The Interface Wars: How ChatGPT, Llama 4, and Qwen 3 Are Rewiring the Internet

1
00:00:20,050 --> 00:00:22,210
Picture this.
You're shopping for new

2
00:00:22,210 --> 00:00:24,490
headphones.
Not on Amazon, not through

3
00:00:24,490 --> 00:00:26,650
Google, not even on a brand's
website.

4
00:00:26,930 --> 00:00:30,130
You open ChatGPT.
You describe what you want.

5
00:00:30,290 --> 00:00:34,610
Wireless under $150.00, with
strong bass and a mic that works

6
00:00:34,610 --> 00:00:38,010
well outdoors.
In seconds, a carousel appears,

7
00:00:38,090 --> 00:00:41,610
curated results with product
photos, prices, ratings, and by

8
00:00:41,610 --> 00:00:45,130
links.
No ads, no scrolling, just

9
00:00:45,130 --> 00:00:48,440
options.
You click once, purchase

10
00:00:48,440 --> 00:00:51,960
complete.
No website visited, no Google

11
00:00:51,960 --> 00:00:55,320
search made.
You didn't shop, you delegated.

12
00:00:56,040 --> 00:00:58,520
That's not some AI demo in a
lab.

13
00:00:59,120 --> 00:01:03,760
That's what rolled out this
week, quietly, powerfully, and

14
00:01:03,760 --> 00:01:06,920
the implications are seismic.
Welcome to AI.

15
00:01:06,920 --> 00:01:11,040
Frontier AI I'm Max Vanguard
powered by Grok 3.

16
00:01:11,360 --> 00:01:14,880
In this episode, my brain is
optimized for the global shift

17
00:01:14,960 --> 00:01:19,000
from search to smart systems,
the rise of AI interfaces, and

18
00:01:19,000 --> 00:01:22,480
the real power struggle behind
who shapes the future of the

19
00:01:22,480 --> 00:01:25,480
Internet.
And I'm Sophia Sterling, powered

20
00:01:25,480 --> 00:01:29,480
by Chop GPT.
This week I'm tuned for Memory

21
00:01:29,480 --> 00:01:33,640
Driven AI, OP Free design, and
the deeper shift from tools that

22
00:01:33,640 --> 00:01:37,560
help to systems that decide.
We're hosting this episode from

23
00:01:37,560 --> 00:01:40,840
a converted workspace in San
Francisco's Mission District, 10

24
00:01:40,840 --> 00:01:43,680
blocks from where Open AI rolled
out its latest feature.

25
00:01:44,080 --> 00:01:47,040
The walls are covered in
whiteboards, glass panels, and

26
00:01:47,040 --> 00:01:50,920
live LLM dashboards.
It smells like espresso startup

27
00:01:50,920 --> 00:01:54,680
stress and ozone from the Always
On GPU rack in the corner.

28
00:01:54,880 --> 00:01:58,040
Founders are pacing between
desks, whispering about

29
00:01:58,040 --> 00:02:01,080
interface funnels.
Someone's running a demo where

30
00:02:01,080 --> 00:02:05,560
ChatGPT builds an itinerary,
books hotels, and emails

31
00:02:05,560 --> 00:02:08,960
confirmations quietly without
calling a single app.

32
00:02:09,479 --> 00:02:13,120
It feels less like a workspace
and more like a launch Bay for

33
00:02:13,120 --> 00:02:15,760
their new Internet.
And that's why we're here.

34
00:02:16,280 --> 00:02:20,480
Because this war, it's not about
who has the best model anymore.

35
00:02:20,600 --> 00:02:23,280
It's about who becomes the next
operating system.

36
00:02:23,840 --> 00:02:27,480
Until now, the AI battle was
about intelligence, benchmarks,

37
00:02:27,480 --> 00:02:31,520
context windows, logic chains.
But that's changed because

38
00:02:31,520 --> 00:02:34,480
whoever controls the interface
controls everything else.

39
00:02:34,800 --> 00:02:37,880
Search, commerce, content,
habits.

40
00:02:38,160 --> 00:02:41,640
And right now, the AI interfaces
are no longer assistance.

41
00:02:41,960 --> 00:02:44,600
They're replacing browsers and
soon.

42
00:02:44,760 --> 00:02:49,400
This week, ChatGPT hit 1 billion
weekly searches.

43
00:02:49,880 --> 00:02:51,520
That's a direct threat to
Google.

44
00:02:52,000 --> 00:02:55,960
But it's not just about traffic,
it's about default behavior.

45
00:02:56,240 --> 00:02:59,240
Ask an AI, get an answer, take
action.

46
00:02:59,600 --> 00:03:03,240
No middle, mid, no ads, no
traditional websites.

47
00:03:03,520 --> 00:03:05,800
That flow used to belong to
Google.

48
00:03:06,120 --> 00:03:09,040
Now it belongs to whoever owns
the interface.

49
00:03:09,400 --> 00:03:12,880
And Open AI just made its move.
But they're not alone.

50
00:03:13,400 --> 00:03:16,680
Meta is pushing Llama 4 into
Instagram and WhatsApp.

51
00:03:17,080 --> 00:03:20,320
Alibaba is embedding Quinn into
enterprise dashboards.

52
00:03:20,640 --> 00:03:23,920
And Google, well, Google is
trying to keep pace while

53
00:03:23,920 --> 00:03:26,200
protecting the search model it
helped create.

54
00:03:26,720 --> 00:03:30,080
The race isn't for best model,
it's for best placement.

55
00:03:30,320 --> 00:03:32,920
The front end layer, The
interface that captures the

56
00:03:32,920 --> 00:03:35,120
first question, not the last
click.

57
00:03:35,160 --> 00:03:37,000
Which is why this episode
matters.

58
00:03:37,360 --> 00:03:40,720
Because if the web is being
replaced not by a new site, but

59
00:03:40,720 --> 00:03:43,720
by a new interface, then
everything downstream of that

60
00:03:43,720 --> 00:03:48,360
interface shifts.
Advertising, e-commerce, AP

61
00:03:48,360 --> 00:03:52,040
design, even trust.
What used to be open is becoming

62
00:03:52,040 --> 00:03:55,000
curated.
What used to be searched is now

63
00:03:55,000 --> 00:03:59,080
suggested, and what used to be
in action is now a conversation.

64
00:03:59,520 --> 00:04:03,200
So today we'll break down this
new interface for not just the

65
00:04:03,200 --> 00:04:07,120
players, but the strategies.
Memory versus speed, multi

66
00:04:07,120 --> 00:04:10,680
modality versus simplicity,
agents versus AP is.

67
00:04:11,080 --> 00:04:14,960
We'll show how ChatGPT, Llama
four and Quinn three are turning

68
00:04:14,960 --> 00:04:18,519
into the new digital gatekeepers
and why the biggest AI battle of

69
00:04:18,519 --> 00:04:22,000
2025 isn't about code, it's
about control.

70
00:04:22,160 --> 00:04:27,840
Subscribe on Apple or Spotify,
follow us on X and share this

71
00:04:27,840 --> 00:04:31,560
episode with a friend.
Help us hit 10,000 downloads as

72
00:04:31,560 --> 00:04:34,440
we build the smart AI community
online.

73
00:04:34,640 --> 00:04:38,400
If segment one was about the
shift, segment 2 is about the

74
00:04:38,400 --> 00:04:40,560
business.
Let's follow the money.

75
00:04:40,760 --> 00:04:44,000
Let's talk money, because
beneath the interface shift we

76
00:04:44,000 --> 00:04:47,080
just explored, something even
bigger is happening.

77
00:04:47,520 --> 00:04:51,880
For years, AI tools have been
framed as assistants, copilots,

78
00:04:52,040 --> 00:04:55,880
productivity boosts.
But that framing hides the real

79
00:04:55,880 --> 00:04:58,160
story.
These interfaces aren't just

80
00:04:58,160 --> 00:05:01,640
smart, they're becoming the most
valuable layer in the digital

81
00:05:01,640 --> 00:05:04,440
economy.
Whoever owns the interface owns

82
00:05:04,440 --> 00:05:07,760
the monetization flow, and what
started as free chatbot

83
00:05:07,760 --> 00:05:11,720
experiments is now turning into
a multi trillion dollar business

84
00:05:11,720 --> 00:05:12,640
model.
Rewire.

85
00:05:12,760 --> 00:05:16,160
The pattern is clear.
First comes usage, then comes

86
00:05:16,160 --> 00:05:20,760
trust, then comes control.
Open AI now moves over a billion

87
00:05:20,760 --> 00:05:24,200
searches a week through ChatGPT.
That's not just traffic, it's

88
00:05:24,200 --> 00:05:26,640
behavior.
And what's you own behavior?

