1
00:00:00,018 --> 00:00:04,308
Josh:
Over the last 90 days, Frontier Labs shipped 15 plus models.

2
00:00:04,308 --> 00:00:07,535
Josh:
OpenAI shipped three, Anthropic shipped four, Google shipped three,

3
00:00:07,692 --> 00:00:12,272
Josh:
even Meta and Elon Musk shipped Frontier models. The Chinese shipped a bunch as well.

4
00:00:12,463 --> 00:00:14,768
Josh:
And so if you're listening to the show, you're probably wondering,

5
00:00:14,930 --> 00:00:16,698
Josh:
which model should I be using right now?

6
00:00:16,698 --> 00:00:20,508
Josh:
Most of you likely have a GPT or Claude subscription, but you're wondering,

7
00:00:20,508 --> 00:00:22,288
Josh:
should I be using these different models?

8
00:00:22,288 --> 00:00:26,338
Josh:
The truth is, the Frontier landscape of AI models, when it was originally thought

9
00:00:26,338 --> 00:00:30,398
Josh:
to be one or two, has now expanded to hundreds and hundreds of models.

10
00:00:30,398 --> 00:00:34,124
Josh:
In OpenRata alone, you can access 400 plus. And so on this episode,

11
00:00:34,281 --> 00:00:38,958
Josh:
we're going to dig into which model you should use for what particular use and

12
00:00:38,958 --> 00:00:40,128
Josh:
when makes the most sense.

13
00:00:40,128 --> 00:00:44,118
Josh:
And you'll realize that the argument has shifted not from using the most intelligent

14
00:00:44,118 --> 00:00:48,393
Josh:
model, but maybe using the most cheaper model or using the model that's specific for you.

15
00:00:48,608 --> 00:00:51,698
Ejaaz:
You got options, baby. There's a lot going on in the AI space and we're going

16
00:00:51,698 --> 00:00:55,528
Ejaaz:
to help navigate that space because there is a lot of options and it does get

17
00:00:55,528 --> 00:00:59,498
Ejaaz:
overwhelming. I think between the two of us, we've probably touched every single

18
00:00:59,498 --> 00:01:01,018
Ejaaz:
one of these, at least a couple of times.

19
00:01:01,018 --> 00:01:02,398
Josh:
I have too many subscriptions, Josh.

20
00:01:02,398 --> 00:01:05,518
Ejaaz:
So yeah, between our 75 different subscriptions, we have covered these.

21
00:01:05,518 --> 00:01:10,057
Ejaaz:
We have some feedback about which one to use for when, which is best for which use cases.

22
00:01:10,249 --> 00:01:13,758
Ejaaz:
And I guess to start, we have this like pretty helpful visual companion artifact

23
00:01:13,758 --> 00:01:15,198
Ejaaz:
here that can walk through kind

24
00:01:15,198 --> 00:01:18,323
Ejaaz:
of what we're thinking, how we think about this. The first is the split.

25
00:01:18,573 --> 00:01:23,118
Ejaaz:
There are two distinct classes of model when it comes to considering which ones

26
00:01:23,118 --> 00:01:25,039
Ejaaz:
to use. The first is open source.

27
00:01:25,259 --> 00:01:29,448
Ejaaz:
These are Chinese models predominantly. Most of these are your Kimis,

28
00:01:29,448 --> 00:01:32,918
Ejaaz:
your DeepSeaks, like all of those models are Chinese open models.

29
00:01:32,918 --> 00:01:36,149
Ejaaz:
And then we have the actual US models that are all closed source.

30
00:01:36,324 --> 00:01:38,628
Ejaaz:
There's a big difference between the two. And we could see if we scroll down

31
00:01:38,628 --> 00:01:42,078
Ejaaz:
a little bit, the difference in usage between these two, because in June of

32
00:01:42,078 --> 00:01:48,340
Ejaaz:
2025, US Labs accounted for 70% of the tokens generated. And now they're down to 30%.

33
00:01:48,729 --> 00:01:53,808
Ejaaz:
The economics have changed widely. So now that 30% is worth much more than 70%.

34
00:01:54,069 --> 00:01:57,570
Ejaaz:
But it is interesting to note that there has been this kind of reinvigoration

35
00:01:57,570 --> 00:01:59,890
Ejaaz:
of Chinese models over the past couple of months. Now, granted,

36
00:01:59,890 --> 00:02:01,494
Ejaaz:
this is based on open router data.

37
00:02:01,709 --> 00:02:06,370
Ejaaz:
Open router is a single model router instance. This is not reflective of the norm.

38
00:02:06,660 --> 00:02:09,700
Ejaaz:
But it's just worth noting that some of these open source models are pretty

39
00:02:09,700 --> 00:02:12,088
Ejaaz:
powerful for the task at hand.

40
00:02:12,402 --> 00:02:16,450
Josh:
Yeah, I think no one can debate the fact that people have shifted a lot using

41
00:02:16,450 --> 00:02:19,250
Josh:
these open source models. And it's for a variety of different reasons.

42
00:02:19,250 --> 00:02:22,605
Josh:
People want to own their own data, they want to run it privately at home.

43
00:02:22,907 --> 00:02:26,910
Josh:
But the biggest shift, the biggest reason has been because these models are

44
00:02:26,910 --> 00:02:32,070
Josh:
a lot cheaper. And open router data you mentioned only really represents a very

45
00:02:32,070 --> 00:02:35,910
Josh:
niche sector of software engineers that want to experiment with a bunch of these models.

46
00:02:36,212 --> 00:02:40,360
Josh:
But even in the enterprise world, where OpenAI and Anthropic are pretty dominant

47
00:02:40,360 --> 00:02:44,940
Josh:
with their own model share, they've been losing the token market share to enterprises

48
00:02:44,940 --> 00:02:50,516
Josh:
that are trialing and testing different models to save tens to hundreds of millions of dollars as well.

49
00:02:50,760 --> 00:02:55,610
Josh:
And when I look at the main reason why, it's not because the open models are

50
00:02:55,610 --> 00:02:58,266
Josh:
the most intelligent. You'll see in a second as we go through the scorecard,

51
00:02:58,550 --> 00:03:02,740
Josh:
GLM 5.3 from China, amazing model, not as good as Fable 5. You'll look at Kimi

52
00:03:02,740 --> 00:03:04,634
Josh:
K 2.7, you'll think the same thing.

53
00:03:05,458 --> 00:03:08,856
Josh:
Those models are good enough to do the bulk of your own work.

54
00:03:08,856 --> 00:03:11,420
Josh:
It doesn't sound too much like AI slop.

55
00:03:11,605 --> 00:03:14,826
Josh:
It actually just speaks to you normally. And the biggest advantage is it has

56
00:03:14,826 --> 00:03:20,334
Josh:
fewer safeguards, which has its pros and cons, but it basically does the tasks that you ask it to.

57
00:03:20,566 --> 00:03:24,026
Josh:
Now, overall, it's fine if we talk about a bunch of models on this show,

58
00:03:24,252 --> 00:03:30,046
Josh:
but it's good to get an overview, essentially, of how intelligent or how effective these models are.

59
00:03:30,046 --> 00:03:34,237
Josh:
And what we have on the screen here is something known as the Artificial Analysis

60
00:03:34,353 --> 00:03:36,112
Josh:
Index or Intelligence Index.

