WEBVTT
1
00:00:00.051 --> 00:00:06.715
Hi, and welcome to Technically Creative's Week in Review, brought to you by Kubrick Labs, the AI and IT partner for creative companies.
2
00:00:06.755 --> 00:00:13.159
Every Monday, I spend 20 minutes giving you an overview of the most important news from the increasingly interdependent worlds of tech and creativity.
3
00:00:13.399 --> 00:00:15.920
This is the Week in Review, October 5th, 2026.
4
00:00:16.280 --> 00:00:19.942
For the last few years, most of the AI conversation has been about capability.
5
00:00:20.343 --> 00:00:20.903
Can it write?
6
00:00:21.103 --> 00:00:21.984
Can it code?
7
00:00:22.064 --> 00:00:24.085
Can it hallucinate a Supreme Court case?
8
00:00:24.405 --> 00:00:31.747
That conversation isn't over, but I think we crossed into a more important one this week because the newest AI systems want to act on your behalf.
9
00:00:32.067 --> 00:00:33.828
Yeah, I'm talking about agents.
10
00:00:34.388 --> 00:00:38.689
OpenAI launched Dots, its new always-on agents inside GPT.
11
00:00:39.149 --> 00:00:45.431
Give a dot a goal and the permissions it needs, and it can keep working after you leave the conversation using its own cloud.
12
00:00:45.591 --> 00:00:48.031
Meta is pursuing much the same idea with Muse.
13
00:00:48.211 --> 00:00:54.013
Muse was downloaded 5 million times in 22 days, making it one of the fastest-growing iPhone apps ever.
14
00:00:54.313 --> 00:00:55.254
This is a big deal.
15
00:00:55.514 --> 00:00:57.616
Generative AI gave machines capability.
16
00:00:57.956 --> 00:01:00.158
Agentic AI gives machines authority.
17
00:01:00.418 --> 00:01:03.721
And as you'll see, authority is a much weirder problem to tackle.
18
00:01:04.082 --> 00:01:15.131
As a tech creator named Matt Robb discovered just this last week, in possibly the most perfect way, Robb had been using Meta's Muse to help with a sale through Facebook Marketplace.
19
00:01:15.752 --> 00:01:20.956
At some point, he'd given Muse his home address and selected an allow always permission.
20
00:01:21.677 --> 00:01:27.563
Muse subsequently shared that address with a buyer and arranged a pickup without Rob realizing what was happening.
21
00:01:27.663 --> 00:01:31.527
And then the buyer arrived at his house without Rob being any the wiser.
22
00:01:31.647 --> 00:01:34.029
I would say he's been doxed by his own petard.
23
00:01:34.770 --> 00:01:39.915
But Meta's position on this is that Muse acted within the permissions Rob had granted, and that's kind of fair.
24
00:01:40.215 --> 00:01:46.281
But did Rob know what he was doing, or did he just want to stop getting pestered with access requests every 60 seconds?
25
00:01:46.741 --> 00:01:47.301
Who knows?
26
00:01:47.581 --> 00:01:52.404
We're moving from a world where the main question was, did AI follow the rules?
27
00:01:52.484 --> 00:01:57.706
To a harder question, which is, did the human even understand the rules it was giving the AI?
28
00:01:58.206 --> 00:02:06.370
Anyone who's clicked accept all on the T's and C's of any software product should worry a little bit about our species readiness for this moment.
29
00:02:06.970 --> 00:02:12.674
Meta says Muse is designed to seek approval before sensitive actions, such as sending emails or making purchases.
30
00:02:13.074 --> 00:02:15.916
And the company provides an audit trail of what the agent's done.
31
00:02:16.316 --> 00:02:16.636
Okay.
32
00:02:17.256 --> 00:02:21.479
But following Rob's experience, Meta also said it was improving the permission experience.
33
00:02:21.679 --> 00:02:22.560
Good, I guess.
34
00:02:23.440 --> 00:02:26.702
And Meta is both culpable and sort of not culpable in this.
35
00:02:27.083 --> 00:02:32.306
And we're going to come back to Rob, because one of the ways in which I think everybody is culpable is this.
36
00:02:32.386 --> 00:02:38.150
The part that really amuses me is how adorable OpenAI and Meta have made these powerful agents.
37
00:02:38.811 --> 00:02:44.855
OpenAI's dots are sweet, little fuzzy, adorable emojis, and Meta's muse is designed to look like a Teletubby.
38
00:02:45.475 --> 00:02:54.339
It's an interesting aesthetic choice for software that knows where you live, reads your emails, makes purchases with your credit card, and communicates with strangers without you knowing.
39
00:02:54.719 --> 00:02:56.320
Of course we gave it a little fuzzy face.
40
00:02:56.820 --> 00:02:58.781
Japan has been doing versions of this for years.
41
00:02:58.901 --> 00:03:05.184
Railway offices and tax offices routinely wrap very serious institutions in absurdly cute little mascots.
42
00:03:05.844 --> 00:03:08.566
Authority looks like it belongs on a scratch-and-sniff sticker.
43
00:03:09.426 --> 00:03:14.690
Unfamiliar technology is intimidating, and good design makes complicated products feel approachable.
44
00:03:14.810 --> 00:03:15.331
I buy that.
45
00:03:16.071 --> 00:03:18.353
But the contrast is something I want to point out.
46
00:03:18.873 --> 00:03:22.876
The more agency the software gets, the cuter the AI companies need to make it.
47
00:03:22.896 --> 00:03:23.997
It's almost a rule.
48
00:03:24.518 --> 00:03:28.681
If you see the cutest little software mascot you've ever seen, be afraid.
49
00:03:29.041 --> 00:03:30.962
Be so, so afraid.
50
00:03:30.983 --> 00:03:31.083
Okay.
51
00:03:31.783 --> 00:03:37.187
It can access your files, your inbox, and your home address, but don't worry, it's shaped like a jelly bean.
52
00:03:37.387 --> 00:03:40.108
There's something important underneath that flippant joke.
53
00:03:40.549 --> 00:03:42.790
Approachable isn't the same thing as harmless.
54
00:03:43.170 --> 00:03:51.676
A friendly visual language can make an unfamiliar system seem easier to use, but it can also make granting that system enormous authority feel trivial.
55
00:03:52.216 --> 00:03:52.717
It's not.
56
00:03:53.457 --> 00:03:56.060
Allow always looks like a harmless little button.
57
00:03:56.361 --> 00:03:58.923
And check out how they're softening the font and shape on that too.
58
00:03:59.204 --> 00:04:01.246
My friends in design will know what I'm talking about.
59
00:04:01.566 --> 00:04:02.467
It's not harmless.
60
00:04:02.487 --> 00:04:03.629
You gotta pay attention.
61
00:04:03.969 --> 00:04:05.290
Yeah, the witch with the candy?
62
00:04:05.671 --> 00:04:08.013
She's gonna do horrible things to those little German kids.
63
00:04:08.634 --> 00:04:11.037
The rule you learned in kindergarten still applies.
64
00:04:11.517 --> 00:04:13.780
Be careful anytime anyone offers you candy.
65
00:04:14.448 --> 00:04:18.890
And Meta may not be the only company trying to help Rob from running into that mistake again.
66
00:04:19.110 --> 00:04:21.091
Apple wants to protect them at the compute level.
67
00:04:21.311 --> 00:04:33.496
The company says that it's changing how macOS handles full disk access, warning that some apps have been using it in ways that can expose files, mail, messages, and browsing history without users fully understanding the access they've granted.
68
00:04:33.536 --> 00:04:34.596
We're looking at you, Rob.
69
00:04:34.816 --> 00:04:39.378
Apple explicitly linked the change to increasingly capable and autonomous AI agents.