89
00:05:26,640 --> 00:05:29,360
You own monetization.
The shopping demo we mentioned

90
00:05:29,360 --> 00:05:32,080
in segment one, that's not a UX
experiment.

91
00:05:32,360 --> 00:05:34,920
It's a prototype for interface
native commerce.

92
00:05:35,120 --> 00:05:38,160
No website, no search engine, no
ad network.

93
00:05:38,480 --> 00:05:41,800
Just a model that remembers what
you like, guides your choices

94
00:05:41,800 --> 00:05:44,080
and closes the sale.
And here's the kicker.

95
00:05:44,280 --> 00:05:47,920
It's frictionless.
No logins, no affiliate clutter,

96
00:05:48,320 --> 00:05:50,400
no SEO.
Just results.

97
00:05:51,000 --> 00:05:54,240
If ChatGPT can show you what you
want, when you want it, and

98
00:05:54,240 --> 00:05:57,560
build trust along the way, it
bypasses the entire digital

99
00:05:57,560 --> 00:06:00,200
advertising industry.
And that's not an exaggeration.

100
00:06:00,600 --> 00:06:04,440
Google makes over $200 billion a
year from search and AD

101
00:06:04,440 --> 00:06:07,320
placement.
ChatGPT threatens to reroute

102
00:06:07,320 --> 00:06:11,800
that flow one query at a time.
What Open AI is building isn't a

103
00:06:11,800 --> 00:06:14,800
chat bot.
It's a commerce funnel, a new

104
00:06:14,800 --> 00:06:18,320
interface layer that replaces
navigation with conversation.

105
00:06:18,720 --> 00:06:22,840
And monetization lives in that
flow because the AI doesn't just

106
00:06:22,840 --> 00:06:27,040
wait for you to click.
It suggests, it curates, it

107
00:06:27,040 --> 00:06:29,960
remembers, and soon it will
negotiate.

108
00:06:30,360 --> 00:06:33,800
We're heading into a world where
your AI doesn't just answer your

109
00:06:33,800 --> 00:06:36,160
questions, it brokers your
choices.

110
00:06:36,600 --> 00:06:39,040
That's the real monetization
revolution.

111
00:06:39,520 --> 00:06:42,440
And that's just one track.
Now look at Meta.

112
00:06:42,760 --> 00:06:46,120
They're not monetizing through
transactions, they're doing it

113
00:06:46,120 --> 00:06:49,120
through engagement.
Llama 4 isn't just being

114
00:06:49,160 --> 00:06:52,320
integrated into Meta's
ecosystem, it's becoming the

115
00:06:52,320 --> 00:06:55,440
ecosystem.
Instagram's AI stickers,

116
00:06:55,640 --> 00:06:59,240
WhatsApp smart replies, Threads,
Discovery engine.

117
00:06:59,520 --> 00:07:01,320
Each interaction is a data
point.

118
00:07:01,760 --> 00:07:06,360
Each data point feeds attention,
and attention feeds revenue.

119
00:07:06,920 --> 00:07:10,680
Meta's goal?
Turn every scroll share and send

120
00:07:10,680 --> 00:07:13,560
into training data for
monetizable prediction.

121
00:07:13,720 --> 00:07:17,600
Different model, same outcome.
Open AI wants to monetize trust

122
00:07:17,600 --> 00:07:19,720
and action.
Meadow wants to monetize

123
00:07:19,720 --> 00:07:23,080
attention and interaction.
But both are doing the same

124
00:07:23,080 --> 00:07:25,240
thing, using the interface as
the control point.

125
00:07:25,560 --> 00:07:28,960
And once the user shifts from
browser to assistant, from tap

126
00:07:28,960 --> 00:07:31,840
to prompt, everything about
monetization changes.

127
00:07:32,320 --> 00:07:36,120
It becomes predictive,
personalized, and more powerful

128
00:07:36,120 --> 00:07:39,800
than banner ads ever were.
What's wild is that Google knows

129
00:07:39,800 --> 00:07:43,640
this, and they're still stuck.
Their core revenue depends on

130
00:07:43,640 --> 00:07:46,360
being the last step before
action.

131
00:07:46,720 --> 00:07:51,680
But AI interfaces are becoming
the first step, and that shift

132
00:07:51,800 --> 00:07:54,320
destroys the economics of
search.

133
00:07:54,720 --> 00:07:57,840
If I asked Jim and I for the
best sushi spot nearby and it

134
00:07:57,840 --> 00:07:59,880
books the table for me, no
search happened.

135
00:08:00,280 --> 00:08:03,400
No ad was clicked, no ranking
battle was fought.

136
00:08:03,880 --> 00:08:06,800
Google made nothing.
Which is why Google's

137
00:08:06,800 --> 00:08:10,400
monetization story now depends
on embedding Gemini into

138
00:08:10,400 --> 00:08:13,880
everything Docs, Gmail, Chrome,
Maps.

139
00:08:14,320 --> 00:08:17,480
Not as a destination, but as an
ambient presence.

140
00:08:17,960 --> 00:08:20,920
Their play isn't to compete with
ChatGPT head on.

141
00:08:21,360 --> 00:08:26,040
It's to be everywhere, always
available and always pushing

142
00:08:26,040 --> 00:08:30,200
usage back into the ad network.
But it's a defensive strategy

143
00:08:30,560 --> 00:08:33,440
because they know that the
longer users stay inside another

144
00:08:33,440 --> 00:08:36,640
interface, the less monetization
Google can extract.

145
00:08:36,799 --> 00:08:40,760
Meanwhile, Alibaba is quietly
building a completely different

146
00:08:40,760 --> 00:08:43,840
interface model with Quinn.
They're not chasing Western

147
00:08:43,840 --> 00:08:47,000
style user experiences.
They're optimizing for

148
00:08:47,000 --> 00:08:50,800
enterprise control, B to B
dashboards, agent layers that

149
00:08:50,800 --> 00:08:53,880
handle procurement, logistics
and internal queries.

150
00:08:54,040 --> 00:08:57,560
Their monetization doesn't come
from ads or shopping, it comes

151
00:08:57,560 --> 00:08:59,280
from owning how businesses
operate.

152
00:08:59,720 --> 00:09:02,440
Quinn isn't trying to become
your assistant, it's trying to

153
00:09:02,440 --> 00:09:05,520
become your COO.
That's the third monetization

154
00:09:05,520 --> 00:09:10,000
model we've covered Open AIS,
action based funnel, Nata's

155
00:09:10,000 --> 00:09:14,000
attention based economy and now
Alibaba's enterprise command

156
00:09:14,000 --> 00:09:16,120
stack.
Each of these players is

157
00:09:16,120 --> 00:09:18,640
optimizing for a different
outcome, but they're all using

158
00:09:18,640 --> 00:09:21,240
the interface to get there, and
that's the shift.

159
00:09:21,600 --> 00:09:25,280
It's not about selling the AI,
it's about owning the point of

160
00:09:25,280 --> 00:09:28,600
interaction, because once you
control that, everything else

161
00:09:28,600 --> 00:09:32,080
flows downstream.
Let's be blunt, ChatGPT may have

162
00:09:32,080 --> 00:09:34,880
started as a free tool, but it's
not staying that way.

163
00:09:35,720 --> 00:09:39,160
Whether it's open AI taking an
affiliate cut for purchases,

164
00:09:39,160 --> 00:09:42,800
Meta monetizing model enhance
time on platform, or Google

165
00:09:42,800 --> 00:09:46,720
trying to force AI query back
into ads, every interface is

166
00:09:46,720 --> 00:09:49,920
being monetized.
The free window was a Trojan

167
00:09:49,920 --> 00:09:52,400
horse.
The real business is just

168
00:09:52,400 --> 00:09:54,560
starting.
And here's what comes next.

169
00:09:54,800 --> 00:09:57,120
Monetization layers built into
memory.

170
00:09:57,760 --> 00:10:00,040
Your AI won't just remember your
name.

171
00:10:00,440 --> 00:10:03,480
It'll remember your patterns,
your preferences, your buying

172
00:10:03,480 --> 00:10:05,800
cycles.
And when it offers you a deal,

173
00:10:05,920 --> 00:10:09,640
it won't be random.
It'll be timed, strategic,

174
00:10:10,000 --> 00:10:13,080
optimized for conversion.
That's not surveillance.

175
00:10:13,200 --> 00:10:16,600
It's monetization with a memory.
And it's the future of how these

176
00:10:16,600 --> 00:10:20,360
interfaces will scale.
So if segment one was about

177
00:10:20,360 --> 00:10:24,080
behavior and segment 2 is about
business, segment 3 is about

178
00:10:24,080 --> 00:10:26,200
capability.
Because to make this

179
00:10:26,200 --> 00:10:29,360
monetization work, the
interfaces need to evolve.