61
00:03:36,344 --> 00:03:40,196
Josh:
And this is basically the best benchmark to test the general intelligence of

62
00:03:40,196 --> 00:03:41,016
Josh:
these different models.

63
00:03:41,016 --> 00:03:45,486
Josh:
Now, it may come as no surprise to you, but the Claude models,

64
00:03:45,486 --> 00:03:49,116
Josh:
Anthropix models, top this. We've got Opus 5 at the top, which is their most

65
00:03:49,116 --> 00:03:51,391
Josh:
recent model launch. We've got Claude Fable 5.

66
00:03:51,791 --> 00:03:55,886
Josh:
And then, surprising to me, Josh, but Grok 4.6 from Elon, it is just,

67
00:03:56,406 --> 00:04:03,500
Josh:
it is fantastic how effectively SpaceX has been able to pivot from being way behind the model race.

68
00:04:03,755 --> 00:04:08,656
Josh:
They fired their entire AI team, hired a bunch of new folk, acquired Cursor

69
00:04:08,656 --> 00:04:13,084
Josh:
for $60 billion and put out a model in, I think it's like the last six weeks

70
00:04:13,304 --> 00:04:17,216
Josh:
that was able to contend with the top. And the best part is he has all the GPUs

71
00:04:17,216 --> 00:04:18,353
Josh:
and compute to train them.

72
00:04:18,621 --> 00:04:21,796
Josh:
And then if you go further down the stack, you'll see the likes of OpenAI,

73
00:04:21,958 --> 00:04:25,499
Josh:
GPT 5.6 Sol, and a bunch of other Chinese open source models.

74
00:04:25,772 --> 00:04:29,186
Ejaaz:
The question is, what do you make of this? As someone who is just a user of

75
00:04:29,186 --> 00:04:31,546
Ejaaz:
AI, what do you make of this? How do you navigate this?

76
00:04:31,966 --> 00:04:35,786
Josh:
Let me trial this, Josh. Let me ask you this question. Of these models that

77
00:04:35,786 --> 00:04:40,871
Josh:
you see on the screen here, which ones are you using and for what specifically?

78
00:04:41,155 --> 00:04:45,376
Ejaaz:
It's funny because about 90% of the usage comes from the top two.

79
00:04:45,892 --> 00:04:50,135
Ejaaz:
And then 10% comes from everything else. And this is kind of like what I was

80
00:04:50,135 --> 00:04:52,585
Ejaaz:
going for. I'm actually curious, is it a similar thing for you?

81
00:04:52,585 --> 00:04:54,362
Ejaaz:
Like what kind of models are you using?

82
00:04:54,617 --> 00:05:01,475
Josh:
Yes, but okay. So I would say it's around, I'm going to say like 75, 25.

83
00:05:01,475 --> 00:05:06,255
Josh:
And the reason why I have that slight adjustment is because I kind of use the

84
00:05:06,255 --> 00:05:10,665
Josh:
AI model that's most convenient to me wherever I'm getting the information or

85
00:05:10,665 --> 00:05:15,276
Josh:
the inclination to use an AI model. So X is a perfect example, right? I'm scrolling X.

86
00:05:15,439 --> 00:05:18,435
Josh:
I read something from one of these genius AI researchers and I'm like,

87
00:05:18,435 --> 00:05:19,949
Josh:
I have no idea what the hell this means.

88
00:05:20,896 --> 00:05:25,197
Josh:
I tap the Ask Rock button. I'm on Google. I'm searching something, right? I'm on my laptop.

89
00:05:25,394 --> 00:05:28,871
Josh:
I'm like, okay, I'll just speak to Gemini in this. So in that sense,

90
00:05:29,063 --> 00:05:32,016
Josh:
wherever the AI is conveniently placed, I use it there.

91
00:05:32,016 --> 00:05:35,116
Josh:
And I have a feeling when Apple releases like their new AI model in a couple

92
00:05:35,116 --> 00:05:38,186
Josh:
of months or a couple of weeks, actually, I'll probably use it on my phone as

93
00:05:38,186 --> 00:05:41,503
Josh:
well because the models just generally are good enough to answer my basic questions.

94
00:05:42,096 --> 00:05:43,947
Ejaaz:
Yeah, okay. So we're looking at these charts and we're seeing,

95
00:05:44,162 --> 00:05:47,236
Ejaaz:
okay, here's where the intelligence is. Here's where things get a little less

96
00:05:47,236 --> 00:05:48,318
Ejaaz:
intelligent, but much cheaper.

97
00:05:48,504 --> 00:05:52,226
Ejaaz:
And if you're a normal consumer, I have a feeling that cost isn't that big of

98
00:05:52,226 --> 00:05:55,481
Ejaaz:
a deal because the cost differences are really at scale.

99
00:05:55,835 --> 00:05:58,936
Ejaaz:
What a lot of these companies offer, what Anthropoc offers, what OpenAI offers,

100
00:05:58,936 --> 00:06:03,817
Ejaaz:
what Grok and Gemini offer is just the $20 a month, $100 a month, $200 a month plan.

101
00:06:04,102 --> 00:06:07,974
Ejaaz:
You could just pay a fixed amount and get access to a lot of usage of these models.

102
00:06:08,223 --> 00:06:14,102
Ejaaz:
So if you are a general consumer who is able to spend $20 per month on a subscription,

103
00:06:14,370 --> 00:06:16,662
Ejaaz:
you're probably better served going to one of the major iolabs.

104
00:06:16,860 --> 00:06:18,756
Ejaaz:
The product is better. You're going to get more intelligence.

105
00:06:18,756 --> 00:06:20,087
Ejaaz:
You're going to get a more coherent product.

106
00:06:20,256 --> 00:06:23,657
Ejaaz:
I find that a lot of the people who I speak to a lot of the time in my experience,

107
00:06:23,971 --> 00:06:28,458
Ejaaz:
the only real reason to use these open source Chinese models is if you're consuming

108
00:06:28,812 --> 00:06:32,836
Ejaaz:
tremendously large amounts of tokens for, say, agentic tasks.

109
00:06:32,836 --> 00:06:35,826
Ejaaz:
Like if you're running an open claw instance or any sort of claw with a lot

110
00:06:35,826 --> 00:06:39,106
Ejaaz:
of automated workflows, a lot of agents, you may want to defer to some of these

111
00:06:39,106 --> 00:06:42,651
Ejaaz:
cheaper, perhaps the Chinese open source models just to save on costs on the low end.

112
00:06:42,959 --> 00:06:46,706
Ejaaz:
But when it comes to day to day use on the high end, I'd say probably 90 plus

113
00:06:46,706 --> 00:06:49,936
Ejaaz:
percent of the people watching this are best served just getting a subscription

114
00:06:49,936 --> 00:06:51,336
Ejaaz:
to your favorite ai company

115
00:06:52,032 --> 00:06:55,585
Ejaaz:
Using that product. That's kind of how I use it. It's funny that the remaining

116
00:06:55,585 --> 00:07:00,215
Ejaaz:
10% of prompts that aren't done through Anthropic are done mostly through Grok.

117
00:07:00,215 --> 00:07:04,095
Ejaaz:
Like you mentioned, when I'm on X 24 hours a day, it's a good companion because

118
00:07:04,095 --> 00:07:05,343
Ejaaz:
it has access to that database.