70
00:04:39.618 --> 00:04:41.039
I think they were talking about Muse.
71
00:04:41.279 --> 00:04:46.921
For a creative company, full disk access sounds technical until you replace it with what's sitting on that disk.
72
00:04:47.261 --> 00:04:57.684
Unreleased scripts, rough cuts, treatments, campaign strategy, client correspondence, budgets, talent agreements, design files, Dropbox folders, pitch decks, contracts.
73
00:04:58.044 --> 00:05:02.025
And especially if your company is mostly remote, that data is your company.
74
00:05:02.185 --> 00:05:04.426
And here's the basic tension at the bottom of all this.
75
00:05:04.866 --> 00:05:09.488
An AI agent becomes more useful precisely because of the access you give it.
76
00:05:10.089 --> 00:05:16.512
An agent that can't see your email, calendar, files, or business systems is a brilliant employee locked outside of the office.
77
00:05:16.712 --> 00:05:21.615
That's why I think the next few years of agent design may be less about raw intelligence than permission architecture.
78
00:05:22.075 --> 00:05:27.062
It goes back to the governance and approving layer we bore our clients with about every day.
79
00:05:27.242 --> 00:05:30.166
Yes, you need to think about stronger IT systems now.
80
00:05:30.306 --> 00:05:30.787
Shocker.
81
00:05:31.227 --> 00:05:36.774
And this agentic takeover of your life is about to hit advertising because of course it's about to hit advertising.
82
00:05:36.835 --> 00:05:37.415
What doesn't?
83
00:05:38.056 --> 00:05:42.661
Imagine telling your agent, find me the best hotel in New York for under $600 a night.
84
00:05:43.021 --> 00:05:44.783
Today, a hotel tries to persuade you.
85
00:05:44.843 --> 00:05:53.851
It buys media, optimizes search, pays influencers, and runs beautiful films of attractive people staring lustfully at each other across an infinity pool.
86
00:05:54.392 --> 00:06:01.795
But if the agent increasingly handles discovery, comparison, and recommendation, the hotel also has to become legible to the agent.
87
00:06:02.215 --> 00:06:03.376
Can the software find the room?
88
00:06:03.476 --> 00:06:05.096
Can it compare cancellation policies?
89
00:06:05.316 --> 00:06:06.457
Does it trust the reviews?
90
00:06:06.697 --> 00:06:14.160
And advertising starts to become mostly about persuading not you or people, but AI agents that this is the best deal for their user.
91
00:06:14.880 --> 00:06:18.341
Meanwhile, the industry is building agents on the other side of the transaction too.
92
00:06:18.521 --> 00:06:28.743
Media buyers, publishers, and ad tech companies are experimenting with autonomous systems that can buy inventory, manage campaigns, and optimize price or price gouge based on buying intent.
93
00:06:29.223 --> 00:06:40.926
Which means we're approaching the delightful possibility that your AI will someday go online to negotiate with Procter & Gamble's AI about which toothpaste you should buy while neither you nor Procter & Gamble is paying much attention.
94
00:06:41.570 --> 00:06:43.292
It's a great moment for humanity.
95
00:06:43.933 --> 00:06:46.816
The agent era is asking us something uncomfortable but important.
96
00:06:47.156 --> 00:06:49.458
How much of your judgment are you prepared to delegate?
97
00:06:49.859 --> 00:06:52.942
And that's a question about humans as much as it's a question about machines.
98
00:06:53.523 --> 00:06:55.204
So that's it for our deep dive this week.
99
00:06:55.565 --> 00:06:58.368
So now we get into some other news from the world of tech and creativity.
100
00:06:58.628 --> 00:07:04.554
So Rich Silverstein is stepping away from the day-to-day running of Goodby Silverstein and Partners after 43 years.
101
00:07:05.094 --> 00:07:15.785
The agency marked the occasion with a GPT Rich, an AI recreation built from decades of material and paired with a synthetic voice so people can still receive Rich-style creative feedback.
102
00:07:15.945 --> 00:07:18.568
It's a clever leaving idea, and the video was very sweet.
103
00:07:18.748 --> 00:07:23.934
It is a joke, and I do get that it's a joke, but the symbolism is potent, and I'm going to run with it a little bit.
104
00:07:24.115 --> 00:07:27.238
AI can mimic competency incredibly well, joke or not.
105
00:07:27.599 --> 00:07:33.225
An AI-rich Silverstein is probably pretty competent, and it zeroes in on exactly the thing I care about most right now.
106
00:07:33.866 --> 00:07:36.570
I don't care whether AI can imitate Rich Silverstein.
107
00:07:37.091 --> 00:07:39.795
I want to know what Rich Silverstein would do with AI.
108
00:07:40.295 --> 00:07:43.900
And that's why I'm sad he's leaving, even though I never worked with or for him.
109
00:07:44.161 --> 00:07:50.510
Because when an industry goes through a change this big, I want the people with the deepest craft and the most history leading from the front.
110
00:07:50.710 --> 00:07:54.952
Rich still sits at the desk that he bought when he started at Parsons School of Design.
111
00:07:55.412 --> 00:07:59.174
There are X-Acto knives and a bowl of colored pencils behind him.
112
00:07:59.674 --> 00:08:05.877
His creative life stretches back through Rolling Stone and physical typography, art direction, and decades of advertising.
113
00:08:06.217 --> 00:08:07.278
He's old school, man.
114
00:08:07.998 --> 00:08:12.079
But he also isn't standing on a hill yelling at the clouds and talking about the good old days.
115
00:08:12.519 --> 00:08:16.880
His recent 13 Days Cuba Missile Crisis musical incorporated AI-assisted work.
116
00:08:17.280 --> 00:08:17.580
Good.
117
00:08:17.960 --> 00:08:18.600
Keep going.
118
00:08:18.921 --> 00:08:21.141
Because we need people like Rich now more than ever.
119
00:08:21.621 --> 00:08:26.242
Rich's value isn't in the archives of things that he's made that AI is going to increasingly train on.
120
00:08:26.842 --> 00:08:30.883
It's 40 years of references, mistakes, taste, and instinct.
121
00:08:31.383 --> 00:08:37.585
The more history somebody carries, the more valuable it is when they show the rest of us what deserves to survive in the next era.
122
00:08:38.105 --> 00:08:40.186
So congratulations on GPT Rich.
123
00:08:40.226 --> 00:08:43.546
It really is such a warm and funny thing that's so current.
124
00:08:44.527 --> 00:08:49.608
But if there's a thing that I actually want to subscribe to, it's rich with GPT, not GPT Rich.
125
00:08:49.988 --> 00:08:54.889
And then speaking about creative leadership in AI, Ben Affleck continued last week's theme rather beautifully.
126
00:08:55.369 --> 00:09:01.631
At Bloomberg's Scream Time, Affleck said he used AI for lots of things while making his new Netflix film Animals.
127
00:09:02.071 --> 00:09:07.452
He declined to identify exactly where, arguing that pulling apart the mechanics of filmmaking can undermine the illusion.
128
00:09:08.272 --> 00:09:09.713
Nice save, Ben.
129
00:09:10.013 --> 00:09:11.514
That man could run for president.
130
00:09:11.934 --> 00:09:16.636
More importantly, though, he described the film as very human and framed the AI use as responsible.
131
00:09:17.156 --> 00:09:28.842
Interpositive, the post-production company he started that was bought by Netflix, was designed as a discrete post-production tool working on footage filmmakers had already created rather than generating a movie from scratch.
132
00:09:29.342 --> 00:09:39.567
The company builds its own controlled training data rather than scraping the internet, and Affleck has consistently emphasized licensed material, consent, human creative control, and preserving filmmakers' judgment.