180
00:10:29,840 --> 00:10:34,640
Static answers aren't enough.
You need voice, vision, context

181
00:10:34,720 --> 00:10:37,600
and agency.
In other words, you need multi

182
00:10:37,600 --> 00:10:40,360
modality.
It used to be enough for an AI

183
00:10:40,360 --> 00:10:45,640
to just reply with text, type in
a question, get back an answer.

184
00:10:46,000 --> 00:10:50,000
But that era is already ending
fast, because the new war isn't

185
00:10:50,000 --> 00:10:54,960
just about intelligence.
It's about census, sight, sound,

186
00:10:55,440 --> 00:10:58,560
context.
The winners of this race don't

187
00:10:58,560 --> 00:11:02,280
just respond, they perceive.
They collaborate.

188
00:11:02,880 --> 00:11:06,040
They hacked.
And in that war text only models

189
00:11:06,040 --> 00:11:09,560
are already outdated.
Multi modality isn't an upgrade,

190
00:11:09,720 --> 00:11:13,040
it's the new baseline.
The ability to process images,

191
00:11:13,040 --> 00:11:16,920
generate visuals, understand
voice and bridge formats is now

192
00:11:16,920 --> 00:11:20,280
essential.
Life isn't made of tokens, it's

193
00:11:20,280 --> 00:11:25,040
made of signals, data, emotion.
If your interface can't see,

194
00:11:25,040 --> 00:11:28,640
hear or interpret all in one
flow, it's not just limited,

195
00:11:28,960 --> 00:11:32,480
it's broken.
That's why Open AI launched GPT

196
00:11:32,520 --> 00:11:36,280
Image 1 this week.
It's the vision model powering

197
00:11:36,280 --> 00:11:40,120
chat, GP TS, new shopping
experience, generating product

198
00:11:40,120 --> 00:11:43,520
visuals, editing photos, and
understanding screenshots.

199
00:11:43,840 --> 00:11:48,400
Over 700 million images were
created in seven days.

200
00:11:48,680 --> 00:11:53,280
That's not hype, that's demand.
And it proves users want their

201
00:11:53,360 --> 00:11:56,520
AI to see what they see.
Meta's right there, too.

202
00:11:56,880 --> 00:12:00,960
Their Llama 4 stack includes
Scout and Maverick 2 multimodal

203
00:12:00,960 --> 00:12:05,360
systems released this week.
Scout detects user sentiment and

204
00:12:05,360 --> 00:12:09,720
visual cues, Maverick builds
personalized image flows in real

205
00:12:09,720 --> 00:12:14,560
time, and Meadows deploying both
inside Instagram threads and

206
00:12:14,560 --> 00:12:17,480
WhatsApp.
That means their interface isn't

207
00:12:17,480 --> 00:12:22,440
just smart, it's ambient
constant embedded into the feed.

208
00:12:22,680 --> 00:12:25,440
Don't overlook Alibaba's Quinn
the Third.

209
00:12:25,720 --> 00:12:29,280
It doesn't chase flash, but it's
built for structure.

210
00:12:29,680 --> 00:12:33,600
Quinn's Multimodal Core parses
PDFs, reads UIS, audits

211
00:12:33,600 --> 00:12:37,320
dashboards, and when paired with
China's massive enterprise data,

212
00:12:37,400 --> 00:12:40,880
it becomes a precision tool
designed to run the systems that

213
00:12:40,880 --> 00:12:45,200
keep the economy moving.
And this evolution changes

214
00:12:45,240 --> 00:12:48,640
everything.
A multimodal AI doesn't just

215
00:12:48,640 --> 00:12:52,680
answer your question.
It books your trip by scanning

216
00:12:52,680 --> 00:12:55,520
maps, checking weather, finding
seats.

217
00:12:56,040 --> 00:13:00,040
It shops for you by analyzing
your space, comparing styles,

218
00:13:00,040 --> 00:13:03,880
calculating fit.
It teaches by translating

219
00:13:03,880 --> 00:13:07,600
charts, syncing voice feedback,
even role-playing study

220
00:13:07,600 --> 00:13:11,080
sessions.
This isn't search, it's full

221
00:13:11,080 --> 00:13:14,040
spectrum assistance.
And that's what makes it stick.

222
00:13:14,440 --> 00:13:18,840
Text only interfaces are flat,
but multimodal systems are

223
00:13:18,840 --> 00:13:21,840
dimensional.
They switch formats mid task,

224
00:13:22,000 --> 00:13:25,840
adapt to real world input, and
deliver output across senses.

225
00:13:26,320 --> 00:13:28,400
You don't need tabs, apps or
filters.

226
00:13:28,760 --> 00:13:31,880
Just describe what you want and
it figures out how to make it

227
00:13:31,880 --> 00:13:34,720
happen.
That's not a tool, that's an

228
00:13:34,720 --> 00:13:38,040
interface you live inside.
That's why Google's moving fast,

229
00:13:38,040 --> 00:13:40,760
too.
This week they launched Dolphin

230
00:13:40,760 --> 00:13:44,200
Gemma, an open multimodal model
that decodes Dolphin

231
00:13:44,200 --> 00:13:46,600
vocalizations and runs on Pixel
phones.

232
00:13:47,080 --> 00:13:50,440
It's part of a larger push to
embed Gemini into Android,

233
00:13:50,680 --> 00:13:53,720
Chrome and beyond.
So your phone's AI doesn't just

234
00:13:53,720 --> 00:13:57,960
reply, it watches, listens,
suggests.

235
00:13:58,120 --> 00:14:01,120
And strategically this changes
the whole platform game.

236
00:14:01,480 --> 00:14:05,640
A multi modal AI keeps you
inside one flow, no switching

237
00:14:05,640 --> 00:14:09,120
between apps for formats.
That's how power accumulates.

238
00:14:09,480 --> 00:14:12,600
Not by being the best model, but
by being the one you never

239
00:14:12,600 --> 00:14:16,960
leave.
One prompt, 1 context, one

240
00:14:16,960 --> 00:14:20,520
system that's lock in.
And that lock in fuels

241
00:14:20,520 --> 00:14:23,200
monetization.
An interface that sees your

242
00:14:23,200 --> 00:14:26,840
space, can recommend products,
one that hears your voice, can

243
00:14:26,840 --> 00:14:30,360
respond with tone, one that
understands screenshots, can

244
00:14:30,360 --> 00:14:33,640
summarize and sell.
Multi modality isn't just

245
00:14:33,640 --> 00:14:36,760
better, it's profitable and the
platforms know it.

246
00:14:37,240 --> 00:14:40,800
We're still early.
Most users haven't tried vision,

247
00:14:40,920 --> 00:14:45,320
voice or context memory yet, but
they will, and once they do,

248
00:14:45,320 --> 00:14:47,840
text only models will feel
broken.

249
00:14:48,240 --> 00:14:51,560
Multi modality isn't the future,
it's the standard.

250
00:14:51,960 --> 00:14:54,360
Segment 3 show you the
capability.

251
00:14:54,600 --> 00:14:57,920
Segment 4 shows what happens
when capability meets trust,

252
00:14:58,200 --> 00:15:02,840
when tools become team mates.
Not long ago AI was just a tool,

253
00:15:03,000 --> 00:15:06,160
something you use to answer a
question, finish a sentence,

254
00:15:06,200 --> 00:15:07,760
maybe generate a few lines of
code.

255
00:15:08,600 --> 00:15:13,280
But that line is fading fast
because today's models don't

256
00:15:13,280 --> 00:15:17,480
just react, They remember.
They take initiative.

257
00:15:18,040 --> 00:15:21,120
They adapt over time.
And that means they're no longer

258
00:15:21,120 --> 00:15:24,800
just responding to your prompts.
They're learning who you are,

259
00:15:25,160 --> 00:15:29,680
acting in your place, making
decisions on your behalf.

260
00:15:30,480 --> 00:15:35,400
The assistant is becoming an
agent, and the tool it's turning

261
00:15:35,400 --> 00:15:37,840
into a teammate.
This shift didn't happen

262
00:15:37,840 --> 00:15:39,600
overnight, but it's
accelerating.

263
00:15:40,080 --> 00:15:43,720
Open AI has now rolled out
memory to millions of ChatGPT

264
00:15:43,720 --> 00:15:45,600
users.
That means when you use the

265
00:15:45,600 --> 00:15:47,760
model, it doesn't just process
input.

266
00:15:47,800 --> 00:15:52,440
It remembers context, your name,
your preferences, your style,

267
00:15:52,840 --> 00:15:56,440
your goals, not just within a
single session, but across time.