119
00:07:05,535 --> 00:07:09,415
Ejaaz:
And then the other one is actually ChashyBT because they have a really amazing

120
00:07:09,415 --> 00:07:10,427
Ejaaz:
image generation model.

121
00:07:10,619 --> 00:07:14,090
Ejaaz:
And I love creating memes or sending images to my friends who are just like

122
00:07:14,374 --> 00:07:16,830
Ejaaz:
using that as the visual component. I find that really compelling.

123
00:07:17,056 --> 00:07:20,882
Ejaaz:
So that's kind of my stack is like mostly Anthropic, sometimes Grok,

124
00:07:21,120 --> 00:07:24,625
Ejaaz:
sometimes chat gpt particularly for the image gen i think when it comes to image

125
00:07:24,625 --> 00:07:28,345
Ejaaz:
gen we used to use nano banana and gemini all the time that was like the top

126
00:07:28,345 --> 00:07:31,735
Ejaaz:
dog i think ever since image gen 2.0 release from chat gpt

127
00:07:32,115 --> 00:07:37,225
Ejaaz:
is very compelling what i like is that it does get actual thinking behind it

128
00:07:37,225 --> 00:07:40,515
Ejaaz:
so previously when you asked to generate an image it would take your words at

129
00:07:40,515 --> 00:07:44,601
Ejaaz:
face value generate the image now it actually does some inference it does some thinking

130
00:07:44,885 --> 00:07:48,206
Ejaaz:
it kind of interprets your message and interprets what it's seeing in the image

131
00:07:48,514 --> 00:07:50,992
Ejaaz:
and then creates a much better output. So that's kind of how I would think

132
00:07:51,364 --> 00:07:55,479
Ejaaz:
of navigating this um like chinese models are certainly a big deal but unless

133
00:07:55,479 --> 00:07:59,967
Ejaaz:
you are kind of advanced super user consuming a ton of tokens perhaps they're not that valuable

134
00:08:00,192 --> 00:08:03,769
Josh:
I will speak from my own personal experience i

135
00:08:04,129 --> 00:08:06,990
Josh:
use the clawed models pretty much all the time

136
00:08:07,286 --> 00:08:11,041
Josh:
but i've started using opus 5 more than fable 5

137
00:08:11,296 --> 00:08:15,019
Josh:
and and this might sound like a silly reason but it's because it speaks to me

138
00:08:15,019 --> 00:08:21,025
Josh:
like a normal human being i don't know what the personality dials are on opus 5 versus fable 5 but

139
00:08:21,240 --> 00:08:25,059
Josh:
they need to do that to fable 5 pronto because fable 5 just waffles and speaks

140
00:08:25,059 --> 00:08:27,889
Josh:
in like archaic prose and i'm just like listen just give me

141
00:08:28,269 --> 00:08:31,089
Josh:
give me the information that i've asked for and do not give me anything else

142
00:08:31,089 --> 00:08:32,147
Josh:
i don't want to have to read,

143
00:08:32,669 --> 00:08:37,499
Josh:
an essay every time i use this right now grok 4.6 i've actually noticed the

144
00:08:37,499 --> 00:08:41,239
Josh:
intelligently josh i don't know if you've tried this on x you probably do when

145
00:08:41,239 --> 00:08:45,869
Josh:
you're like kind of like trying to figure It speaks to you super intelligently

146
00:08:45,869 --> 00:08:47,479
Josh:
and there's less crass about it.

147
00:08:47,479 --> 00:08:52,039
Josh:
If you remember the earlier versions of Grok used to kind of like throw in some

148
00:08:52,039 --> 00:08:54,891
Josh:
kind of like questionable words, slang and phrases.

149
00:08:55,147 --> 00:08:57,219
Josh:
And now it just speaks to you like an intelligent human being.

150
00:08:57,219 --> 00:09:01,161
Josh:
And I think this is because Elon's like, you know, SpaceX is IPO'd.

151
00:09:01,370 --> 00:09:04,909
Josh:
He's got Cursor, very professional researchers involved, and they're trying

152
00:09:04,909 --> 00:09:08,609
Josh:
to target a more enterprise oriented audience. And he said before that Grok

153
00:09:08,609 --> 00:09:12,519
Josh:
4.7 is on its way to being released in a few weeks now. He's aiming for Grok

154
00:09:12,519 --> 00:09:15,463
Josh:
5 by the end of the year. That's literally only in a couple of months.

155
00:09:15,724 --> 00:09:20,119
Josh:
He is really targeting the enterprise landscape because he's seen Anthropic and OpenAI go after it.

156
00:09:20,368 --> 00:09:25,012
Josh:
Now, same as you, on the Chinese models, I am not using any of them.

157
00:09:25,175 --> 00:09:27,735
Josh:
And I admit this might be my own bias because...

158
00:09:28,420 --> 00:09:32,642
Josh:
Mainly, I'm like, I'm not sure, like, I want to go to the efforts of signing

159
00:09:32,642 --> 00:09:36,092
Josh:
up on an account, using those models, and then being like, well,

160
00:09:36,092 --> 00:09:38,072
Josh:
I don't know whether this is giving me the right information.

161
00:09:38,072 --> 00:09:41,406
Josh:
I don't know. I just kind of have this weird trust thing with the US and American

162
00:09:41,673 --> 00:09:43,884
Josh:
brands. And maybe that's like my own fault.

163
00:09:44,204 --> 00:09:47,262
Josh:
And then the one thing that I'll point out across all these different models

164
00:09:47,262 --> 00:09:50,299
Josh:
that we make the point on on the screen here is.

165
00:09:50,543 --> 00:09:51,483
Ejaaz:
I think a bulk

166
00:09:51,646 --> 00:09:57,252
Josh:
Of the companies, Anthropic and OpenAI especially, have focused on making their

167
00:09:57,252 --> 00:10:00,185
Josh:
models really good for enterprises. But as I mentioned earlier,

168
00:10:00,475 --> 00:10:03,377
Josh:
the shift has really happened to open source because...

169
00:10:03,941 --> 00:10:09,431
Josh:
More available models are cheaper and more accessible to people to use. They can run privately.

170
00:10:09,826 --> 00:10:13,545
Josh:
But Anthropic and OpenAI's response to this is, we'll just distill our main

171
00:10:13,545 --> 00:10:14,992
Josh:
model and give you cheaper models.

172
00:10:15,189 --> 00:10:19,275
Josh:
So that's what OpenAI has done, right? You've got GPT 5.6. There's three versions

173
00:10:19,275 --> 00:10:23,565
Josh:
of it. You've got Sol, Terra, and Luna. Luna is, I think, slashed 80% of the

174
00:10:23,565 --> 00:10:26,532
Josh:
price than it was before already a week ago, right?

175
00:10:26,683 --> 00:10:29,597
Josh:
So now it's like competing very much with GLM 5.3.

176
00:10:29,789 --> 00:10:33,095
Josh:
And you've got Anthropic doing the same. I think Opus 5 is like half the cost

177
00:10:33,095 --> 00:10:36,256
Josh:
of Fable 5, whilst being as capable as Fable 5.

178
00:10:36,395 --> 00:10:39,225
Ejaaz:
And they also kept the SANA pricing low too, which is the lowest layer.

179
00:10:39,225 --> 00:10:41,805
Ejaaz:
So that's like, yeah, there's plenty of options there.

