 'cause exploration is not super important, or here is the tree of sub-tasks.

849
02:13:33.544 --> 02:13:46.384
I feel like there's so many different fun things that could be done if like a really smart person like dedicated like a whole lot of time to rethink like the, the coding agent experience, and that would be super duper sick.

850
02:13:46.824 --> 02:13:55.304
Yeah. Man, I, that would be my dream Uh, I, I would point you towards PrimeAgent if you haven't looked at it. Uh, so this, uh, uh, works together with the RLM work.

851
02:13:55.344 --> 02:14:05.684
We just, uh, talked to Alex, uh, who's a buddy of Alan's, um, i- in the chair before you. Oh, cool. Uh, and like, yeah. It, it is being worked on, but it's not super popular yet. Yep. And if, like- Well, yeah.

852
02:14:05.724 --> 02:14:10.444
I mean, but the hope-- Yeah, I would want everyone to, like, just play around- Yeah... with, with like weird things.

853
02:14:10.504 --> 02:14:18.703
I have no guarantees that it'll work, but it seems really, really interesting from like a technical perspective. So yeah, that, that seems cool. And cool- It seems cool...

854
02:14:18.724 --> 02:14:26.204
like, like, uh, like once we figure out how to give credits out, I would love to like give credits out to people like this. Yeah. You, you will be in a position to fund research for sure. Yeah.

855
02:14:26.304 --> 02:14:34.784
Uh, no, anyway, congrats on all your success. Uh, you, you've like, uh, come s- such a long way since I first met you, like, and, and the whole team as well. I'd like to think I'm the same person as well. Yeah. Yeah.

856
02:14:34.804 --> 02:14:42.544
Yeah. I, I think, but I think, like you are energized in a way that I have never seen you before because you found your mission. Yeah. You know? That's true. That's definitely true. Um, and like- I was-...

857
02:14:42.944 --> 02:14:50.784
you are articulating your mission because you, you, you, uh, for many years, you complained about the problems, but you didn't have a solution yet, right?

858
02:14:51.244 --> 02:15:01.904
And you're like, you had, you had the rough shape, and that-- then you had to do it, put in the work. I will say that that is partially because I, I describe myself as, as zeroth percent entrepreneurial.

859
02:15:01.904 --> 02:15:14.284
[laughs] I, I, I don't like startups. Um, I never wanted to be a CEO in my life. I can't imagine anyone doing this twice. It seems horrible. Honestly, doing it once is pretty bad.

860
02:15:14.844 --> 02:15:24.064
Um, when we first were fundraising, an investor asked me like, "Which CEOs do you look up to?" And I was like, "Ew, why would I look up to those people?" [laughs] Um, no offense to anyone.

861
02:15:24.304 --> 02:15:32.244
You know, I'm trying to be like, I'm trying to be genuine good, and I've met like a lot of really good people, but like the famous ones have like a lot of like skeletons in their closet, it seems.

862
02:15:32.684 --> 02:15:39.364
And I think I just really did feel disempowered when I was at OpenAI. You know, like I felt, um...

863
02:15:40.984 --> 02:15:46.644
Yeah, like, like, like it's a little bit easier to be truthful now because, like I have at least some proof that the direction has legs.

864
02:15:46.984 --> 02:15:57.164
Like, I just felt like in the, the, the insane house where everyone is just like, "ChatGPT, yeah." Like, "Where do we put ChatGPT in everything? How do we make ChatGPT good for like, you know, developers and stuff?"

865
02:15:57.204 --> 02:16:04.684
And I like, "What? What are you talking about?" Like, "The function calling interface is insane. Why would you deploy this?" Um, like this is just so anti-developer.

866
02:16:04.684 --> 02:16:16.124
It's sort of a hacky way on top of hacks on top of hacks. Well- Yeah... not just that. Like the thing I often said was if there was like a, uh,

867
02:16:17.204 --> 02:16:28.584
this is also probably tea I don't have time for right now, but I always used to say like, "I want to be removed from any project involving like function calling if you did not get a legit bias for each function."

868
02:16:29.184 --> 02:16:36.864
Like, so very- Mm... ve-very simple ask in my part because- Which is something like a confidence but not calibrated. Or, or a probability for it, right? Yeah.

869
02:16:36.904 --> 02:16:47.504
Like we need to give users the ability to control like, you know, like let's say they have actions- Oh, yeah... are refuse or allow. Um, yeah, Disney needs to set a, a different refusal threshold than AI Dungeon.

870
02:16:47.764 --> 02:16:53.424
The only way to control that with function calling right now is to say like, "Pretty please." You know? That's nuts.

871
02:16:53.644 --> 02:17:01.084
That's a nuts interface for developers and like people have been like dealing with this for years now, right? Like they still have it with skills.

872
02:17:01.304 --> 02:17:08.864
Like, um, you know, like, um, um, the existing coding agents are like highly overfit to their existing harness because they're jagged.

873
02:17:09.124 --> 02:17:21.064
They don't tend to use like external like tools and MCPs super well because of overfitting, of course. And like why can't like big companies a-allow for like slight nudges to be like, "Call this more.