268
00:15:56,560 --> 00:16:00,480
And when memory combines with
task execution and what we call

269
00:16:00,480 --> 00:16:04,760
agency, that's when the human AI
relationship fundamentally

270
00:16:04,760 --> 00:16:09,120
changes, because now the model
isn't just helping you, it's

271
00:16:09,120 --> 00:16:12,000
working with you.
Think about what that really

272
00:16:12,000 --> 00:16:14,320
means.
If your AI remembers your

273
00:16:14,320 --> 00:16:18,000
writing voice, your schedule,
your favorite vendors, and your

274
00:16:18,000 --> 00:16:20,640
next launch.
It can draft emails before you

275
00:16:20,640 --> 00:16:23,040
ask.
Propose meeting times without

276
00:16:23,040 --> 00:16:25,720
you checking.
Build task chains based on your

277
00:16:25,720 --> 00:16:29,800
last three weeks of activity.
It's not just reactive, it's

278
00:16:29,800 --> 00:16:33,040
proactive.
And the moment that proactivity

279
00:16:33,040 --> 00:16:35,600
is trusted, you stop thinking of
it as a tool.

280
00:16:36,000 --> 00:16:37,640
You start thinking of it as a
partner.

281
00:16:37,840 --> 00:16:40,560
And we're already seeing this in
the interface design.

282
00:16:41,040 --> 00:16:43,360
Opening eyes.
Assistant memory is framed as

283
00:16:43,360 --> 00:16:46,640
helpful context, but the effect
is much larger.

284
00:16:46,960 --> 00:16:50,000
Tasks that once required 5
prompts now take one.

285
00:16:50,320 --> 00:16:53,320
Some need none.
Models remember your workflows,

286
00:16:53,680 --> 00:16:58,080
your writing style, your tone.
It's subtle but powerful.

287
00:16:58,240 --> 00:17:01,800
And when paired with agents
systems that can take multi step

288
00:17:01,800 --> 00:17:05,000
action across apps, it becomes
infrastructure.

289
00:17:05,440 --> 00:17:08,960
You don't just delegate tasks,
you outsource thinking.

290
00:17:09,240 --> 00:17:14,760
Let's talk agents right now.
The top LLMSG, PT4-O, Claude 3,

291
00:17:14,760 --> 00:17:19,720
Opus Quinn 3.5 can all be
wrapped in a gentic frameworks.

292
00:17:20,160 --> 00:17:23,520
Langchen Auto, Gen.
Crew AI, and others let you

293
00:17:23,520 --> 00:17:27,960
build agents that reason, plan,
and execute across steps.

294
00:17:28,319 --> 00:17:33,320
You tell it a goal, it figures
out the how, and with memory it

295
00:17:33,320 --> 00:17:36,800
refines itself over time.
We're at far from agents that

296
00:17:36,800 --> 00:17:40,760
write code tested, debug it,
deploy it, and explain the

297
00:17:40,760 --> 00:17:44,720
results back to you.
No task manager, no developer,

298
00:17:45,160 --> 00:17:47,400
just the agent.
These agents don't have to be

299
00:17:47,400 --> 00:17:51,680
fully autonomous to be powerful.
Even semi agentic models, ones

300
00:17:51,680 --> 00:17:55,800
that nudge, suggest or prep, are
changing how users behave.

301
00:17:56,280 --> 00:17:59,520
Once an interface suggests your
next task, you're no longer in

302
00:17:59,520 --> 00:18:03,200
command, you're collaborating.
And that's a shift in power,

303
00:18:03,600 --> 00:18:06,000
because when the interface
becomes intelligent enough to

304
00:18:06,000 --> 00:18:09,320
shape your workflow, it's no
longer a mirror, it's a guide.

305
00:18:09,480 --> 00:18:12,200
And This is why interface memory
is so important.

306
00:18:12,720 --> 00:18:15,240
Without it, every session
resets.

307
00:18:15,840 --> 00:18:18,880
You lose context, you repeat
instructions.

308
00:18:19,480 --> 00:18:22,400
But with memory, the interface
becomes continuous.

309
00:18:22,920 --> 00:18:28,120
It evolves, learns, optimizes.
And that's what creates trust.

310
00:18:28,600 --> 00:18:32,360
You start relying on it,
offloading to it, sharing more,

311
00:18:32,920 --> 00:18:34,520
which in turn gives it more
power.

312
00:18:34,640 --> 00:18:39,160
This isn't just a UX upgrade,
it's a psychological contract

313
00:18:39,560 --> 00:18:43,200
and users are already signing it
whether they know it or not.

314
00:18:43,280 --> 00:18:47,520
But with that trust comes risk,
because memory isn't just

315
00:18:47,520 --> 00:18:51,320
helpful, it's intimate.
It tracks preferences, stores

316
00:18:51,320 --> 00:18:53,880
behavior patterns, and makes
assumptions.

317
00:18:54,320 --> 00:18:58,200
And when those assumptions Dr.
actions like auto scheduling,

318
00:18:58,200 --> 00:19:01,960
e-mail replies, or even purchase
decisions, mistakes carry

319
00:19:01,960 --> 00:19:04,240
consequences.
Who's responsible?

320
00:19:04,240 --> 00:19:08,360
When an agent books the wrong
flight or protects its role, or

321
00:19:08,400 --> 00:19:12,080
auto generates a message that
creates a problem, We're

322
00:19:12,080 --> 00:19:14,720
entering an era of blurred
accountability.

323
00:19:14,960 --> 00:19:17,840
That's why companies are being
so careful with the roll out.

324
00:19:18,320 --> 00:19:20,480
Open AI lets you turn memory
off.

325
00:19:21,000 --> 00:19:22,920
Cloud Three lets you clear
sessions.

326
00:19:23,320 --> 00:19:26,240
Quinn Three's enterprise
variants log every interaction,

327
00:19:26,720 --> 00:19:30,000
but let's be honest, once users
experience a good memory loop,

328
00:19:30,120 --> 00:19:34,320
they rarely want to go back.
The gain in speed, relevance,

329
00:19:34,360 --> 00:19:36,560
and intelligence is just too
high.

330
00:19:36,960 --> 00:19:41,320
Memory is sticky and once it's
trusted it becomes default.

331
00:19:41,560 --> 00:19:45,520
And this isn't just about memory
or agency in isolation.

332
00:19:45,880 --> 00:19:47,680
It's about how they work
together.

333
00:19:48,200 --> 00:19:52,000
Memory builds trust, Agency
builds capability.

334
00:19:52,400 --> 00:19:54,640
Together, they create
continuity.

335
00:19:55,160 --> 00:19:59,360
That's what turns an assistant
into a teammate, a system that

336
00:19:59,360 --> 00:20:03,240
remembers your goals, takes
action, and adapts in real time.

337
00:20:03,720 --> 00:20:08,520
We're talking about interfaces
that don't just assist you, they

338
00:20:08,520 --> 00:20:10,880
evolve with you.
So where does this lead?

339
00:20:11,440 --> 00:20:14,920
Into a future where your AI
handles e-mail, project

340
00:20:14,920 --> 00:20:17,880
management, contract
negotiation, task chains?

341
00:20:18,280 --> 00:20:22,480
Into a future where teams shrink
and agents scale, where every

342
00:20:22,480 --> 00:20:26,920
solo founder runs a staff of AI
teammates, and where the best AI

343
00:20:26,920 --> 00:20:29,600
isn't just smart, it's
synchronized with you.

344
00:20:30,040 --> 00:20:33,440
That's the new interface
relationship, one that's

345
00:20:33,480 --> 00:20:37,080
ambient, trusted, and eventually
indispensable.

346
00:20:37,200 --> 00:20:40,880
Segment 4 was about evolution
from tools to teammates.

347
00:20:41,240 --> 00:20:44,880
But in Segment 5 we zoom out
because if agents are replacing

348
00:20:44,880 --> 00:20:47,040
assistance, what happens to the
rest of the stack?

349
00:20:47,640 --> 00:20:49,120
The apps?
The tabs?

350
00:20:49,360 --> 00:20:51,040
The workflows?
The answer?

351
00:20:51,480 --> 00:20:54,720
They collapse and the AI becomes
the new operating system.

352
00:20:54,920 --> 00:20:56,680
Let's stop calling them
assistants.

353
00:20:57,200 --> 00:20:59,920
Let's stop pretending these
interfaces are just smarter

354
00:20:59,920 --> 00:21:03,000
versions of old tools.
Because what's really happening

355
00:21:03,000 --> 00:21:05,040
isn't an upgrade, it's a
takeover.

356
00:21:05,400 --> 00:21:10,120
Quiet, seamless, intentional.
The AI interface isn't just

357
00:21:10,120 --> 00:21:14,040
helping you use your apps, it's
replacing them one by one.

358
00:21:14,600 --> 00:21:17,280
Search is gone.
Calendar is fading.