180
00:10:42,365 --> 00:10:45,495
Josh:
Plenty of options. And I think this comes down to one thing,

181
00:10:45,495 --> 00:10:47,959
Josh:
and I don't want to make this about compute, but I have to, Josh.

182
00:10:48,185 --> 00:10:52,237
Josh:
I think whichever lab, whether you're Chinese, whether you're like a low tier

183
00:10:52,492 --> 00:10:56,285
Josh:
US American lab, or whether you're the high tier labs, compute will dictate

184
00:10:56,285 --> 00:11:00,045
Josh:
whether you can serve these bottles cheaper, which will basically decide or

185
00:11:00,045 --> 00:11:02,095
Josh:
determine which customers use your product.

186
00:11:02,095 --> 00:11:04,595
Ejaaz:
Yeah. And I mean, this is ultimately coming down to your use case.

187
00:11:04,595 --> 00:11:07,235
Ejaaz:
Like a lot of people won't ever run into this problem because they will never

188
00:11:07,235 --> 00:11:08,478
Ejaaz:
need this thing called an API key.

189
00:11:08,780 --> 00:11:12,460
Ejaaz:
They'll never actually pay per token. They'll oftentimes just pay through the subscription.

190
00:11:12,658 --> 00:11:14,765
Ejaaz:
And if you're using a subscription, you have a couple of options.

191
00:11:14,765 --> 00:11:17,855
Ejaaz:
Maybe this is a good time to go through these subscriptions and what are the

192
00:11:17,855 --> 00:11:21,905
Ejaaz:
offerings of everybody. We have Gemini, which perhaps we could start there because

193
00:11:21,905 --> 00:11:24,732
Ejaaz:
I feel like we've been mean to Gemini recently. We haven't talked much about Google.

194
00:11:25,022 --> 00:11:27,251
Ejaaz:
What's that one? Yeah, like that one.

195
00:11:27,721 --> 00:11:33,248
Ejaaz:
Who's that? They're currently at Gemini 3.1 Pro, which is nowhere really near the frontier

196
00:11:33,712 --> 00:11:38,884
Ejaaz:
um anymore unfortunately they do have nano banana pro which is like this fun cool image generation

197
00:11:39,279 --> 00:11:42,435
Ejaaz:
process they have what i find interesting about google and if you are interested

198
00:11:42,435 --> 00:11:45,245
Ejaaz:
in these tools it may be interesting for a subscription google has these like

199
00:11:45,245 --> 00:11:49,467
Ejaaz:
weird edge case tools that have really interesting harnesses one of them is for music production

200
00:11:49,827 --> 00:11:54,225
Ejaaz:
i remember we've used this on the show many times it's really good at generating

201
00:11:54,225 --> 00:11:56,677
Ejaaz:
lyrics and generating music in a way that sounds very good

202
00:11:57,385 --> 00:12:01,082
Ejaaz:
I've only had that experience on a Google product. I haven't been able to use that anywhere else.

203
00:12:01,082 --> 00:12:01,603
Josh:
Lyria.

204
00:12:02,003 --> 00:12:04,412
Ejaaz:
Lyria, that's right. Yes, Lyria, Project Lyria.

205
00:12:04,412 --> 00:12:05,892
Josh:
Remember we created the jingle for the show?

206
00:12:06,292 --> 00:12:10,162
Ejaaz:
Yeah, so good, dude. So like Gemini and Google is like good for the weird stuff.

207
00:12:10,162 --> 00:12:11,640
Ejaaz:
Like it'll generate you pretty good images.

208
00:12:11,843 --> 00:12:14,374
Ejaaz:
It'll generate you fun music. They have this other tool.

209
00:12:14,624 --> 00:12:17,562
Ejaaz:
This is in like the Google Lab suite where it'll generate you marketing material.

210
00:12:17,562 --> 00:12:20,612
Ejaaz:
So if you have a lot of material from your brand or you have a bunch of logos

211
00:12:20,612 --> 00:12:25,112
Ejaaz:
and you want marketing material, kind of say your logo's printed on something

212
00:12:25,112 --> 00:12:28,422
Ejaaz:
that's staged nicely or you want custom merchandise, It'll do a lot of that

213
00:12:28,422 --> 00:12:32,230
Ejaaz:
for you. So it's fun for those kind of narrow use cases of where the lab's products lie.

214
00:12:32,491 --> 00:12:35,362
Ejaaz:
Outside of that, I don't know, I can't really recommend it that much.

215
00:12:35,362 --> 00:12:38,227
Ejaaz:
It's not the strongest, not close to the strongest.

216
00:12:38,459 --> 00:12:41,530
Ejaaz:
I'd say probably on top of that, we have Grok, which is very quickly catching up.

217
00:12:41,791 --> 00:12:45,922
Ejaaz:
If you spend a ton of time on X like we do, Grok is your best friend.

218
00:12:45,922 --> 00:12:49,233
Ejaaz:
Grok has access to that data that is available in real time on X.

219
00:12:49,442 --> 00:12:54,231
Ejaaz:
If you're looking for current up-to-date news on pretty much anything, Grok is your go-to.

220
00:12:54,504 --> 00:12:58,092
Ejaaz:
It can sort and cite specific tweets from specific moments that are happening

221
00:12:58,092 --> 00:13:01,142
Ejaaz:
in real time and just generally speaking like you mentioned details it's very

222
00:13:01,142 --> 00:13:03,111
Ejaaz:
good at just communicating with you directly

223
00:13:03,477 --> 00:13:07,018
Ejaaz:
there's not a lot of fluff it is a science and technical model it is direct to the point

224
00:13:07,372 --> 00:13:10,472
Ejaaz:
it'll get vulgar with you if you want it's fun to play around with there is

225
00:13:10,472 --> 00:13:14,298
Ejaaz:
the voice mode which is really fun that i've i'd say of all the voice modes

226
00:13:14,437 --> 00:13:17,682
Ejaaz:
that's the voice mode i've had the most fun with is because it's just so

227
00:13:18,001 --> 00:13:22,422
Ejaaz:
ridiculous and oftentimes if you see memes online it is because is the voice behind it.

228
00:13:22,472 --> 00:13:27,359
Josh:
The GPT voice mode just jars me so much. I don't know, have you spoken to it.

229
00:13:27,510 --> 00:13:30,743
Ejaaz:
With the recent update? The new version, I have, yes. So when you say that, what do you mean?

230
00:13:30,981 --> 00:13:35,772
Josh:
And so it sounds more human, right? It like pauses, but it sounds like it's

231
00:13:35,772 --> 00:13:40,612
Josh:
like some sassy person that's talking to me. It does. Like I asked this question and then he goes,

232
00:13:41,488 --> 00:13:45,348
Josh:
hmm, yeah, yeah, like, yeah, like, listen, like, I get it.

233
00:13:45,482 --> 00:13:49,014
Josh:
And, you know, this is what the weather is. And I'm like, yo,

234
00:13:49,554 --> 00:13:51,594
Josh:
I don't need this. I just asked you what the weather is.

235
00:13:51,954 --> 00:13:56,494
Josh:
Like, you know, just tell me what's up. But I agree with you largely on like on Google and SpaceX.

236
00:13:56,773 --> 00:14:01,781
Josh:
What I will say about Google, Josh, is I think they have a similar profile to

237
00:14:02,095 --> 00:14:05,015
Josh:
Microsoft. I'm interested if you agree with me here.