133
00:09:40.127 --> 00:09:47.730
And that's what I want from people with experience, showing us, in practice, where they think the technology belongs and what value should govern its use.
134
00:09:48.250 --> 00:09:50.471
Because then the rest of us have something to learn from.
135
00:09:50.571 --> 00:09:54.113
We can adopt the framework, argue with it, improve it, or reject parts of it.
136
00:09:54.533 --> 00:09:56.715
but at least someone credible is leading from the front.
137
00:09:57.135 --> 00:10:02.379
That's way more useful than another panel asking whether Hollywood is for or against AI.
138
00:10:02.579 --> 00:10:07.842
And then in other news, everybody's freaking out about Skydance, but as usual, they're paying attention to the wrong thing.
139
00:10:08.383 --> 00:10:14.107
People are losing their minds over the fact that the combined Paramount-Warner Brothers company is going to sit under the Skydance name.
140
00:10:14.347 --> 00:10:16.588
And I don't really think that's the part to pay attention to.
141
00:10:17.088 --> 00:10:18.209
Sony bought Columbia.
142
00:10:18.249 --> 00:10:19.369
Columbia didn't disappear.
143
00:10:19.729 --> 00:10:21.730
Tristar didn't disappear, kinda.
144
00:10:22.190 --> 00:10:29.894
Those studio identities carried on perfectly happy under a parent company called Sony, which, if we're being honest, is a pretty shitty name for a movie studio.
145
00:10:30.234 --> 00:10:32.555
And I'll say something nobody else seems to be willing to say.
146
00:10:33.235 --> 00:10:34.816
Skydance is a badass name.
147
00:10:35.656 --> 00:10:39.897
It also has one of the best vanity cards in the business, that epic fireball in space.
148
00:10:39.957 --> 00:10:40.517
Come on.
149
00:10:40.997 --> 00:10:42.958
It's up there with the other two best in the business.
150
00:10:42.998 --> 00:10:44.898
And if you're wondering, that's Disney and Universal.
151
00:10:45.519 --> 00:10:51.640
So if your complaint about this merger is that the word Skydance is going to appear on some corporate paperwork, I think we can all survive.
152
00:10:52.220 --> 00:10:55.461
But the story is that Skydance was smaller than Paramount and took it over.
153
00:10:55.981 --> 00:10:58.942
Then Paramount is smaller than Warner Brothers, and it's taking that over.
154
00:10:59.502 --> 00:11:02.104
That's one hell of a corporate ladder to climb in two years.
155
00:11:02.665 --> 00:11:05.807
And the hires tell us something about what David Ellison wants to build.
156
00:11:05.907 --> 00:11:12.293
Ynon Kreitz, and I apologize if I screwed up that pronunciation, from Mattel is coming in as co-CEO.
157
00:11:12.993 --> 00:11:14.655
Mattel isn't his whole story, though.
158
00:11:14.675 --> 00:11:17.016
He was at Fox, Makers, and Endemol before that.
159
00:11:17.357 --> 00:11:18.938
So he has entertainment experience.
160
00:11:19.278 --> 00:11:28.126
Dana Goldberg and Josh Greenstein, the current heads of Paramount, are expected to take broader control of the film operation, with Pam Abdi and Mike DeLuca out at WBD.
161
00:11:28.766 --> 00:11:32.447
Diane Glasgow came over from meta Facebook to a senior product role.
162
00:11:32.788 --> 00:11:35.769
Taken together, this doesn't just look like bringing some tech people.
163
00:11:36.149 --> 00:11:42.331
It looks like an attempt to turn a collection of legacy media businesses into one more centralized, product-driven company.
164
00:11:42.831 --> 00:11:49.094
Shared infrastructure, shared technology, shared data, fewer overlapping power centers, and more centralized decision-making.
165
00:11:49.414 --> 00:11:57.697
And that could be good, but it's troubling that the systems might be governed by the people operating the smaller companies, not the bigger company.
166
00:11:59.148 --> 00:12:05.150
But listen, there are absolutely systems inside legacy Hollywood companies that deserve to be dragged into this century.
167
00:12:05.790 --> 00:12:10.572
But Warner Brothers is where the rubber meets the road, because creatively, quite a lot has been working.
168
00:12:11.072 --> 00:12:16.814
It backed distinctive filmmakers, original theatrical movies, and serious creative swings at a meaningful scale.
169
00:12:17.154 --> 00:12:24.897
And that culminated in an extraordinary 2025 run at the box office and during awards season, including the Best Picture Oscar for one battle after another.
170
00:12:25.477 --> 00:12:26.698
But Mike and Pam out?
171
00:12:26.718 --> 00:12:27.058
Pfft.
172
00:12:27.618 --> 00:12:29.359
Maybe that's just Hollywood being Hollywood.
173
00:12:29.419 --> 00:12:31.921
Twelve months can take you from genius to unemployed.
174
00:12:32.422 --> 00:12:35.284
But maybe Ellicent wants his own people and his own operating model.
175
00:12:35.704 --> 00:12:37.765
Probably both.
176
00:12:38.326 --> 00:12:40.207
But the hires make the bigger question clearer.
177
00:12:40.587 --> 00:12:43.750
Can you centralize the machinery without centralizing the taste?
178
00:12:44.190 --> 00:12:45.571
That's the thing to keep an eye on.
179
00:12:46.031 --> 00:12:47.212
Modernize the technology.
180
00:12:47.272 --> 00:12:48.493
Modernize the infrastructure.
181
00:12:48.593 --> 00:12:49.674
Fix the bureaucracy.
182
00:12:49.774 --> 00:12:50.534
Use better data.
183
00:12:50.575 --> 00:12:51.615
Make company faster.
184
00:12:51.835 --> 00:12:52.156
Sure.
185
00:12:52.716 --> 00:12:58.023
But don't turn Warner Brothers into just another application running on top of Skydance's operating systems.
186
00:12:58.644 --> 00:13:01.888
Because Warner's value isn't in the library, the lot, or the logo.
187
00:13:02.429 --> 00:13:08.397
It's the fact that generations of filmmakers have believed that Warner Brothers was a place that would let them make unique, original movies.
188
00:13:08.677 --> 00:13:14.462
The thing everyone says Hollywood needs more of, less IP, more original voices, that's what Warner's does.
189
00:13:15.302 --> 00:13:18.965
2025 was weapons and sinners in one battle after another.
190
00:13:19.505 --> 00:13:24.209
Original, artist-driven, no-IP-in-sight work that's special and makes money.
191
00:13:24.429 --> 00:13:28.833
Of course they do the IP-driven stuff as well, but it's worth noting that Warner's Walks the Walk.
192
00:13:29.513 --> 00:13:34.557
So if Ellison can preserve that while making the machinery underneath it dramatically better, fantastic.
193
00:13:35.158 --> 00:13:41.683
If he optimizes away the thing that makes Warner special, however, that can be an extremely modern way to screw up a wonderful company.
194
00:13:42.140 --> 00:13:45.741
And then in legal news, copyright wins one and antitrust loses one.
195
00:13:46.041 --> 00:13:48.282
We got two useful legal signals this week.
196
00:13:48.502 --> 00:13:58.584
The Third Circuit upheld Thomson Reuters' copyright victory against Ross Intelligence, making this the first U.S. federal appellate decision to address fair use in an AI-related context.
197
00:13:58.724 --> 00:14:05.946
Ross had used material derived from Westlaw headnotes to build a competing legal research product, and the court held that the use was not fair—
198
00:14:06.426 --> 00:14:15.171
Ross was not training a contemporary generative LLM, and the factual circumstances differ materially from the major cases against OpenAI, Anthropic, and Meta.