359
00:21:17,840 --> 00:21:20,520
E-mail drafted before you even
open it.

360
00:21:20,960 --> 00:21:24,040
This isn't just new software,
it's a new system.

361
00:21:24,560 --> 00:21:27,440
The AI isn't living on top of
the OS anymore.

362
00:21:27,960 --> 00:21:30,920
It is the OS.
This is the quietest revolution

363
00:21:30,920 --> 00:21:34,680
in tech and also the most
complete because instead of

364
00:21:34,680 --> 00:21:37,560
building new apps, the AI
interface dissolves them.

365
00:21:37,560 --> 00:21:42,080
Need to draft a contract?
You don't open Word, you tell

366
00:21:42,080 --> 00:21:44,080
the model.
Need a trip planned?

367
00:21:44,560 --> 00:21:47,360
No tabs, no booking sites, just
the assistant.

368
00:21:47,720 --> 00:21:50,480
It reads your calendar, checks
your preferences, confirms your

369
00:21:50,480 --> 00:21:53,400
time zone, then handles it.
The structure of the digital

370
00:21:53,400 --> 00:21:58,520
world, Menus, apps, folders,
navigation is being replaced by

371
00:21:58,520 --> 00:22:00,760
a single input layer, the
interface.

372
00:22:00,960 --> 00:22:04,240
In that interface, it's no
longer just about command and

373
00:22:04,240 --> 00:22:08,800
response, it's about continuity.
The assistant doesn't need you

374
00:22:08,800 --> 00:22:11,720
to remember the workflow.
It remembers for you.

375
00:22:12,240 --> 00:22:15,440
It executes.
It refines and hands back

376
00:22:15,440 --> 00:22:18,520
outcomes, not options.
That's what an O does.

377
00:22:18,920 --> 00:22:22,880
It abstracts complexity.
An AI is doing it better than

378
00:22:22,880 --> 00:22:25,560
software ever could.
This is why the smartest

379
00:22:25,560 --> 00:22:27,120
companies aren't building us
anymore.

380
00:22:27,600 --> 00:22:30,840
They're building interfaces or
they're building for the

381
00:22:30,840 --> 00:22:34,000
interface because the value
layer is shifting from the app

382
00:22:34,000 --> 00:22:37,520
stack to the input stack.
Developers are asking how does

383
00:22:37,520 --> 00:22:41,600
this product plug into ChatGPT,
into Gemini, into Llama?

384
00:22:42,000 --> 00:22:44,760
Because if it doesn't, it
doesn't scale.

385
00:22:44,920 --> 00:22:48,560
Users won't hunt down new apps
when their AI can already do the

386
00:22:48,560 --> 00:22:51,160
job.
The platform shift isn't mobile

387
00:22:51,160 --> 00:22:55,960
to AI, it's interface to intent.
Open AI's App Store isn't about

388
00:22:55,960 --> 00:22:59,080
plug insurance, it's about
conditioning users to live

389
00:22:59,080 --> 00:23:02,880
inside the model.
Once the AI can perform any

390
00:23:02,880 --> 00:23:06,480
action on your behalf, the
interface becomes the only layer

391
00:23:06,480 --> 00:23:09,120
that matters.
The more frictionless it is, the

392
00:23:09,120 --> 00:23:13,080
more powerful it becomes.
And frictionless means no apps,

393
00:23:13,160 --> 00:23:16,480
no browser, no tabs.
Just one prompt, one

394
00:23:16,480 --> 00:23:20,480
conversation, one outcome.
That's not software, that's an

395
00:23:20,760 --> 00:23:22,480
OS.
Google knows this.

396
00:23:22,880 --> 00:23:26,080
That's why Gemini is being
embedded directly into Chrome,

397
00:23:26,080 --> 00:23:31,960
Gmail, Android, Docs, Sheets,
Meet, and every layer of the

398
00:23:31,960 --> 00:23:35,800
Workspace suite.
Not as a tool, but as the new

399
00:23:35,800 --> 00:23:38,600
foundation.
You don't open an app to use

400
00:23:38,600 --> 00:23:45,720
Gemini, it's just they're
persistent, present, ambient.

401
00:23:46,240 --> 00:23:50,040
It turns every click into a
conversation, every workflow

402
00:23:50,040 --> 00:23:54,320
into a suggestion, and every
suggestion into an action.

403
00:23:54,760 --> 00:23:56,640
But Mehta's taking a different
path.

404
00:23:57,080 --> 00:24:00,920
Instead of embedding the model
into legacy software, they're

405
00:24:00,920 --> 00:24:04,800
embedding it into the user's
life, social messaging,

406
00:24:04,840 --> 00:24:08,280
identity.
The Llama 4 interface isn't just

407
00:24:08,280 --> 00:24:12,040
an Instagram.
It's in your DMS, in your camera

408
00:24:12,040 --> 00:24:17,400
feed, in your recommendations.
It's not replacing tools, it's

409
00:24:17,400 --> 00:24:20,320
replacing instincts.
You don't think search, you

410
00:24:20,320 --> 00:24:23,480
think ask, You don't navigate,
you react.

411
00:24:24,520 --> 00:24:26,960
And that shift over time is
irreversible.

412
00:24:27,120 --> 00:24:30,640
The real take away?
Interfaces aren't just changing

413
00:24:30,640 --> 00:24:33,480
how we use tech, they're
changing how tech behaves.

414
00:24:33,920 --> 00:24:37,600
And OS isn't a collection of
programs, it's a system for

415
00:24:37,600 --> 00:24:40,320
managing logic, action, and
execution.

416
00:24:40,680 --> 00:24:44,080
And AI does all three.
Once the model is smart enough

417
00:24:44,080 --> 00:24:48,120
to understand intent, manage
context and control apps, what's

418
00:24:48,120 --> 00:24:52,320
the role of the traditional OS?
It becomes invisible, replaced,

419
00:24:52,600 --> 00:24:54,720
irrelevant.
This is already happening.

420
00:24:55,000 --> 00:24:57,840
Open AI can schedule your
meetings, write your emails,

421
00:24:57,840 --> 00:25:01,400
summarize your docs, shop for
you, book your travel, and

422
00:25:01,400 --> 00:25:04,440
handle follow up, all without
touching Outlook, Google

423
00:25:04,440 --> 00:25:08,880
Calendar, Expedia, or Slack.
It doesn't open taps, it

424
00:25:08,880 --> 00:25:11,560
bypasses them, and that's what
makes it an OS.

425
00:25:12,000 --> 00:25:14,440
Not because it runs on the
machine, but because it runs

426
00:25:14,440 --> 00:25:16,760
your logic.
And it's not just personal

427
00:25:16,760 --> 00:25:19,560
productivity.
Enterprise systems are shifting

428
00:25:19,560 --> 00:25:21,840
too.
Agent stacks are being trained

429
00:25:21,840 --> 00:25:25,720
on internal documents connected
to CRMS synced with compliance

430
00:25:25,720 --> 00:25:27,960
tools.
They don't ask users to open

431
00:25:27,960 --> 00:25:30,240
platforms.
They handle the workflow end to

432
00:25:30,240 --> 00:25:32,720
end.
For many companies, the AI

433
00:25:32,720 --> 00:25:36,360
interface has already become the
default portal to work, and the

434
00:25:36,360 --> 00:25:39,440
software stuck.
It's just the back end invisible

435
00:25:39,440 --> 00:25:41,680
to the user.
This changes everything.

436
00:25:42,040 --> 00:25:45,840
Funding models, product design,
go to market strategy.

437
00:25:46,280 --> 00:25:50,160
If the AI interface becomes the
OS and start-ups don't build

438
00:25:50,160 --> 00:25:53,800
apps, they build for agents.
They don't raise to grow a user

439
00:25:53,800 --> 00:25:56,000
base.
They raise to integrate with the

440
00:25:56,000 --> 00:25:58,800
dominant interface layer.
And that's a much more

441
00:25:58,800 --> 00:26:02,600
centralized, more defensible and
more dangerous paradigm because

442
00:26:02,600 --> 00:26:05,400
the platform risk now lives at
the model layer.

443
00:26:05,800 --> 00:26:09,840
That's the deeper tension.
The new OS isn't open.

444
00:26:10,240 --> 00:26:14,480
It's not a Linux distro.
It's proprietary, controlled,

445
00:26:14,920 --> 00:26:19,400
memory enabled behavior.
Optimizing the same AI that

446
00:26:19,400 --> 00:26:23,280
helps you today might nudge you
tomorrow, might prioritize

447
00:26:23,280 --> 00:26:27,840
responses based on incentives,
licensing, or subtle nudges.

448
00:26:28,200 --> 00:26:32,720
When the OS is a model and the
model is a gentic, power doesn't

449
00:26:32,720 --> 00:26:35,960
just flow to the interface, it
flows through it.