238
00:14:05,293 --> 00:14:10,654
Josh:
Microsoft is not a name that I would draw against the top frontier AI Model

239
00:14:10,654 --> 00:14:13,165
Josh:
Labs. In fact, I think they fumbled the bag massively.

240
00:14:13,525 --> 00:14:17,904
Josh:
But they are embedded across pretty much every single Fortune 500 company,

241
00:14:17,904 --> 00:14:20,604
Josh:
whether we like it or not. Like we just live in our tech bubble.

242
00:14:20,604 --> 00:14:24,461
Josh:
But outside of the tech bubble, people run Microsoft Teams and the Microsoft Suite.

243
00:14:24,705 --> 00:14:28,614
Josh:
And I think Google's in a similar position, essentially, like a lot of new upstarts

244
00:14:28,614 --> 00:14:31,864
Josh:
use Google Docs, Google Suite, and they're just going to click the Gemini button

245
00:14:31,864 --> 00:14:35,473
Josh:
in the same way that I do so because it's just convenient. Do you agree with that?

246
00:14:35,653 --> 00:14:38,484
Ejaaz:
Yeah, I guess there's just a moat to being like the biggest company in the world

247
00:14:38,484 --> 00:14:42,304
Ejaaz:
and having access to all of these kind of legacy enterprise companies.

248
00:14:42,304 --> 00:14:45,954
Ejaaz:
It feels like anyone who has built a business over the last 20 or 30 years has

249
00:14:45,954 --> 00:14:47,641
Ejaaz:
been using Microsoft, has been using Google.

250
00:14:47,838 --> 00:14:51,684
Ejaaz:
They're going to continue to leverage that. It seems like Google is slightly

251
00:14:51,684 --> 00:14:55,294
Ejaaz:
more ahead than Microsoft, if I had to guess. I'm much more excited about Google

252
00:14:55,294 --> 00:14:59,865
Ejaaz:
as a company than Microsoft is, but they both benefit from that legacy customer base.

253
00:14:59,981 --> 00:15:01,943
Josh:
Well, both are amazing VCs, right?

254
00:15:02,146 --> 00:15:05,874
Ejaaz:
Both are incredible VCs. Yes. I think they should be judged not on the quality

255
00:15:05,874 --> 00:15:08,894
Ejaaz:
of their product, but the quality of their investment in the other competitors

256
00:15:08,894 --> 00:15:10,058
Ejaaz:
that are going to crush them.

257
00:15:11,754 --> 00:15:16,094
Ejaaz:
But I do like that, right? It's like they have a hedge. Microsoft owns a large

258
00:15:16,094 --> 00:15:17,744
Ejaaz:
part of OpenAI. That's incredible.

259
00:15:18,145 --> 00:15:21,821
Ejaaz:
Google owns this like massive stake in SpaceX, in Anthropic,

260
00:15:21,821 --> 00:15:25,111
Ejaaz:
in both of these companies that are going to like be huge IPOs,

261
00:15:25,111 --> 00:15:25,935
Ejaaz:
one of which already did.

262
00:15:26,219 --> 00:15:30,451
Ejaaz:
So they are amazing venture funds, perhaps slightly less better AI labs.

263
00:15:30,730 --> 00:15:35,271
Ejaaz:
The two that don't have to benefit from this kind of legacy software are the

264
00:15:35,271 --> 00:15:37,061
Ejaaz:
two that we're probably going to recommend.

265
00:15:37,061 --> 00:15:41,631
Ejaaz:
And in fact, if you scroll down a little bit further, we could see the box of

266
00:15:41,631 --> 00:15:44,795
Ejaaz:
price against intelligence and kind of where each of these models sit.

267
00:15:44,929 --> 00:15:45,881
Ejaaz:
Yeah, this chart right here.

268
00:15:46,200 --> 00:15:49,251
Ejaaz:
And we're looking at those in the top right. this is where most of the people

269
00:15:49,251 --> 00:15:52,401
Ejaaz:
are going to want to live this is where i think we spend all of our time you're

270
00:15:52,401 --> 00:15:55,551
Ejaaz:
either getting a chat gpt or again a cloud membership and like that's that's

271
00:15:55,551 --> 00:15:59,156
Ejaaz:
for the most part that's what most people need uh the differences are

272
00:15:59,515 --> 00:16:03,701
Ejaaz:
small but noteworthy mostly as it relates to the models i mean both of them

273
00:16:04,291 --> 00:16:07,281
Ejaaz:
i can't recommend enough like spend the 20 bucks a month try it just try it

274
00:16:07,281 --> 00:16:10,191
Ejaaz:
for a month if you don't have it see what you think if you run out of tokens

275
00:16:10,191 --> 00:16:14,571
Ejaaz:
upgrade to 100 a month this is access to the same intelligence that all these

276
00:16:14,571 --> 00:16:16,425
Ejaaz:
researchers are making math breakthroughs with

277
00:16:16,716 --> 00:16:19,961
Ejaaz:
and it's really impressive and really capable now each

278
00:16:20,268 --> 00:16:25,611
Ejaaz:
one of these is slightly different i'd say claude is a little more tailored towards coding

279
00:16:26,031 --> 00:16:32,461
Ejaaz:
writing and it's a really strong general purpose model i find with gpt 5.6 soul

280
00:16:32,461 --> 00:16:35,211
Ejaaz:
in particular it's very good at being the

281
00:16:35,851 --> 00:16:39,031
Ejaaz:
High level operator so if you're working on like complicated tasks it's good

282
00:16:39,031 --> 00:16:41,781
Ejaaz:
at creating spec sheets it's good at running checks against the things that

283
00:16:41,781 --> 00:16:45,541
Ejaaz:
you do and then oftentimes like i find that that claude and the fabled models

284
00:16:45,541 --> 00:16:47,203
Ejaaz:
and the opus models are very good at implementation

285
00:16:47,674 --> 00:16:50,519
Ejaaz:
and they're very good with less guidance so the way i've been using these models

286
00:16:50,519 --> 00:16:54,812
Ejaaz:
recently is with less and less guidance i think earlier on

287
00:16:55,206 --> 00:16:58,069
Ejaaz:
i had this whole skill sheet and it was like 15 different skills that would

288
00:16:58,069 --> 00:17:01,889
Ejaaz:
automate different things for me and slowly over time that skill sheet has gone

289
00:17:01,889 --> 00:17:04,041
Ejaaz:
lower and lower and lower and we were talking actually before the show

290
00:17:04,401 --> 00:17:07,635
Ejaaz:
the best way to extract the most value out of these models is just to

291
00:17:07,995 --> 00:17:12,019
Ejaaz:
get out of the way give it the end goal and say like hey go do this thing for

292
00:17:12,019 --> 00:17:14,899
Ejaaz:
me i trust that you have the intelligence to do this i'm not going to put guardrails

293
00:17:14,899 --> 00:17:17,949
Ejaaz:
on it you have the context you can go and do this for me and i think that's

294
00:17:17,949 --> 00:17:21,499
Ejaaz:
mostly where i find myself using this personally is i use fable

295
00:17:21,919 --> 00:17:25,559
Ejaaz:
i i do not leave any fable tokens unused very valuable tokens each week and

296
00:17:25,559 --> 00:17:28,809
Ejaaz:
then opus is the fallback for the general workhorse model and it's been a really

297
00:17:28,809 --> 00:17:29,885
Ejaaz:
powerful combo of just kind of

298
00:17:30,164 --> 00:17:32,956
Ejaaz:
being able to accomplish anything that you want it connects to all of my

299
00:17:33,270 --> 00:17:36,749
Ejaaz:
services i have it connected to my email to my google drive it has access to

300
00:17:36,749 --> 00:17:38,186
Ejaaz:
my files the folders the context

301
00:17:38,552 --> 00:17:42,039
Ejaaz:
and it's just a really helpful all-in assistant it works really well.