199
00:14:15.752 --> 00:14:19.554
So anyone posting fair use is dead should perhaps wait until after breakfast.
200
00:14:20.054 --> 00:14:27.539
Still, publishers and rights holders now have something they didn't have before, an appellate decision rejecting an AI-related fair use defense.
201
00:14:28.339 --> 00:14:34.583
And then also Penske Media and Chegg lost their antitrust cases against Google over AI overviews.
202
00:14:35.064 --> 00:14:37.065
Their basic complaint is easy to understand.
203
00:14:37.485 --> 00:14:38.866
Publishers create the reporting.
204
00:14:39.207 --> 00:14:42.669
Google can use material from that reporting to produce an AI answer.
205
00:14:43.129 --> 00:14:48.033
The answer may reduce the need for the user to visit the publisher that paid for the reporting in the first place.
206
00:14:48.673 --> 00:14:59.477
The court rejected the antitrust theories, concluding that the claims did not establish the legal arrangement required under antitrust law, while acknowledging the broader economic concern facing publishers.
207
00:14:59.697 --> 00:15:05.639
Penske is particularly relevant here because it owns creative industry publications including Variety, Rolling Stone, and Billboard.
208
00:15:05.740 --> 00:15:09.081
Put the two cases side by side and you get an interesting contrast.
209
00:15:09.821 --> 00:15:17.126
rights holders are attacking essentially the same economic problem through copyright, licensing, contract, competition law, and legislation.
210
00:15:17.567 --> 00:15:21.630
And this week, copyright gained some traction and antitrust didn't.
211
00:15:21.910 --> 00:15:26.513
It's far too early to conclude that copyright is where publishers will ultimately have the greatest leverage.
212
00:15:26.974 --> 00:15:30.276
The facts and legal theories are completely different, but I would keep an eye on it.
213
00:15:30.376 --> 00:15:36.781
Because we're leaving the era where every AI copyright argument consisted of somebody shouting fair use at somebody else.
214
00:15:37.221 --> 00:15:39.003
And now we're starting to get case law.
215
00:15:39.023 --> 00:15:41.064
So those conversations should get more nuanced.
216
00:15:41.585 --> 00:15:43.646
So those were the stories I felt were most important.
217
00:15:43.686 --> 00:15:51.492
But the question I keep asking myself is, what responsibility do we take when we hand these systems authority over our own lives?
218
00:15:52.113 --> 00:15:54.855
Because there's an answer nobody particularly enjoys hearing.
219
00:15:55.415 --> 00:15:56.997
You do, you silly bastard.
220
00:15:57.896 --> 00:15:58.797
That's the deal.
221
00:15:59.177 --> 00:16:05.461
It's like somebody continually asking how to lose weight while hoping the answer won't involve eating less and moving more.
222
00:16:06.081 --> 00:16:13.586
You can spend years looking for a more interesting answer, but eventually the boring one is still sitting there waiting for you to acknowledge it.
223
00:16:14.466 --> 00:16:21.333
These systems ask for permission, and convenience has an extraordinary track record of getting us to click yes.
224
00:16:21.574 --> 00:16:23.876
Now, that doesn't mean that companies get a free pass.
225
00:16:24.256 --> 00:16:28.981
We need responsible leaders building this stuff, and we need regulators and institutions paying attention.
226
00:16:29.782 --> 00:16:32.704
But none of that removes our own personal responsibility.
227
00:16:33.284 --> 00:16:44.570
If you give an agent permanent access to your home address and permission to act, and it uses that address to complete a task, some of the blame may need to be redirected to the person who clicked allow.
228
00:16:45.130 --> 00:16:46.151
That may sound harsh.
229
00:16:46.571 --> 00:16:53.695
I'm sorry, Rob, but I think we're gonna have to get comfortable with it because the agent era is gonna offer us an enormous bargain.
230
00:16:54.335 --> 00:16:56.937
Give me more access and I'll make your life easier.
231
00:16:57.757 --> 00:16:59.979
Most of us are going to be very tempted to take it.
232
00:17:00.299 --> 00:17:00.700
Fine.
233
00:17:01.420 --> 00:17:04.603
But every convenience is also a decision about authority.
234
00:17:04.864 --> 00:17:10.369
And we need to get much better at taking responsibility for our own data, our own permissions, and our own choices.
235
00:17:10.929 --> 00:17:12.711
The machines are getting way more capable.
236
00:17:13.111 --> 00:17:14.652
The interfaces are getting cuter.
237
00:17:15.093 --> 00:17:16.894
But the responsibility is still ours.
238
00:17:17.515 --> 00:17:19.096
So thank you so much for listening.
239
00:17:19.817 --> 00:17:20.858
I really appreciate it.
240
00:17:21.218 --> 00:17:26.563
We are about to have our 50th episode tomorrow of Technically Creative, which is so exciting.
241
00:17:26.803 --> 00:17:43.398
We also were just recently nominated for two Signal Awards, one as Best Overall Podcast in Business and Entrepreneurship, and one in Culture and Marketing for our Super Bowl episode with Chris Bellinger, with Chris Bellinger from PepsiCo and Mark Gross, the co-founder of High Dive.
242
00:17:45.069 --> 00:17:46.471
We feel so lucky.
243
00:17:46.711 --> 00:17:48.173
It's been a tremendous ride.
244
00:17:48.193 --> 00:17:50.656
We've been doing this for 18 months, but not continually.
245
00:17:50.756 --> 00:17:53.659
So 50 episodes over 18 months.
246
00:17:53.740 --> 00:17:55.542
Now these Week in Review episodes...
247
00:17:56.625 --> 00:17:57.186
We're growing.
248
00:17:57.226 --> 00:17:58.287
It's growing very fast.
249
00:17:58.507 --> 00:18:03.211
We are now, though, at the limit of my ability to make it grow while still running Kubrick Labs.
250
00:18:03.892 --> 00:18:08.096
And we do this for free because Kubrick Labs pays for the podcast and our partner Future of Life.
251
00:18:08.697 --> 00:18:10.498
But it's incredibly fun to do.
252
00:18:10.558 --> 00:18:14.122
But we are not trying to turn it into an economic engine for us.
253
00:18:14.742 --> 00:18:16.184
Kubrick Labs is growing just fine.
254
00:18:16.224 --> 00:18:18.606
Maybe because of some of the people who listen to this podcast.
255
00:18:19.106 --> 00:18:19.307
Yeah.
256
00:18:20.542 --> 00:18:24.726
And we will keep doing it as long as everybody keeps listening.
257
00:18:25.087 --> 00:18:26.248
So thank you for listening.
258
00:18:26.368 --> 00:18:28.370
Please go vote in the Signal Awards.
259
00:18:28.390 --> 00:18:29.812
I'll put the links in the show notes.
260
00:18:30.532 --> 00:18:31.293
Please go vote.
261
00:18:32.034 --> 00:18:32.995
It would mean a lot to us.
262
00:18:33.095 --> 00:18:36.999
In the business and entrepreneurship one, we are up against the Prof G podcast.
263
00:18:37.420 --> 00:18:39.221
And Scott Galloway is basically my hero.
264
00:18:39.342 --> 00:18:41.464
Yeah, so that's a total honor.
265
00:18:42.865 --> 00:18:43.706
It's a total honor.
266
00:18:45.087 --> 00:18:46.889
But yes, there is a listener's choice vote.
267
00:18:47.810 --> 00:18:48.391
Please listen.
268
00:18:48.831 --> 00:18:49.532
Please choose.
269
00:18:50.073 --> 00:18:51.334
And thank you for listening here.
270
00:18:51.514 --> 00:18:52.976
And we'll see you in the episode tomorrow.
271
00:18:52.996 --> 00:18:53.256
Bye-bye.