874
02:17:21.104 --> 02:17:31.604
It's really useful," right? [laughs] And, and like the, the, the, the, the, the solution is begging in a system message. That's nuts. Hell yeah. Um, but, uh, uh, okay, I, I think I, I think I get you.

875
02:17:31.824 --> 02:17:41.564
Um, a-and like, man, uh, it is so exciting to, to talk about all this stuff. Thank you. It's, it's really cool to get you on a podcast. Uh, you're gonna go, uh, do amazing things, man. Like, I'm excited- Yeah...

876
02:17:41.564 --> 02:17:51.304
for your next, uh, big launches, whatever it is. Oh, hell yeah. Yeah. Just you wait. Yeah, yeah. Just you wait. It might be sooner than you think. So hiring data people, infra people, I assume. Marketer. Why?

877
02:17:51.324 --> 02:18:01.864
[laughs] Depends on who you ask. Community person. If you ask me- Yeah... I feel like I'm a pretty good founding marketer. But if you ask anyone on my team, they say, "Shut the fuck up, Yogo. You need to do CEO stuff."

878
02:18:01.884 --> 02:18:08.264
[laughs] So yes, founding marketer- It, it is not just about spice. Like I think you're very spice-oriented, which like you, like that's- Mm-hmm... your unique talent.

879
02:18:08.564 --> 02:18:17.424
But sometimes you just need to say like- I know, I know... yeah, the routine marketing things. I, I would really love-- Yes. I-- Nothing teaches you delegation like having a tidal wave of stuff to do.

880
02:18:17.824 --> 02:18:28.764
Um, hiring data people, or we call them model capabilities, like, uh- Yeah... but they are data people, both-- Like, like, uh, data is kind of a slur in the industry and like I want to make sure- I don't think so.

881
02:18:28.804 --> 02:18:35.464
We are very pro-data here. Yeah, but I want them to be the highest status of like, you know, the people actually working on the model that actually sounds weird.

882
02:18:35.524 --> 02:18:42.964
I want everyone to have equal status, but like I want to even that out, and I want to know that that's really valuable. These are more equal than others. [laughs] Well, I mean...

883
02:18:43.304 --> 02:18:53.264
[laughs] I, I don't like weird hierarchies, and I, I think one of the things I'm most proud about in the company is that they don't respect me that much, or they don't show that.

884
02:18:53.304 --> 02:18:59.564
They, they troll me and like joke with me, and they treat me poorly sometimes and all of that, and I think that that's a good sign of a culture.

885
02:18:59.944 --> 02:19:09.764
We're hiring like platform people, like people to like build out Jev everywhere. Like we are so much more sensitive to location because speed of light is more of a bottleneck. Mm. Right?

886
02:19:09.804 --> 02:19:20.184
Like I'm so sad for the European users that they were only like three times as fast instead of like 100 times as fast because like we don't have servers there right now, and like, like that's insane, right?

887
02:19:20.284 --> 02:19:30.384
But like- It's, it's okay. Life in Europe goes a bit slower as well. It's okay. [laughs] Wow. I can't believe you, you said it, not me. Um, uh, or, or everywhere. Yeah.

888
02:19:30.424 --> 02:19:41.024
You know, like if, if intelligence per second is a metric that matters, like we, we want this all over the place. Like we care about... Like if they're a developer building on top of us, I care a lot about you.

889
02:19:41.504 --> 02:19:54.044
Um, and we are hiring for people to keep building more s- like not just, like the goal is not to just be like Jev as a company. The goal is to like ship more shapes of intelligence beyond that.

890
02:19:54.364 --> 02:20:03.024
So we are hiring people to like build those things too. You know, like we want to not just be like, yeah, like the one-trick pony of like the simple model.

891
02:20:03.364 --> 02:20:11.284
But like I think that there's gonna be like an AWS of like intelligence, you know? Which- Which is gonna be you, by the way, right? Yes. Uh, I mean, that's a direction I want to go down. Yes. Okay.

892
02:20:11.304 --> 02:20:21.944
It would be arrogant to say it will be me. Yeah. Uh, like we-- Like I'm going to do anything I can to make sure that happens. Yeah. Like I think that that's gonna be so, so cool.

893
02:20:22.264 --> 02:20:34.404
You know, like we are playing with like system one intelligence right now. Imagine the layers, you know? Like this is like the TCP of it. Mm. Yeah. Uh, seven more layers to go. Yep. And, uh, who, who knows what else?

894
02:20:34.744 --> 02:20:43.444
I've also pitched Temporal, by the way. I, I don't know. We, we need to talk about Temporal as layer eight- Ooh... uh, out, out of the seven layers. Um- Ooh... but anyway, we can talk forever. Hell yeah.

895
02:20:43.484 --> 02:20:51.804
You gotta get back to work or sleep. Yep. Uh, thank you for coming. Oh, boy. Yeah. Cool. You're most welcome. It was a pleasure, man. Yeah. So excited. Yeah. So excited. You picked my first time. Not the last time.

896
02:20:51.824 --> 02:20:52.284
Not the last time