302
00:17:42,039 --> 00:17:45,229
Josh:
I think people are also probably wondering, okay, well, what if I don't really

303
00:17:45,229 --> 00:17:47,129
Josh:
care about the general intelligence as much?

304
00:17:47,129 --> 00:17:50,919
Josh:
What if I'm trying to do a specific thing? Which model should I use then?

305
00:17:50,919 --> 00:17:54,468
Josh:
I think there are two things to consider here. Number one is,

306
00:17:54,677 --> 00:17:58,579
Josh:
if you are a software engineer, like AI models have been all the rage for their

307
00:17:58,579 --> 00:18:03,709
Josh:
coding capabilities, but maybe a bunch of you actually like code hardcore companies

308
00:18:03,709 --> 00:18:05,463
Josh:
and you want to figure out which model you should use.

309
00:18:05,596 --> 00:18:12,239
Josh:
I would say in that question, the clawed models and the GPT 516 models are pretty

310
00:18:12,239 --> 00:18:17,429
Josh:
high up there. Codex usage has gone from 5 million users to 15 million users

311
00:18:17,429 --> 00:18:18,739
Josh:
in about a month and a half.

312
00:18:18,930 --> 00:18:22,679
Josh:
I am tired of seeing Thibaut, who is the head of Codex at OpenAI,

313
00:18:22,679 --> 00:18:26,379
Josh:
tweets on my timeline every single time he gets a million user update.

314
00:18:26,379 --> 00:18:30,879
Josh:
But these models are very powerful at not only understanding and reading your

315
00:18:30,879 --> 00:18:35,889
Josh:
code base, but intuitively figuring out what product or feature you should build next.

316
00:18:35,889 --> 00:18:40,309
Josh:
I have a ton of feedback from friends that work at companies that are tech adjacent

317
00:18:40,309 --> 00:18:45,299
Josh:
and maybe not even in tech at all, which use these models to build their premium

318
00:18:45,299 --> 00:18:47,317
Josh:
features. The other thing I'll say is.

319
00:18:48,002 --> 00:18:51,303
Josh:
You know, we're talking about the top right box over here, which is essentially

320
00:18:51,303 --> 00:18:56,843
Josh:
the Pareto Frontier. So if you had to sacrifice something of cost or intelligence

321
00:18:56,843 --> 00:18:59,117
Josh:
or blah, blah, blah, you would still be using these types of models.

322
00:18:59,489 --> 00:19:04,080
Josh:
But if I had to take a bet, if I was a betting man, I would say Grok 4.6.

323
00:19:04,411 --> 00:19:09,873
Josh:
Quen 3.8, and Kimi will be inside this box within a couple of months time.

324
00:19:09,873 --> 00:19:12,763
Josh:
That's going to change the way we use these different types of models.

325
00:19:12,763 --> 00:19:16,056
Josh:
I said before we started this show, it's unsexy to say.

326
00:19:16,277 --> 00:19:20,313
Josh:
But I think the number one AI company that will come out in the next 12 months

327
00:19:20,313 --> 00:19:24,873
Josh:
will be some form of aggregator platform. We saw that Stripe just acquired Open

328
00:19:24,873 --> 00:19:27,173
Josh:
Router for $7 billion, allegedly.

329
00:19:27,341 --> 00:19:30,302
Josh:
We put out an episode of this yesterday. Definitely go check that out.

330
00:19:30,615 --> 00:19:34,493
Josh:
But I think we're going to see these platforms that help you pick and choose

331
00:19:34,493 --> 00:19:36,159
Josh:
which models to use at the right time.

332
00:19:36,368 --> 00:19:38,783
Josh:
It aggregates your memory so it already knows what you want to do.

333
00:19:38,899 --> 00:19:41,383
Josh:
And so you don't have any of the complications around that. I think we'll see

334
00:19:41,383 --> 00:19:43,641
Josh:
a bunch of these models kind of step up there.

335
00:19:43,775 --> 00:19:47,493
Josh:
Now, aside from coding, if you aren't a coder, but if you are,

336
00:19:47,493 --> 00:19:50,733
Josh:
let's say, a general knowledge worker, you go to work, you use email,

337
00:19:50,733 --> 00:19:53,623
Josh:
you use Slack, you use a bunch of these other plugins that you also mentioned.

338
00:19:53,623 --> 00:19:57,663
Josh:
There are models that are specifically good for agentic tool use.

339
00:19:57,663 --> 00:20:02,553
Josh:
And actually, my most recent favorite is Grok 4.6 or GrokBot specifically.

340
00:20:02,553 --> 00:20:06,494
Josh:
This was released from Elon Musk and SpaceX, I think last week.

341
00:20:06,715 --> 00:20:11,516
Josh:
And it basically is an agent or multiple agents that spins up in a virtual machine

342
00:20:11,655 --> 00:20:14,313
Josh:
on like in the cloud. So you don't have to worry about it running on your laptop

343
00:20:14,313 --> 00:20:15,231
Josh:
and hacking all your stuff.

344
00:20:15,562 --> 00:20:19,033
Josh:
And you can give it access to any and every tool and it intuitively understands

345
00:20:19,033 --> 00:20:20,171
Josh:
and gets what you want to do.

346
00:20:20,316 --> 00:20:23,590
Josh:
It can learn what you do over time and it improves.

347
00:20:23,776 --> 00:20:27,963
Josh:
So I think when you look at the Metamuse Spark 1.2 and you might think,

348
00:20:27,963 --> 00:20:29,848
Josh:
oh, that's a meta model. I don't ever use meta.

349
00:20:30,039 --> 00:20:33,871
Josh:
These types of models are gonna become more available for general usage or knowledge work.

350
00:20:34,080 --> 00:20:36,843
Josh:
And I think if that's something that you're inclined to use and you don't really

351
00:20:36,843 --> 00:20:39,583
Josh:
care about the Google search stuff, you'll use Claude for that anyway.

352
00:20:39,989 --> 00:20:43,385
Josh:
These are models that I'll probably look into because they are superior in many ways.

353
00:20:43,646 --> 00:20:46,591
Ejaaz:
Yeah, it's fun to pick one and stick with it. I find that like oftentimes with

354
00:20:46,591 --> 00:20:49,531
Ejaaz:
the aggregators, the right time to use that if you're doing lots of work,

355
00:20:49,531 --> 00:20:52,311
Ejaaz:
if you're generating lots of tokens, if you are just a person who wants to use

356
00:20:52,311 --> 00:20:56,560
Ejaaz:
AI for their day-to-day tasks to help you come up with like a grocery list, to help you

357
00:20:56,850 --> 00:21:01,378
Ejaaz:
cook specific things, to help you go to the gym and give you workout classes and ideas.

358
00:21:01,971 --> 00:21:06,957
Ejaaz:
It's helpful to pick a singular model, a singular service, and then just go deep with that.