1
00:00:00.051 --> 00:00:06.715
Hi, and welcome to Technically Creative's Week in Review, brought to you by Kubrick Labs, the AI and IT partner for creative companies.
2
00:00:06.755 --> 00:00:13.159
Every Monday, I spend 20 minutes giving you an overview of the most important news from the increasingly interdependent worlds of tech and creativity.
3
00:00:13.399 --> 00:00:15.920
This is the Week in Review, October 5th, 2026.
4
00:00:16.280 --> 00:00:19.942
For the last few years, most of the AI conversation has been about capability.
5
00:00:20.343 --> 00:00:20.903
Can it write?
6
00:00:21.103 --> 00:00:21.984
Can it code?
7
00:00:22.064 --> 00:00:24.085
Can it hallucinate a Supreme Court case?
8
00:00:24.405 --> 00:00:31.747
That conversation isn't over, but I think we crossed into a more important one this week because the newest AI systems want to act on your behalf.
9
00:00:32.067 --> 00:00:33.828
Yeah, I'm talking about agents.
10
00:00:34.388 --> 00:00:38.689
OpenAI launched Dots, its new always-on agents inside GPT.
11
00:00:39.149 --> 00:00:45.431
Give a dot a goal and the permissions it needs, and it can keep working after you leave the conversation using its own cloud.
12
00:00:45.591 --> 00:00:48.031
Meta is pursuing much the same idea with Muse.
13
00:00:48.211 --> 00:00:54.013
Muse was downloaded 5 million times in 22 days, making it one of the fastest-growing iPhone apps ever.
14
00:00:54.313 --> 00:00:55.254
This is a big deal.
15
00:00:55.514 --> 00:00:57.616
Generative AI gave machines capability.
16
00:00:57.956 --> 00:01:00.158
Agentic AI gives machines authority.
17
00:01:00.418 --> 00:01:03.721
And as you'll see, authority is a much weirder problem to tackle.
18
00:01:04.082 --> 00:01:15.131
As a tech creator named Matt Robb discovered just this last week, in possibly the most perfect way, Robb had been using Meta's Muse to help with a sale through Facebook Marketplace.
19
00:01:15.752 --> 00:01:20.956
At some point, he'd given Muse his home address and selected an allow always permission.
20
00:01:21.677 --> 00:01:27.563
Muse subsequently shared that address with a buyer and arranged a pickup without Rob realizing what was happening.
21
00:01:27.663 --> 00:01:31.527
And then the buyer arrived at his house without Rob being any the wiser.
22
00:01:31.647 --> 00:01:34.029
I would say he's been doxed by his own petard.
23
00:01:34.770 --> 00:01:39.915
But Meta's position on this is that Muse acted within the permissions Rob had granted, and that's kind of fair.
24
00:01:40.215 --> 00:01:46.281
But did Rob know what he was doing, or did he just want to stop getting pestered with access requests every 60 seconds?
25
00:01:46.741 --> 00:01:47.301
Who knows?
26
00:01:47.581 --> 00:01:52.404
We're moving from a world where the main question was, did AI follow the rules?
27
00:01:52.484 --> 00:01:57.706
To a harder question, which is, did the human even understand the rules it was giving the AI?
28
00:01:58.206 --> 00:02:06.370
Anyone who's clicked accept all on the T's and C's of any software product should worry a little bit about our species readiness for this moment.
29
00:02:06.970 --> 00:02:12.674
Meta says Muse is designed to seek approval before sensitive actions, such as sending emails or making purchases.
30
00:02:13.074 --> 00:02:15.916
And the company provides an audit trail of what the agent's done.
31
00:02:16.316 --> 00:02:16.636
Okay.
32
00:02:17.256 --> 00:02:21.479
But following Rob's experience, Meta also said it was improving the permission experience.
33
00:02:21.679 --> 00:02:22.560
Good, I guess.
34
00:02:23.440 --> 00:02:26.702
And Meta is both culpable and sort of not culpable in this.
35
00:02:27.083 --> 00:02:32.306
And we're going to come back to Rob, because one of the ways in which I think everybody is culpable is this.
36
00:02:32.386 --> 00:02:38.150
The part that really amuses me is how adorable OpenAI and Meta have made these powerful agents.
37
00:02:38.811 --> 00:02:44.855
OpenAI's dots are sweet, little fuzzy, adorable emojis, and Meta's muse is designed to look like a Teletubby.
38
00:02:45.475 --> 00:02:54.339
It's an interesting aesthetic choice for software that knows where you live, reads your emails, makes purchases with your credit card, and communicates with strangers without you knowing.
39
00:02:54.719 --> 00:02:56.320
Of course we gave it a little fuzzy face.
40
00:02:56.820 --> 00:02:58.781
Japan has been doing versions of this for years.
41
00:02:58.901 --> 00:03:05.184
Railway offices and tax offices routinely wrap very serious institutions in absurdly cute little mascots.
42
00:03:05.844 --> 00:03:08.566
Authority looks like it belongs on a scratch-and-sniff sticker.
43
00:03:09.426 --> 00:03:14.690
Unfamiliar technology is intimidating, and good design makes complicated products feel approachable.
44
00:03:14.810 --> 00:03:15.331
I buy that.
45
00:03:16.071 --> 00:03:18.353
But the contrast is something I want to point out.
46
00:03:18.873 --> 00:03:22.876
The more agency the software gets, the cuter the AI companies need to make it.
47
00:03:22.896 --> 00:03:23.997
It's almost a rule.
48
00:03:24.518 --> 00:03:28.681
If you see the cutest little software mascot you've ever seen, be afraid.
49
00:03:29.041 --> 00:03:30.962
Be so, so afraid.
50
00:03:30.983 --> 00:03:31.083
Okay.
51
00:03:31.783 --> 00:03:37.187
It can access your files, your inbox, and your home address, but don't worry, it's shaped like a jelly bean.
52
00:03:37.387 --> 00:03:40.108
There's something important underneath that flippant joke.
53
00:03:40.549 --> 00:03:42.790
Approachable isn't the same thing as harmless.
54
00:03:43.170 --> 00:03:51.676
A friendly visual language can make an unfamiliar system seem easier to use, but it can also make granting that system enormous authority feel trivial.
55
00:03:52.216 --> 00:03:52.717
It's not.
56
00:03:53.457 --> 00:03:56.060
Allow always looks like a harmless little button.
57
00:03:56.361 --> 00:03:58.923
And check out how they're softening the font and shape on that too.
58
00:03:59.204 --> 00:04:01.246
My friends in design will know what I'm talking about.
59
00:04:01.566 --> 00:04:02.467
It's not harmless.
60
00:04:02.487 --> 00:04:03.629
You gotta pay attention.
61
00:04:03.969 --> 00:04:05.290
Yeah, the witch with the candy?
62
00:04:05.671 --> 00:04:08.013
She's gonna do horrible things to those little German kids.
63
00:04:08.634 --> 00:04:11.037
The rule you learned in kindergarten still applies.
64
00:04:11.517 --> 00:04:13.780
Be careful anytime anyone offers you candy.
65
00:04:14.448 --> 00:04:18.890
And Meta may not be the only company trying to help Rob from running into that mistake again.
66
00:04:19.110 --> 00:04:21.091
Apple wants to protect them at the compute level.
67
00:04:21.311 --> 00:04:33.496
The company says that it's changing how macOS handles full disk access, warning that some apps have been using it in ways that can expose files, mail, messages, and browsing history without users fully understanding the access they've granted.
68
00:04:33.536 --> 00:04:34.596
We're looking at you, Rob.
69
00:04:34.816 --> 00:04:39.378
Apple explicitly linked the change to increasingly capable and autonomous AI agents.