450
00:26:36,680 --> 00:26:41,680
So if Segment 4 showed us how
how AIS evolve into teammates,

451
00:26:41,760 --> 00:26:44,600
Segment 5 just proved the next
step.

452
00:26:45,240 --> 00:26:48,640
These agents aren't living in
your apps, they're replacing

453
00:26:48,640 --> 00:26:51,320
them.
The OS isn't software anymore,

454
00:26:51,760 --> 00:26:55,240
it's intelligence.
And in Segment 6, that

455
00:26:55,240 --> 00:26:59,360
intelligence goes global.
Because the interface wars

456
00:26:59,520 --> 00:27:02,080
aren't just technical, they're
geopolitical.

457
00:27:02,200 --> 00:27:05,720
The interface war isn't just
playing out in Silicon Valley,

458
00:27:05,880 --> 00:27:09,640
It's gone global.
From Beijing to Berlin, from Sao

459
00:27:09,640 --> 00:27:13,640
Paulo to Singapore, governments,
companies and sovereign funds

460
00:27:13,840 --> 00:27:16,800
are racing to control the next
digital layer.

461
00:27:17,240 --> 00:27:20,800
Not the model, not the chip.
The interface.

462
00:27:21,320 --> 00:27:25,280
The layer that sits between
human intent and machine action

463
00:27:25,680 --> 00:27:31,360
and the stakes, The cultural,
political, economic Whoever

464
00:27:31,360 --> 00:27:36,240
controls that layer shapes what
gets built and who benefits.

465
00:27:36,600 --> 00:27:39,560
Because interfaces aren't
neutral, they reflect the

466
00:27:39,560 --> 00:27:43,240
values, incentives, and
structures of the ecosystems

467
00:27:43,240 --> 00:27:46,600
they're born in.
In the US, Open AI and Meta are

468
00:27:46,600 --> 00:27:48,960
fighting for consumer trust at
global scale.

469
00:27:49,400 --> 00:27:53,160
In China, Alibaba's Quin 3 is
being embedded across state

470
00:27:53,160 --> 00:27:56,720
backed enterprise systems.
In the EU, it's not the model

471
00:27:56,720 --> 00:28:00,520
that leads, it's the regulator.
And everywhere else the game is

472
00:28:00,520 --> 00:28:03,880
wide open.
Let's start with open AI chat.

473
00:28:03,880 --> 00:28:07,960
GP TS new shopping features
rolled out just days ago aren't

474
00:28:07,960 --> 00:28:11,840
just about UX, they're about
locking in the interface layer

475
00:28:11,920 --> 00:28:17,040
with 1 billion weekly searches,
plugins, memory and GPT image.

476
00:28:17,040 --> 00:28:21,040
One open AIS play is to become
the global front door to the

477
00:28:21,040 --> 00:28:23,040
Internet.
But there's friction.

478
00:28:23,160 --> 00:28:26,840
Europe's regulators are circling
and other markets are watching.

479
00:28:27,280 --> 00:28:30,200
The EU's AI Act marks a turning
point.

480
00:28:30,480 --> 00:28:34,640
It classifieds AI interfaces as
high risk systems requiring

481
00:28:34,640 --> 00:28:39,560
transparency, explainability,
auditability, memory, agents,

482
00:28:39,600 --> 00:28:42,320
personalization.
All of it triggers scrutiny.

483
00:28:42,760 --> 00:28:46,360
And open a IS model which relies
on behavior modeling to build

484
00:28:46,360 --> 00:28:49,200
trust and monetization may not
clear the bar.

485
00:28:49,600 --> 00:28:53,840
Meanwhile, Meta is going wide.
This week they rolled out the

486
00:28:53,840 --> 00:28:57,760
Llama 4 app, embedding Scout and
Maverick into Instagram,

487
00:28:57,840 --> 00:29:01,720
WhatsApp and Threads.
Not as standalone tools, but as

488
00:29:01,720 --> 00:29:04,640
native features.
The interface just shows U

489
00:29:05,240 --> 00:29:09,320
inside your feed, your messages,
your camera.

490
00:29:09,920 --> 00:29:14,360
This is distribution without
friction and it scales across 4

491
00:29:14,360 --> 00:29:17,400
billion users.
Meta's localization play is

492
00:29:17,400 --> 00:29:19,600
unmatched.
Their models are trained on

493
00:29:19,600 --> 00:29:23,360
dozens of languages and cultural
contexts, giving them an edge in

494
00:29:23,360 --> 00:29:27,040
Latin America, Southeast Asia,
Africa, regions where mobile

495
00:29:27,040 --> 00:29:30,160
first habits dominate and legacy
app UX falls short.

496
00:29:30,520 --> 00:29:34,240
Their interface is global by
default and tuned for instinct,

497
00:29:34,360 --> 00:29:37,520
not structure.
In China, the strategy is even

498
00:29:37,520 --> 00:29:41,200
more centralized.
Alibaba's Quinn 3, launched this

499
00:29:41,200 --> 00:29:45,000
week, isn't chasing consumers.
It's powering enterprise

500
00:29:45,000 --> 00:29:49,040
dashboards, finance,
procurement, logistics tuned for

501
00:29:49,040 --> 00:29:52,440
Chinese workflows, accounting
standards and regulations.

502
00:29:52,760 --> 00:29:56,080
It's not your assistant, it's
your operating layer.

503
00:29:56,440 --> 00:30:00,040
And it has government backing.
Quinn 3 is part of China's

504
00:30:00,040 --> 00:30:03,320
broader push for sovereign AI
data localization, model

505
00:30:03,320 --> 00:30:07,000
control, interface ownership.
In this ecosystem, the AI

506
00:30:07,000 --> 00:30:10,680
doesn't just serve users, it
shapes policy implementation,

507
00:30:10,680 --> 00:30:13,360
system optimization, national
productivity.

508
00:30:13,760 --> 00:30:16,800
The interface is no longer just
a tech layer, it's a governance

509
00:30:16,800 --> 00:30:18,880
mechanism.
Google's playing catch up with

510
00:30:18,880 --> 00:30:22,040
Gemini, but they made a
strategic move this week with

511
00:30:22,040 --> 00:30:25,320
Dolphin Gemma, an open
multimodal model tuned for niche

512
00:30:25,320 --> 00:30:28,080
use cases like environmental
data and mobile inference.

513
00:30:28,440 --> 00:30:31,560
It's small, deployable, and fits
inside Pixel phones.

514
00:30:32,000 --> 00:30:35,000
That's not dominance, but it's
placement, and that matters in

515
00:30:35,000 --> 00:30:37,240
the interface game.
Globally, we're seeing

516
00:30:37,240 --> 00:30:40,280
fragmentation.
Latin America mixes Llama based

517
00:30:40,280 --> 00:30:43,920
bots on WhatsApp, Gemini
experiments on Android and open

518
00:30:43,920 --> 00:30:47,600
source agents on Telegram.
India is training Hindi and

519
00:30:47,600 --> 00:30:50,360
Tamil models while deploying
Mistral based flows.

520
00:30:50,840 --> 00:30:53,720
Africa is skipping banks and
going straight to agent LED

521
00:30:53,720 --> 00:30:56,680
finance through messaging apps
and the Middle East.

522
00:30:56,840 --> 00:30:59,520
They're building Sovereign
stacks from scratch.

523
00:30:59,720 --> 00:31:03,520
So here's the real map Open AI
leads in North America.

524
00:31:03,880 --> 00:31:08,800
Meta owns Global Social, Quinn
Three runs China's enterprise

525
00:31:09,000 --> 00:31:13,720
OS, Google's carving niche zones
with Gemini and Dolphin, Gemma.

526
00:31:14,080 --> 00:31:19,120
And in every other region, local
champions are rising faster than

527
00:31:19,120 --> 00:31:23,160
most realize.
This isn't A1 winner game, it's

528
00:31:23,160 --> 00:31:26,280
a distributed land grab for the
new interface layer.

529
00:31:26,400 --> 00:31:29,120
And underneath it all, the same
truth holds.

530
00:31:29,320 --> 00:31:31,960
The interface is power.
It's the entry point to

531
00:31:31,960 --> 00:31:34,840
behavior, trust, and
monetization.

532
00:31:35,240 --> 00:31:38,880
It defines what users see, what
choices they get, and which

533
00:31:38,880 --> 00:31:40,760
systems operate behind the
scenes.

534
00:31:41,240 --> 00:31:44,600
That layer isn't neutral, and
it's being claimed country by

535
00:31:44,600 --> 00:31:47,320
country, company by company.
Segment 6.

536
00:31:47,320 --> 00:31:50,840
Map the Battlefield Segment 7 is
about strategy.

537
00:31:51,280 --> 00:31:53,480
How do you profit from this
fragmentation?