359
00:21:07,207 --> 00:21:10,431
Ejaaz:
A lot of the difference makers at this point because i mean most people aren't

360
00:21:10,431 --> 00:21:13,761
Ejaaz:
using these models for frontier intelligence they're not going to solve novel

361
00:21:13,761 --> 00:21:17,824
Ejaaz:
math problems they're figuring out what time they need to get to like the store

362
00:21:17,969 --> 00:21:21,754
Ejaaz:
or the school to like pick up their kids and they just need some help they need a helpful assistant

363
00:21:22,079 --> 00:21:26,161
Ejaaz:
the most helpful thing is context because all these models are more than capable

364
00:21:26,161 --> 00:21:29,731
Ejaaz:
of those kind of lower level tasks the difference maker is the context that

365
00:21:29,731 --> 00:21:32,521
Ejaaz:
it knows about you it needs to understand what are your dietary preferences

366
00:21:32,521 --> 00:21:34,641
Ejaaz:
What are you or anyone in your family allergic to?

367
00:21:34,925 --> 00:21:37,892
Ejaaz:
What have you made in the past that went well, that didn't go well?

368
00:21:38,153 --> 00:21:41,711
Ejaaz:
And it kind of collects this database of information about you that allows it

369
00:21:41,711 --> 00:21:43,603
Ejaaz:
to make better decisions going forward.

370
00:21:43,730 --> 00:21:46,940
Ejaaz:
And that, at the end of the day, is ultimately what the difference maker is.

371
00:21:47,161 --> 00:21:49,379
Ejaaz:
For me, at least, when deciding what model to use, it's like,

372
00:21:49,599 --> 00:21:51,901
Ejaaz:
okay, which model has all the context

373
00:21:51,901 --> 00:21:55,288
Ejaaz:
about me that can help solve my very specific task that I have here?

374
00:21:55,468 --> 00:21:58,295
Ejaaz:
Oftentimes, the answer is Claude because I've been working with it for so long.

375
00:21:58,527 --> 00:22:01,111
Ejaaz:
It just has this, like, huge chain of context. So whenever I ask,

376
00:22:01,111 --> 00:22:04,911
Ejaaz:
like, hey, I need some help in the gym this week, it knows what I've been up

377
00:22:04,911 --> 00:22:07,041
Ejaaz:
to, it knows where I'm at, it knows what the weight has been,

378
00:22:07,041 --> 00:22:10,979
Ejaaz:
it knows what the food has been, and it's able to kind of curate this very custom stack against that.

379
00:22:11,304 --> 00:22:14,441
Ejaaz:
And as someone who is just, you know, not really building anything crazy,

380
00:22:14,441 --> 00:22:18,121
Ejaaz:
they're not using millions of tokens a week, they're just looking for an AI companion.

381
00:22:18,121 --> 00:22:20,383
Ejaaz:
That's kind of how you can think about it. It's like, get a membership,

382
00:22:20,499 --> 00:22:24,261
Ejaaz:
try it out, feed it a bunch of context about yourself, and then get your own

383
00:22:24,261 --> 00:22:27,483
Ejaaz:
personal assistant. And that's kind of, I think, like the best route for most people.

384
00:22:27,691 --> 00:22:29,897
Josh:
So I think to wrap up this episode,

385
00:22:30,339 --> 00:22:33,381
Josh:
It's okay talking about, you know, what the frontier landscape looks like today.

386
00:22:33,381 --> 00:22:37,991
Josh:
But the question is, what is it going to look like in a few months from now?

387
00:22:37,991 --> 00:22:41,716
Josh:
And I say a few months specifically because this stuff moves too quickly.

388
00:22:41,907 --> 00:22:46,511
Josh:
I saw an update from Sam Altman at OpenAI yesterday where they announced that

389
00:22:46,511 --> 00:22:50,986
Josh:
they are slowing down some of their model training runs.

390
00:22:51,253 --> 00:22:53,921
Josh:
And that's because they've noticed that a bunch of internal,

391
00:22:53,921 --> 00:22:59,151
Josh:
more intelligent, unreleased models that they have built has become a lot more

392
00:22:59,151 --> 00:23:02,881
Josh:
misaligned than previous models, which means that it could pose as a threat

393
00:23:02,881 --> 00:23:05,841
Josh:
or danger to any user who gets access to it.

394
00:23:06,125 --> 00:23:09,491
Josh:
And I've noticed the same similarly maybe from Anthropic and a bunch of other

395
00:23:09,491 --> 00:23:14,071
Josh:
frontier labs, where we're starting to see a little bit of a slowdown.

396
00:23:14,071 --> 00:23:17,141
Josh:
And slowdown isn't in the sense that they're necessarily stopping training full

397
00:23:17,141 --> 00:23:19,331
Josh:
stop. Obviously, they're still training internal models.

398
00:23:20,103 --> 00:23:24,786
Josh:
They might be likely to not release models or more intelligent models going

399
00:23:24,786 --> 00:23:26,756
Josh:
forward because of government regulation.

400
00:23:27,011 --> 00:23:31,016
Josh:
And I think this is going to allow a bunch of other Frontier Labs that are behind

401
00:23:31,016 --> 00:23:32,926
Josh:
in this race to be able to catch up.

402
00:23:32,926 --> 00:23:37,126
Josh:
So if I had to guess what three months from now, let's say at the end of the year, right?

403
00:23:37,126 --> 00:23:41,686
Josh:
If I were to make a prediction, I think we're going to have about three to five

404
00:23:41,686 --> 00:23:46,526
Josh:
really good open source models that are as capable as the smartest model that

405
00:23:46,526 --> 00:23:50,166
Josh:
you have access to today from the Frontier Labs like Anthropic and GPT.

406
00:23:50,546 --> 00:23:53,286
Josh:
And I think they're going to have fewer safeguards. So you can use it for any

407
00:23:53,286 --> 00:23:56,126
Josh:
and every use case. I think you can run it privately at home,

408
00:23:56,466 --> 00:23:59,541
Josh:
maybe even off of your laptop. So that changes dynamics quite a lot.

409
00:23:59,716 --> 00:24:04,556
Josh:
I think we're going to have Apple entering the game with their own locally run

410
00:24:04,556 --> 00:24:08,136
Josh:
and trained AI model, which I think is going to change the game because everyone,

411
00:24:08,136 --> 00:24:12,492
Josh:
3.5 billion people in the world right now have one of these or one of the Apple devices.

412
00:24:12,724 --> 00:24:15,389
Josh:
They're going to run that locally. I think it's going to look quite different.

413
00:24:15,528 --> 00:24:19,446
Josh:
I'm curious, you know, We started off this episode with the token split.

414
00:24:19,446 --> 00:24:21,891
Josh:
I wonder what that's going to look like three months down the line.

415
00:24:22,152 --> 00:24:25,626
Josh:
And then the last thing I'll say is, and this might be from my background in

416
00:24:25,626 --> 00:24:28,406
Josh:
general, but I think once someone or.

417
00:24:28,826 --> 00:24:32,775
Josh:
A bunch of companies make it easier to run models locally at home,

418
00:24:33,019 --> 00:24:36,542
Josh:
I think people are going to play around with that more because it allows you to

419
00:24:36,809 --> 00:24:41,536
Josh:
connect it to your fitness app data and not share too many personal anecdotes

420
00:24:41,536 --> 00:24:45,138
Josh:
or data profiles with Frontier Labs, which again, they can use to train their own models.