70
00:04:39.618 --> 00:04:41.039
I think they were talking about Muse.
71
00:04:41.279 --> 00:04:46.921
For a creative company, full disk access sounds technical until you replace it with what's sitting on that disk.
72
00:04:47.261 --> 00:04:57.684
Unreleased scripts, rough cuts, treatments, campaign strategy, client correspondence, budgets, talent agreements, design files, Dropbox folders, pitch decks, contracts.
73
00:04:58.044 --> 00:05:02.025
And especially if your company is mostly remote, that data is your company.
74
00:05:02.185 --> 00:05:04.426
And here's the basic tension at the bottom of all this.
75
00:05:04.866 --> 00:05:09.488
An AI agent becomes more useful precisely because of the access you give it.
76
00:05:10.089 --> 00:05:16.512
An agent that can't see your email, calendar, files, or business systems is a brilliant employee locked outside of the office.
77
00:05:16.712 --> 00:05:21.615
That's why I think the next few years of agent design may be less about raw intelligence than permission architecture.
78
00:05:22.075 --> 00:05:27.062
It goes back to the governance and approving layer we bore our clients with about every day.
79
00:05:27.242 --> 00:05:30.166
Yes, you need to think about stronger IT systems now.
80
00:05:30.306 --> 00:05:30.787
Shocker.
81
00:05:31.227 --> 00:05:36.774
And this agentic takeover of your life is about to hit advertising because of course it's about to hit advertising.
82
00:05:36.835 --> 00:05:37.415
What doesn't?
83
00:05:38.056 --> 00:05:42.661
Imagine telling your agent, find me the best hotel in New York for under $600 a night.
84
00:05:43.021 --> 00:05:44.783
Today, a hotel tries to persuade you.
85
00:05:44.843 --> 00:05:53.851
It buys media, optimizes search, pays influencers, and runs beautiful films of attractive people staring lustfully at each other across an infinity pool.
86
00:05:54.392 --> 00:06:01.795
But if the agent increasingly handles discovery, comparison, and recommendation, the hotel also has to become legible to the agent.
87
00:06:02.215 --> 00:06:03.376
Can the software find the room?
88
00:06:03.476 --> 00:06:05.096
Can it compare cancellation policies?
89
00:06:05.316 --> 00:06:06.457
Does it trust the reviews?
90
00:06:06.697 --> 00:06:14.160
And advertising starts to become mostly about persuading not you or people, but AI agents that this is the best deal for their user.
91
00:06:14.880 --> 00:06:18.341
Meanwhile, the industry is building agents on the other side of the transaction too.
92
00:06:18.521 --> 00:06:28.743
Media buyers, publishers, and ad tech companies are experimenting with autonomous systems that can buy inventory, manage campaigns, and optimize price or price gouge based on buying intent.
93
00:06:29.223 --> 00:06:40.926
Which means we're approaching the delightful possibility that your AI will someday go online to negotiate with Procter & Gamble's AI about which toothpaste you should buy while neither you nor Procter & Gamble is paying much attention.
94
00:06:41.570 --> 00:06:43.292
It's a great moment for humanity.
95
00:06:43.933 --> 00:06:46.816
The agent era is asking us something uncomfortable but important.
96
00:06:47.156 --> 00:06:49.458
How much of your judgment are you prepared to delegate?
97
00:06:49.859 --> 00:06:52.942
And that's a question about humans as much as it's a question about machines.
98
00:06:53.523 --> 00:06:55.204
So that's it for our deep dive this week.
99
00:06:55.565 --> 00:06:58.368
So now we get into some other news from the world of tech and creativity.
100
00:06:58.628 --> 00:07:04.554
So Rich Silverstein is stepping away from the day-to-day running of Goodby Silverstein and Partners after 43 years.
101
00:07:05.094 --> 00:07:15.785
The agency marked the occasion with a GPT Rich, an AI recreation built from decades of material and paired with a synthetic voice so people can still receive Rich-style creative feedback.
102
00:07:15.945 --> 00:07:18.568
It's a clever leaving idea, and the video was very sweet.
103
00:07:18.748 --> 00:07:23.934
It is a joke, and I do get that it's a joke, but the symbolism is potent, and I'm going to run with it a little bit.
104
00:07:24.115 --> 00:07:27.238
AI can mimic competency incredibly well, joke or not.
105
00:07:27.599 --> 00:07:33.225
An AI-rich Silverstein is probably pretty competent, and it zeroes in on exactly the thing I care about most right now.
106
00:07:33.866 --> 00:07:36.570
I don't care whether AI can imitate Rich Silverstein.
107
00:07:37.091 --> 00:07:39.795
I want to know what Rich Silverstein would do with AI.
108
00:07:40.295 --> 00:07:43.900
And that's why I'm sad he's leaving, even though I never worked with or for him.
109
00:07:44.161 --> 00:07:50.510
Because when an industry goes through a change this big, I want the people with the deepest craft and the most history leading from the front.
110
00:07:50.710 --> 00:07:54.952
Rich still sits at the desk that he bought when he started at Parsons School of Design.
111
00:07:55.412 --> 00:07:59.174
There are X-Acto knives and a bowl of colored pencils behind him.
112
00:07:59.674 --> 00:08:05.877
His creative life stretches back through Rolling Stone and physical typography, art direction, and decades of advertising.
113
00:08:06.217 --> 00:08:07.278
He's old school, man.
114
00:08:07.998 --> 00:08:12.079
But he also isn't standing on a hill yelling at the clouds and talking about the good old days.
115
00:08:12.519 --> 00:08:16.880
His recent 13 Days Cuba Missile Crisis musical incorporated AI-assisted work.
116
00:08:17.280 --> 00:08:17.580
Good.
117
00:08:17.960 --> 00:08:18.600
Keep going.
118
00:08:18.921 --> 00:08:21.141
Because we need people like Rich now more than ever.
119
00:08:21.621 --> 00:08:26.242
Rich's value isn't in the archives of things that he's made that AI is going to increasingly train on.
120
00:08:26.842 --> 00:08:30.883
It's 40 years of references, mistakes, taste, and instinct.
121
00:08:31.383 --> 00:08:37.585
The more history somebody carries, the more valuable it is when they show the rest of us what deserves to survive in the next era.
122
00:08:38.105 --> 00:08:40.186
So congratulations on GPT Rich.
123
00:08:40.226 --> 00:08:43.546
It really is such a warm and funny thing that's so current.
124
00:08:44.527 --> 00:08:49.608
But if there's a thing that I actually want to subscribe to, it's rich with GPT, not GPT Rich.
125
00:08:49.988 --> 00:08:54.889
And then speaking about creative leadership in AI, Ben Affleck continued last week's theme rather beautifully.
126
00:08:55.369 --> 00:09:01.631
At Bloomberg's Scream Time, Affleck said he used AI for lots of things while making his new Netflix film Animals.
127
00:09:02.071 --> 00:09:07.452
He declined to identify exactly where, arguing that pulling apart the mechanics of filmmaking can undermine the illusion.
128
00:09:08.272 --> 00:09:09.713
Nice save, Ben.
129
00:09:10.013 --> 00:09:11.514
That man could run for president.
130
00:09:11.934 --> 00:09:16.636
More importantly, though, he described the film as very human and framed the AI use as responsible.
131
00:09:17.156 --> 00:09:28.842
Interpositive, the post-production company he started that was bought by Netflix, was designed as a discrete post-production tool working on footage filmmakers had already created rather than generating a movie from scratch.
132
00:09:29.342 --> 00:09:39.567
The company builds its own controlled training data rather than scraping the internet, and Affleck has consistently emphasized licensed material, consent, human creative control, and preserving filmmakers' judgment.