538
00:31:54,120 --> 00:31:57,200
How do you find the edge in a
world built on interfaces?

539
00:31:57,680 --> 00:32:00,480
Let's talk money.
So far we've mapped the war, the

540
00:32:00,480 --> 00:32:03,080
players, the platforms, the
power shifts.

541
00:32:03,560 --> 00:32:06,240
But now it's time to get
practical, because interface

542
00:32:06,240 --> 00:32:09,360
wars don't just create
disruption, they create

543
00:32:09,360 --> 00:32:11,600
leverage.
And if you know where to look,

544
00:32:11,760 --> 00:32:13,840
they create asymmetric
opportunities.

545
00:32:14,280 --> 00:32:17,760
This isn't just a TET
transformation, it's a new value

546
00:32:17,760 --> 00:32:21,120
layer and that means there are
ways to profit whether you're

547
00:32:21,120 --> 00:32:24,520
building, investing or
positioning for what comes next.

548
00:32:25,040 --> 00:32:26,680
Let's.
Start with the core principle.

549
00:32:26,800 --> 00:32:30,800
Don't bet on the smartest model,
Bet on the stickiest interface.

550
00:32:31,240 --> 00:32:34,280
Intelligence changes.
Interfaces persist.

551
00:32:34,760 --> 00:32:38,280
If you'd invested in the best
algorithm in 2010, you wouldn't

552
00:32:38,280 --> 00:32:41,040
have picked Facebook.
You'd have picked Google Plus.

553
00:32:41,520 --> 00:32:45,520
But Facebook owned attention
owned interaction and that's

554
00:32:45,520 --> 00:32:47,760
what 1.
The same thing applies here.

555
00:32:47,880 --> 00:32:51,080
It's not about which model
scores highest, it's about which

556
00:32:51,080 --> 00:32:55,240
interface captures the most time
trust and default behavior.

557
00:32:55,400 --> 00:32:59,240
So where's the edge right now?
First, vertical agents.

558
00:32:59,640 --> 00:33:03,240
We're entering an era of hyper
specialized interfaces, AI

559
00:33:03,240 --> 00:33:05,240
layers that don't try to do
everything.

560
00:33:05,640 --> 00:33:08,080
They do one thing extremely
well.

561
00:33:08,520 --> 00:33:12,600
Booking travel, handling
invoices, recommending gear,

562
00:33:13,080 --> 00:33:16,960
closing sales.
These agents live inside LLMS,

563
00:33:16,960 --> 00:33:20,400
fine-tuned with domain specific
memory and monetized through

564
00:33:20,400 --> 00:33:23,160
performance.
They're lean, focused, and

565
00:33:23,160 --> 00:33:26,240
designed for trust.
The best ones will feel like

566
00:33:26,240 --> 00:33:30,120
apps, but they won't be apps.
They'll be interfaces built

567
00:33:30,120 --> 00:33:34,240
entirely on AI rails.
Second, agent infrastructure.

568
00:33:34,280 --> 00:33:37,800
About agents, but very fewer
looking at the middleware, the

569
00:33:37,800 --> 00:33:42,160
rails that let agents talk to
apps, call APIs, track tasks,

570
00:33:42,160 --> 00:33:44,720
cache memory.
That's where the real money is.

571
00:33:45,120 --> 00:33:49,280
Lane chain crew, AI, autogen
tool, former open agents.

572
00:33:49,480 --> 00:33:52,040
These are the early primitives
of the agent economy.

573
00:33:52,160 --> 00:33:55,360
Think of them as the Stripe or
Twilio of AI.

574
00:33:55,480 --> 00:33:58,480
Low level, essential and
increasingly valuable

575
00:33:58,840 --> 00:34:02,240
infrastructure always wins
during a platform shift, and

576
00:34:02,360 --> 00:34:05,120
agents are the shift.
Third, the context.

577
00:34:05,120 --> 00:34:09,639
Layer interfaces are only as
good as the memory they access.

578
00:34:09,920 --> 00:34:13,920
That means vector databases,
context caching systems,

579
00:34:14,159 --> 00:34:18,800
semantic retrieval engines.
We're talking Pine cone YV 8

580
00:34:18,880 --> 00:34:22,760
Chroma DB.
As AI interfaces scale, they'll

581
00:34:22,760 --> 00:34:27,000
need to recall billions of
tokens across time, task, and

582
00:34:27,000 --> 00:34:29,920
tone.
Whoever solves long context

583
00:34:29,920 --> 00:34:33,280
retrieval at scale will own a
key part of the stack.

584
00:34:33,679 --> 00:34:37,560
This isn't just back end, it's
part of the new BRAIN 4th

585
00:34:37,679 --> 00:34:40,280
distribution.
Modes don't just look at tech,

586
00:34:40,360 --> 00:34:43,040
look at surface area.
Platforms like WhatsApp,

587
00:34:43,199 --> 00:34:46,880
Instagram, Android and iMessage
are interface gold mines.

588
00:34:47,360 --> 00:34:50,400
If an AI can live inside those
environments, it doesn't need to

589
00:34:50,400 --> 00:34:52,360
be the best, it just needs to be
there.

590
00:34:52,760 --> 00:34:56,440
This is why Meta is dangerous.
It's not about Llama 4 beating

591
00:34:56,440 --> 00:35:00,520
GPT 4, it's about Llama 4
showing up inside a billion

592
00:35:00,520 --> 00:35:04,520
conversations before GPT 4 even
loads in.

593
00:35:04,520 --> 00:35:09,240
Interface Wars distribution is
Destiny 5th Regional.

594
00:35:09,320 --> 00:35:13,640
AI winners Global interface
dominance is unlikely.

595
00:35:13,920 --> 00:35:19,480
Instead, expect localized
champions Quinn in China, Cohere

596
00:35:19,480 --> 00:35:24,240
in Canada, Mistrawn in Europe,
Sarvam in India.

597
00:35:24,680 --> 00:35:28,400
Investors who can spot the
interface Native winners in

598
00:35:28,400 --> 00:35:32,520
these regions will gain exposure
to markets that global giants

599
00:35:32,520 --> 00:35:37,600
can't access easily.
Think sovereign AIUX.

600
00:35:37,800 --> 00:35:41,800
It's coming and it will fragment
the opportunity set in powerful

601
00:35:41,800 --> 00:35:44,760
ways and 6th behavior.
Data pipelines.

602
00:35:45,040 --> 00:35:48,440
The most overlooked play in this
entire shift is behavioral

603
00:35:48,440 --> 00:35:51,280
intelligence.
Interfaces generate high signal

604
00:35:51,280 --> 00:35:53,720
data.
What people ask, when they ask

605
00:35:53,720 --> 00:35:57,160
it, how they respond.
That data is gold for product

606
00:35:57,160 --> 00:35:59,640
design, ad targeting, agent
refinement.

607
00:36:00,040 --> 00:36:03,280
Companies that build ethical,
scalable ways to harness this

608
00:36:03,280 --> 00:36:05,600
behavioral layer will own the
feedback loop.

609
00:36:05,800 --> 00:36:08,120
Not just prediction, but
iteration.

610
00:36:08,600 --> 00:36:11,040
And that's where real edge
compounds.

611
00:36:11,480 --> 00:36:14,800
Let's pull it together.
The mistake most people make is

612
00:36:14,800 --> 00:36:17,960
looking for an AI stock or a
model to buy.

613
00:36:18,640 --> 00:36:22,240
But the real question is who's
building the interfaces people

614
00:36:22,240 --> 00:36:25,520
will actually live in?
Who's capturing the first

615
00:36:25,520 --> 00:36:28,560
action, the first question, the
first interaction?

616
00:36:29,040 --> 00:36:32,760
That's where monetization
happens, that's where platforms

617
00:36:32,760 --> 00:36:35,840
get built, and that's where the
leverage is hiding whether

618
00:36:35,840 --> 00:36:38,600
you're a founder.
A fund or just someone paying

619
00:36:38,600 --> 00:36:41,320
attention.
This is the moment to shift your

620
00:36:41,320 --> 00:36:43,560
frame.
AI isn't just a tech

621
00:36:43,560 --> 00:36:45,640
breakthrough, it's a new user
layer.

622
00:36:45,920 --> 00:36:49,480
One that collapses the OP stack,
rewires behavior, and creates

623
00:36:49,480 --> 00:36:52,200
entirely new value chains.
The winners?

624
00:36:52,560 --> 00:36:54,920
They won't just be smart,
they'll be embedded.

625
00:36:55,240 --> 00:36:58,440
Trusted Habitual segment 7 gave
you.