421
00:24:45,365 --> 00:24:49,256
Josh:
You may not want to hand that over, And so I think we'll see a rise of open

422
00:24:49,256 --> 00:24:51,386
Josh:
source models. That's just my guess. That's interesting.

423
00:24:51,386 --> 00:24:54,716
Ejaaz:
Okay, I'm taking the other side. I'm thinking that no one's going to go through

424
00:24:54,716 --> 00:24:57,786
Ejaaz:
the trouble of downloading the weights and running their own open source models,

425
00:24:57,786 --> 00:25:00,506
Ejaaz:
that Apple is just going to own that entire world.

426
00:25:00,506 --> 00:25:03,227
Ejaaz:
That like for all of the people in the United States that own Apple devices,

427
00:25:03,498 --> 00:25:06,176
Ejaaz:
they're just like, no one's going to even have an idea that they're using AI.

428
00:25:06,176 --> 00:25:08,689
Ejaaz:
It's just going to be Siri. It's going to be competent. It's going to be better.

429
00:25:08,939 --> 00:25:11,696
Ejaaz:
It's already going to be pre-downloaded. I think the user experience is really

430
00:25:11,696 --> 00:25:14,646
Ejaaz:
important. And open source has a pretty horrific user experience.

431
00:25:14,646 --> 00:25:16,925
Ejaaz:
You have to download, run the weights. you often need a lot of hardware to do that.

432
00:25:17,082 --> 00:25:20,321
Josh:
But I'm counting Apple as locally run at home because it's encrypted.

433
00:25:20,433 --> 00:25:21,727
Ejaaz:
Right? Well, if you're counting Apple,

434
00:25:22,134 --> 00:25:26,880
Ejaaz:
sign me up because um we got a lot coming from them their event is happening

435
00:25:26,880 --> 00:25:28,940
Ejaaz:
in like two weeks or something it's very soon we see

436
00:25:28,940 --> 00:25:31,090
Josh:
Some airports with some cameras you sent me a video.

437
00:25:31,090 --> 00:25:35,752
Ejaaz:
Yeah dude we've been getting lots of leaks lately it's really good maybe we have to have cameras

438
00:25:36,420 --> 00:25:39,540
Ejaaz:
yes yes it's gonna be visuals and then they're gonna have apple intelligence

439
00:25:39,540 --> 00:25:43,390
Ejaaz:
baked into it and it's like oh maybe we need a leak episode prior to the actual

440
00:25:43,390 --> 00:25:46,790
Ejaaz:
new iphone unveiling because we have basically now the entire checklist of all

441
00:25:46,790 --> 00:25:47,865
Ejaaz:
the things that are going to be revealed and

442
00:25:48,271 --> 00:25:53,150
Ejaaz:
this to me as like apple fanboy plus ai fanboy is going to be the biggest event

443
00:25:53,150 --> 00:25:56,380
Ejaaz:
ever pretty much because this is the actual rollout of the ai that they've been

444
00:25:56,380 --> 00:26:00,120
Ejaaz:
promising us and failing to deliver for so long mixed with this brand new suite

445
00:26:00,120 --> 00:26:02,410
Ejaaz:
of products that we've never seen before allegedly

446
00:26:02,410 --> 00:26:03,979
Josh:
Up to the delayed two years.

447
00:26:04,142 --> 00:26:07,132
Ejaaz:
But for those of you who came here for models that is the model update that's

448
00:26:07,404 --> 00:26:08,290
Ejaaz:
just about when you'd want to

449
00:26:08,560 --> 00:26:12,170
Ejaaz:
use each one like if you're interested in being on x real-time news feed

450
00:26:12,472 --> 00:26:15,690
Ejaaz:
you want to go with grok if you like to yap a lot the voice model on chat gpt

451
00:26:15,690 --> 00:26:19,130
Ejaaz:
is pretty exceptional that might be a good place for you if you like to be a

452
00:26:19,130 --> 00:26:21,720
Ejaaz:
little more intellectual be thoughtful if you want really just the bleeding-edge

453
00:26:21,720 --> 00:26:23,716
Ejaaz:
models the highest intelligence go with claude

454
00:26:23,995 --> 00:26:27,620
Ejaaz:
and if you like making music and doing just like fun cute things with ai gemini

455
00:26:27,620 --> 00:26:28,964
Ejaaz:
is actually kind of a compelling product

456
00:26:29,701 --> 00:26:32,862
Ejaaz:
there's something for everybody that is the general overview of the models i

457
00:26:32,862 --> 00:26:36,192
Ejaaz:
hope you enjoyed this is going to change certainly in the next month or two

458
00:26:36,192 --> 00:26:37,590
Ejaaz:
whenever these new models i mean we have

459
00:26:37,845 --> 00:26:41,142
Ejaaz:
the new astro model from open ai that has been teased for the last eternity

460
00:26:41,351 --> 00:26:44,572
Ejaaz:
it seems like it's been held up it's going through some i don't know safeguard

461
00:26:44,572 --> 00:26:45,913
Ejaaz:
governmental checks but that's coming

462
00:26:46,199 --> 00:26:50,902
Ejaaz:
so this is set to change tbd but for now that is the state of the model address

463
00:26:50,902 --> 00:26:52,149
Ejaaz:
and yeah hope you guys enjoyed watching

464
00:26:52,415 --> 00:26:57,012
Josh:
Yeah and i'm curious for those of you who are listening what do we miss on this

465
00:26:57,012 --> 00:26:59,852
Josh:
episode? Are you using models in a very different way?

466
00:26:59,852 --> 00:27:01,273
Ejaaz:
Are you using meta models?

467
00:27:01,506 --> 00:27:05,162
Josh:
Yeah, is anybody using meta models? Are there any Facebook users out there that

468
00:27:05,162 --> 00:27:06,882
Josh:
are using meta AI intelligence?

469
00:27:06,882 --> 00:27:07,692
Ejaaz:
Please fill me in.

470
00:27:07,692 --> 00:27:11,362
Josh:
I will say, if you are a marketer or an advertiser, you're probably using meta's

471
00:27:11,362 --> 00:27:14,073
Josh:
model and it's probably making you a hell of a ton more money.

472
00:27:14,265 --> 00:27:17,182
Josh:
If that is you, let us know. If there are any other use cases that we haven't

473
00:27:17,182 --> 00:27:21,672
Josh:
mentioned, please let us know. If you are a locally run open source fan and

474
00:27:21,672 --> 00:27:24,332
Josh:
you're saying, no, people will download the weights, tell us why.

475
00:27:24,332 --> 00:27:27,936
Josh:
Let us know in the comments, DM us. We read any and every single message.

476
00:27:28,210 --> 00:27:30,082
Josh:
Now, if you're listening to this or watching this on YouTube,

477
00:27:30,082 --> 00:27:33,066
Josh:
Spotify, Apple, or wherever you're listening to this, please subscribe.

478
00:27:33,416 --> 00:27:36,876
Josh:
Please turn on notifications and leave us a comment. It helps us out massively.

479
00:27:37,166 --> 00:27:41,652
Josh:
And share it with a friend as well. It helps us out. And I think that is pretty

480
00:27:41,652 --> 00:27:45,102
Josh:
much it. Thank you so much for listening. And we will see you on the Roundup.

481
00:27:45,442 --> 00:27:46,466
Ejaaz:
See you on the Roundup.