133
00:09:40.127 --> 00:09:47.730
And that's what I want from people with experience, showing us, in practice, where they think the technology belongs and what value should govern its use.
134
00:09:48.250 --> 00:09:50.471
Because then the rest of us have something to learn from.
135
00:09:50.571 --> 00:09:54.113
We can adopt the framework, argue with it, improve it, or reject parts of it.
136
00:09:54.533 --> 00:09:56.715
but at least someone credible is leading from the front.
137
00:09:57.135 --> 00:10:02.379
That's way more useful than another panel asking whether Hollywood is for or against AI.
138
00:10:02.579 --> 00:10:07.842
And then in other news, everybody's freaking out about Skydance, but as usual, they're paying attention to the wrong thing.
139
00:10:08.383 --> 00:10:14.107
People are losing their minds over the fact that the combined Paramount-Warner Brothers company is going to sit under the Skydance name.
140
00:10:14.347 --> 00:10:16.588
And I don't really think that's the part to pay attention to.
141
00:10:17.088 --> 00:10:18.209
Sony bought Columbia.
142
00:10:18.249 --> 00:10:19.369
Columbia didn't disappear.
143
00:10:19.729 --> 00:10:21.730
Tristar didn't disappear, kinda.
144
00:10:22.190 --> 00:10:29.894
Those studio identities carried on perfectly happy under a parent company called Sony, which, if we're being honest, is a pretty shitty name for a movie studio.
145
00:10:30.234 --> 00:10:32.555
And I'll say something nobody else seems to be willing to say.
146
00:10:33.235 --> 00:10:34.816
Skydance is a badass name.
147
00:10:35.656 --> 00:10:39.897
It also has one of the best vanity cards in the business, that epic fireball in space.
148
00:10:39.957 --> 00:10:40.517
Come on.
149
00:10:40.997 --> 00:10:42.958
It's up there with the other two best in the business.
150
00:10:42.998 --> 00:10:44.898
And if you're wondering, that's Disney and Universal.
151
00:10:45.519 --> 00:10:51.640
So if your complaint about this merger is that the word Skydance is going to appear on some corporate paperwork, I think we can all survive.
152
00:10:52.220 --> 00:10:55.461
But the story is that Skydance was smaller than Paramount and took it over.
153
00:10:55.981 --> 00:10:58.942
Then Paramount is smaller than Warner Brothers, and it's taking that over.
154
00:10:59.502 --> 00:11:02.104
That's one hell of a corporate ladder to climb in two years.
155
00:11:02.665 --> 00:11:05.807
And the hires tell us something about what David Ellison wants to build.
156
00:11:05.907 --> 00:11:12.293
Ynon Kreitz, and I apologize if I screwed up that pronunciation, from Mattel is coming in as co-CEO.
157
00:11:12.993 --> 00:11:14.655
Mattel isn't his whole story, though.
158
00:11:14.675 --> 00:11:17.016
He was at Fox, Makers, and Endemol before that.
159
00:11:17.357 --> 00:11:18.938
So he has entertainment experience.
160
00:11:19.278 --> 00:11:28.126
Dana Goldberg and Josh Greenstein, the current heads of Paramount, are expected to take broader control of the film operation, with Pam Abdi and Mike DeLuca out at WBD.
161
00:11:28.766 --> 00:11:32.447
Diane Glasgow came over from meta Facebook to a senior product role.
162
00:11:32.788 --> 00:11:35.769
Taken together, this doesn't just look like bringing some tech people.
163
00:11:36.149 --> 00:11:42.331
It looks like an attempt to turn a collection of legacy media businesses into one more centralized, product-driven company.
164
00:11:42.831 --> 00:11:49.094
Shared infrastructure, shared technology, shared data, fewer overlapping power centers, and more centralized decision-making.
165
00:11:49.414 --> 00:11:57.697
And that could be good, but it's troubling that the systems might be governed by the people operating the smaller companies, not the bigger company.
166
00:11:59.148 --> 00:12:05.150
But listen, there are absolutely systems inside legacy Hollywood companies that deserve to be dragged into this century.
167
00:12:05.790 --> 00:12:10.572
But Warner Brothers is where the rubber meets the road, because creatively, quite a lot has been working.
168
00:12:11.072 --> 00:12:16.814
It backed distinctive filmmakers, original theatrical movies, and serious creative swings at a meaningful scale.
169
00:12:17.154 --> 00:12:24.897
And that culminated in an extraordinary 2025 run at the box office and during awards season, including the Best Picture Oscar for one battle after another.
170
00:12:25.477 --> 00:12:26.698
But Mike and Pam out?
171
00:12:26.718 --> 00:12:27.058
Pfft.
172
00:12:27.618 --> 00:12:29.359
Maybe that's just Hollywood being Hollywood.
173
00:12:29.419 --> 00:12:31.921
Twelve months can take you from genius to unemployed.
174
00:12:32.422 --> 00:12:35.284
But maybe Ellicent wants his own people and his own operating model.
175
00:12:35.704 --> 00:12:37.765
Probably both.
176
00:12:38.326 --> 00:12:40.207
But the hires make the bigger question clearer.
177
00:12:40.587 --> 00:12:43.750
Can you centralize the machinery without centralizing the taste?
178
00:12:44.190 --> 00:12:45.571
That's the thing to keep an eye on.
179
00:12:46.031 --> 00:12:47.212
Modernize the technology.
180
00:12:47.272 --> 00:12:48.493
Modernize the infrastructure.
181
00:12:48.593 --> 00:12:49.674
Fix the bureaucracy.
182
00:12:49.774 --> 00:12:50.534
Use better data.
183
00:12:50.575 --> 00:12:51.615
Make company faster.
184
00:12:51.835 --> 00:12:52.156
Sure.
185
00:12:52.716 --> 00:12:58.023
But don't turn Warner Brothers into just another application running on top of Skydance's operating systems.
186
00:12:58.644 --> 00:13:01.888
Because Warner's value isn't in the library, the lot, or the logo.
187
00:13:02.429 --> 00:13:08.397
It's the fact that generations of filmmakers have believed that Warner Brothers was a place that would let them make unique, original movies.
188
00:13:08.677 --> 00:13:14.462
The thing everyone says Hollywood needs more of, less IP, more original voices, that's what Warner's does.
189
00:13:15.302 --> 00:13:18.965
2025 was weapons and sinners in one battle after another.
190
00:13:19.505 --> 00:13:24.209
Original, artist-driven, no-IP-in-sight work that's special and makes money.
191
00:13:24.429 --> 00:13:28.833
Of course they do the IP-driven stuff as well, but it's worth noting that Warner's Walks the Walk.
192
00:13:29.513 --> 00:13:34.557
So if Ellison can preserve that while making the machinery underneath it dramatically better, fantastic.
193
00:13:35.158 --> 00:13:41.683
If he optimizes away the thing that makes Warner special, however, that can be an extremely modern way to screw up a wonderful company.
194
00:13:42.140 --> 00:13:45.741
And then in legal news, copyright wins one and antitrust loses one.
195
00:13:46.041 --> 00:13:48.282
We got two useful legal signals this week.
196
00:13:48.502 --> 00:13:58.584
The Third Circuit upheld Thomson Reuters' copyright victory against Ross Intelligence, making this the first U.S. federal appellate decision to address fair use in an AI-related context.
197
00:13:58.724 --> 00:14:05.946
Ross had used material derived from Westlaw headnotes to build a competing legal research product, and the court held that the use was not fair—
198
00:14:06.426 --> 00:14:15.171
Ross was not training a contemporary generative LLM, and the factual circumstances differ materially from the major cases against OpenAI, Anthropic, and Meta.