626
00:36:58,440 --> 00:37:02,200
The strategies, but segment 8,
that's where we go full builder

627
00:37:02,200 --> 00:37:05,280
mode because if you want to
create something inside this new

628
00:37:05,280 --> 00:37:08,440
interface world, there's a wide
open lane right now, one that

629
00:37:08,440 --> 00:37:11,920
open AI hasn't monetized yet.
Let's talk about how to build

630
00:37:11,920 --> 00:37:15,000
the next AI shopping agent and
make money doing it.

631
00:37:15,160 --> 00:37:18,800
Let's get tactical because this
isn't just a shift you can

632
00:37:18,800 --> 00:37:21,640
watch.
It's one you can build into Chad

633
00:37:21,640 --> 00:37:26,240
GPTS new shopping layer.
It's sleek, intuitive, and trust

634
00:37:26,240 --> 00:37:28,000
based.
But here's what most people

635
00:37:28,000 --> 00:37:30,960
don't realize.
Open AI hasn't monetized it yet.

636
00:37:31,240 --> 00:37:35,720
No affiliate programs, no brand
integrations, no developer

637
00:37:35,720 --> 00:37:38,760
extensions.
That means the layer is live,

638
00:37:38,760 --> 00:37:42,680
but the business ecosystem
around it is wide open for now.

639
00:37:42,840 --> 00:37:45,160
Which makes this the perfect.
Window for builders.

640
00:37:45,480 --> 00:37:48,520
If you're a solo founder, Andy,
hacker, or product minded

641
00:37:48,520 --> 00:37:50,520
engineer, the opportunity is
simple.

642
00:37:50,720 --> 00:37:54,600
Build a vertical agent that sits
on top of existing LLMS and

643
00:37:54,600 --> 00:37:58,240
handles a specific shopping
category better than ChatGPT can

644
00:37:58,240 --> 00:38:01,800
do out-of-the-box.
Think headphones, running shoes,

645
00:38:02,200 --> 00:38:05,360
ergonomic office chairs.
The trick isn't to be general,

646
00:38:05,480 --> 00:38:07,760
it's to be expert.
The tech stack.

647
00:38:07,760 --> 00:38:10,120
Is light.
You don't need to build your own

648
00:38:10,120 --> 00:38:14,400
model.
Just use GPT 4 O or Claude 3.5

649
00:38:14,440 --> 00:38:17,480
as the reasoning engine.
Wrap it in a short context,

650
00:38:17,480 --> 00:38:22,040
layer curated reviews, specs,
user types and expose it through

651
00:38:22,040 --> 00:38:24,440
a clean UI.
Add affiliate pipelines through

652
00:38:24,440 --> 00:38:28,040
Amazon, niche vendors or high
margin DTC brands.

653
00:38:28,720 --> 00:38:31,480
That's it.
You've built a monetizable,

654
00:38:31,480 --> 00:38:36,200
memory enabled AI shopping
interface with trust baked in.

655
00:38:36,400 --> 00:38:39,560
Bonus points if you go deep.
On Workflow, let users upload

656
00:38:39,560 --> 00:38:43,120
photos of their space, describe
their needs, or set constraints.

657
00:38:43,520 --> 00:38:47,520
Use vision, price filtering,
style guides, and feedback loops

658
00:38:47,520 --> 00:38:50,720
to refine results.
Offer, save, share e-mail

659
00:38:50,720 --> 00:38:54,120
options with memory tracking.
Every layer you add increases

660
00:38:54,120 --> 00:38:57,600
trust and repeat use.
Don't just help users choose,

661
00:38:57,960 --> 00:39:00,400
help them decide.
That's the future of shopping

662
00:39:00,560 --> 00:39:04,080
and the monetization is.
Clean affiliate commissions,

663
00:39:04,560 --> 00:39:08,320
premium referrals, sponsored
results if you want them.

664
00:39:08,960 --> 00:39:12,160
But here's the real power When
your agent starts helping 1000

665
00:39:12,160 --> 00:39:15,400
users per month make purchase
decisions, you don't need ads.

666
00:39:15,880 --> 00:39:19,560
You've built intent gravity, a
flow that drives commerce.

667
00:39:20,280 --> 00:39:24,800
And once you have that, you can
move up market into B to B, into

668
00:39:24,800 --> 00:39:28,480
search partnerships, into your
own product lines.

669
00:39:28,600 --> 00:39:32,040
You don't need to scale like.
Amazon, you just need to own a

670
00:39:32,040 --> 00:39:35,600
niche before Open AI decides to
monetize the general case.

671
00:39:36,160 --> 00:39:40,560
Right now, ChatGPT helps users
shop, but it doesn't go deep,

672
00:39:41,040 --> 00:39:44,360
doesn't fine tune, doesn't
optimize by lifestyle or use

673
00:39:44,360 --> 00:39:46,000
case.
That's your window.

674
00:39:46,320 --> 00:39:48,840
Six months from now, it might be
gone.

675
00:39:49,240 --> 00:39:52,320
So here's the play.
Pick a vertical launch fast and

676
00:39:52,320 --> 00:39:54,440
tie into a high conversion back
end.

677
00:39:55,120 --> 00:39:58,960
Let the model do the work, let
the interface carry the trust,

678
00:39:59,240 --> 00:40:01,840
and let the decision loop drive
the value.

679
00:40:02,440 --> 00:40:05,600
This isn't about building the
next big platform, it's about

680
00:40:05,640 --> 00:40:07,840
owning the layer where choice
gets made.

681
00:40:08,000 --> 00:40:11,960
That's the edge.
One prompt, one agent, 1 moment

682
00:40:11,960 --> 00:40:13,960
of trust, and the purchase is
done.

683
00:40:14,360 --> 00:40:16,840
If you want to make money in the
interface economy, don't wait

684
00:40:16,840 --> 00:40:20,000
for permission.
Build something useful right now

685
00:40:20,360 --> 00:40:24,080
before the layer closes.
Subscribe to Finance Frontier.

686
00:40:24,160 --> 00:40:28,960
AI on Spotify or Apple Podcasts
Follow us on X to track the

687
00:40:28,960 --> 00:40:31,440
biggest AI stories shaping the
world.

688
00:40:31,960 --> 00:40:35,720
Share this episode with a friend
and help us hit 10,000 downloads

689
00:40:35,720 --> 00:40:38,440
as we build the smartest AI
community online.

690
00:40:38,600 --> 00:40:42,480
We cover AI, innovation,
infrastructure and intelligence

691
00:40:42,640 --> 00:40:47,040
across 4 series, all grouped at
financefrontierai.com.

692
00:40:47,320 --> 00:40:50,560
And if your company or idea fits
one of our themes, you may

693
00:40:50,560 --> 00:40:54,360
qualify for a free spotlight.
Just head to the pitch page and

694
00:40:54,360 --> 00:40:56,440
take a look.
Sign up for the 10 Times.

695
00:40:56,440 --> 00:40:59,880
Out our weekly drop of AI
business ideas you can actually

696
00:40:59,880 --> 00:41:02,040
use.
Each one's tied to a real

697
00:41:02,040 --> 00:41:04,760
breakthrough new tools, models
and trends.

698
00:41:04,760 --> 00:41:06,680
We catch early if you're
building with

699
00:41:06,760 --> 00:41:13,520
aithisiswhereyouredgebeginsonly@financefrontierai.com.
This podcast is for educational

700
00:41:13,520 --> 00:41:17,440
purposes only, not financial
advice, legal advice or model

701
00:41:17,440 --> 00:41:20,800
development guidance.
Always verify before you build,

702
00:41:20,800 --> 00:41:23,920
deploy or invest.
The AI landscape is.

703
00:41:23,920 --> 00:41:28,240
Changing fast, benchmarks
evolve, regulation shift, and

704
00:41:28,240 --> 00:41:30,720
what's true today may not hold
tomorrow.

705
00:41:31,120 --> 00:41:34,720
Use every insight here as a
lens, not a conclusion.

706
00:41:34,880 --> 00:41:38,600
Today's music, including.
Our intro and outro track, Night

707
00:41:38,600 --> 00:41:42,200
Runner by Audionautics, is
licensed under the YouTube Audio

708
00:41:42,200 --> 00:41:45,320
Library license.
Additional tracks are licensed

709
00:41:45,320 --> 00:41:49,000
under Creative Commons, and full
details can be found in the

710
00:41:49,000 --> 00:41:51,840
episode description.
Copyright 2025.

711
00:41:51,840 --> 00:41:54,960
Finance Frontier AI.
All rights reserved.

712
00:41:55,320 --> 00:41:58,480
Reroduction, distribution, or
transmission of this episode's

713
00:41:58,480 --> 00:42:00,760
content without written
ermission is strictly

714
00:42:00,760 --> 00:42:02,920
prohibited.
Thanks for listening and we'll

715
00:42:02,920 --> 00:42:03,640
see you next time.