199
00:14:15.752 --> 00:14:19.554
So anyone posting fair use is dead should perhaps wait until after breakfast.
200
00:14:20.054 --> 00:14:27.539
Still, publishers and rights holders now have something they didn't have before, an appellate decision rejecting an AI-related fair use defense.
201
00:14:28.339 --> 00:14:34.583
And then also Penske Media and Chegg lost their antitrust cases against Google over AI overviews.
202
00:14:35.064 --> 00:14:37.065
Their basic complaint is easy to understand.
203
00:14:37.485 --> 00:14:38.866
Publishers create the reporting.
204
00:14:39.207 --> 00:14:42.669
Google can use material from that reporting to produce an AI answer.
205
00:14:43.129 --> 00:14:48.033
The answer may reduce the need for the user to visit the publisher that paid for the reporting in the first place.
206
00:14:48.673 --> 00:14:59.477
The court rejected the antitrust theories, concluding that the claims did not establish the legal arrangement required under antitrust law, while acknowledging the broader economic concern facing publishers.
207
00:14:59.697 --> 00:15:05.639
Penske is particularly relevant here because it owns creative industry publications including Variety, Rolling Stone, and Billboard.
208
00:15:05.740 --> 00:15:09.081
Put the two cases side by side and you get an interesting contrast.
209
00:15:09.821 --> 00:15:17.126
rights holders are attacking essentially the same economic problem through copyright, licensing, contract, competition law, and legislation.
210
00:15:17.567 --> 00:15:21.630
And this week, copyright gained some traction and antitrust didn't.
211
00:15:21.910 --> 00:15:26.513
It's far too early to conclude that copyright is where publishers will ultimately have the greatest leverage.
212
00:15:26.974 --> 00:15:30.276
The facts and legal theories are completely different, but I would keep an eye on it.
213
00:15:30.376 --> 00:15:36.781
Because we're leaving the era where every AI copyright argument consisted of somebody shouting fair use at somebody else.
214
00:15:37.221 --> 00:15:39.003
And now we're starting to get case law.
215
00:15:39.023 --> 00:15:41.064
So those conversations should get more nuanced.
216
00:15:41.585 --> 00:15:43.646
So those were the stories I felt were most important.
217
00:15:43.686 --> 00:15:51.492
But the question I keep asking myself is, what responsibility do we take when we hand these systems authority over our own lives?
218
00:15:52.113 --> 00:15:54.855
Because there's an answer nobody particularly enjoys hearing.
219
00:15:55.415 --> 00:15:56.997
You do, you silly bastard.
220
00:15:57.896 --> 00:15:58.797
That's the deal.
221
00:15:59.177 --> 00:16:05.461
It's like somebody continually asking how to lose weight while hoping the answer won't involve eating less and moving more.
222
00:16:06.081 --> 00:16:13.586
You can spend years looking for a more interesting answer, but eventually the boring one is still sitting there waiting for you to acknowledge it.
223
00:16:14.466 --> 00:16:21.333
These systems ask for permission, and convenience has an extraordinary track record of getting us to click yes.
224
00:16:21.574 --> 00:16:23.876
Now, that doesn't mean that companies get a free pass.
225
00:16:24.256 --> 00:16:28.981
We need responsible leaders building this stuff, and we need regulators and institutions paying attention.
226
00:16:29.782 --> 00:16:32.704
But none of that removes our own personal responsibility.
227
00:16:33.284 --> 00:16:44.570
If you give an agent permanent access to your home address and permission to act, and it uses that address to complete a task, some of the blame may need to be redirected to the person who clicked allow.
228
00:16:45.130 --> 00:16:46.151
That may sound harsh.
229
00:16:46.571 --> 00:16:53.695
I'm sorry, Rob, but I think we're gonna have to get comfortable with it because the agent era is gonna offer us an enormous bargain.
230
00:16:54.335 --> 00:16:56.937
Give me more access and I'll make your life easier.
231
00:16:57.757 --> 00:16:59.979
Most of us are going to be very tempted to take it.
232
00:17:00.299 --> 00:17:00.700
Fine.
233
00:17:01.420 --> 00:17:04.603
But every convenience is also a decision about authority.
234
00:17:04.864 --> 00:17:10.369
And we need to get much better at taking responsibility for our own data, our own permissions, and our own choices.
235
00:17:10.929 --> 00:17:12.711
The machines are getting way more capable.
236
00:17:13.111 --> 00:17:14.652
The interfaces are getting cuter.
237
00:17:15.093 --> 00:17:16.894
But the responsibility is still ours.
238
00:17:17.515 --> 00:17:19.096
So thank you so much for listening.
239
00:17:19.817 --> 00:17:20.858
I really appreciate it.
240
00:17:21.218 --> 00:17:26.563
We are about to have our 50th episode tomorrow of Technically Creative, which is so exciting.
241
00:17:26.803 --> 00:17:43.398
We also were just recently nominated for two Signal Awards, one as Best Overall Podcast in Business and Entrepreneurship, and one in Culture and Marketing for our Super Bowl episode with Chris Bellinger, with Chris Bellinger from PepsiCo and Mark Gross, the co-founder of High Dive.
242
00:17:45.069 --> 00:17:46.471
We feel so lucky.
243
00:17:46.711 --> 00:17:48.173
It's been a tremendous ride.
244
00:17:48.193 --> 00:17:50.656
We've been doing this for 18 months, but not continually.
245
00:17:50.756 --> 00:17:53.659
So 50 episodes over 18 months.
246
00:17:53.740 --> 00:17:55.542
Now these Week in Review episodes...
247
00:17:56.625 --> 00:17:57.186
We're growing.
248
00:17:57.226 --> 00:17:58.287
It's growing very fast.
249
00:17:58.507 --> 00:18:03.211
We are now, though, at the limit of my ability to make it grow while still running Kubrick Labs.
250
00:18:03.892 --> 00:18:08.096
And we do this for free because Kubrick Labs pays for the podcast and our partner Future of Life.
251
00:18:08.697 --> 00:18:10.498
But it's incredibly fun to do.
252
00:18:10.558 --> 00:18:14.122
But we are not trying to turn it into an economic engine for us.
253
00:18:14.742 --> 00:18:16.184
Kubrick Labs is growing just fine.
254
00:18:16.224 --> 00:18:18.606
Maybe because of some of the people who listen to this podcast.
255
00:18:19.106 --> 00:18:19.307
Yeah.
256
00:18:20.542 --> 00:18:24.726
And we will keep doing it as long as everybody keeps listening.
257
00:18:25.087 --> 00:18:26.248
So thank you for listening.
258
00:18:26.368 --> 00:18:28.370
Please go vote in the Signal Awards.
259
00:18:28.390 --> 00:18:29.812
I'll put the links in the show notes.
260
00:18:30.532 --> 00:18:31.293
Please go vote.
261
00:18:32.034 --> 00:18:32.995
It would mean a lot to us.
262
00:18:33.095 --> 00:18:36.999
In the business and entrepreneurship one, we are up against the Prof G podcast.
263
00:18:37.420 --> 00:18:39.221
And Scott Galloway is basically my hero.
264
00:18:39.342 --> 00:18:41.464
Yeah, so that's a total honor.
265
00:18:42.865 --> 00:18:43.706
It's a total honor.
266
00:18:45.087 --> 00:18:46.889
But yes, there is a listener's choice vote.
267
00:18:47.810 --> 00:18:48.391
Please listen.
268
00:18:48.831 --> 00:18:49.532
Please choose.
269
00:18:50.073 --> 00:18:51.334
And thank you for listening here.
270
00:18:51.514 --> 00:18:52.976
And we'll see you in the episode tomorrow.
271
00:18:52.996 --> 00:18:53.256
Bye-bye.
