Wrtn Expands AI Character Platform; Palo Alto Networks Acquires Applied Compute
Summary
Adina Specter and Jamie Ross discuss consumer sentiment regarding AI in film, citing an NRG survey where nearly two-thirds of Gen Z respondents rejected AI lead characters. They analyze Wrtn's expansion of its Crack platform into games and media at G-Star, noting its creator economy and revenue growth. The hosts also examine Palo Alto Networks' $4.5 billion acquisition of Applied Compute, highlighting the strategic combination of inference capabilities with security supervision.
Key takeaways
- 1 The rap and an entertainment news outlet reported reports that nearly two-thirds of Generation Z respondents in one survey found a lead character played by artificial intelligence unacceptable.
Jamie Ross · 1:21
Episode sections
- 0:05 – 7:04
Viewer Reactions to AI in Film
Adina Specter and Jamie Ross discuss moviegoer Nick Rampey's negative reaction to AI-generated animation in Eli Roth's film Ice Cream Man. They note that a survey by NRG found nearly two-thirds of Generation Z respondents found an AI lead character unacceptable, while 39% said some AI material would increase their interest.
- 7:04 – 12:54
Wrtn's Crack Platform Expansion
The hosts cover Wrtn's announcement at G-Star to expand its Crack character service into games, comics, and films. They discuss the platform's creator economy, the lack of clear content rules, and competition from other AI character services like Scatter Lab's Zeta and Krafton's KIF.
- 12:54 – 20:26
Palo Alto Networks Acquisition of Applied Compute
Adina Specter and Jamie Ross analyze Palo Alto Networks' agreement to acquire Applied Compute for $4.5 billion. They examine the strategic rationale of combining post-training and inference capabilities with security supervision, noting Applied Compute's $100 million annualized revenue and existing customers like Microsoft and Shopify.
Assets discussed
| Asset | Kind | Stance | Context |
|---|---|---|---|
| MicrosoftMSFT | equity | neutral | Applied compute does have customers of its own. It names software company Microsoft, online store platform Shopify, legal technology company Harvey, and delivery company DoorDash among them. Those customers aren't being presented as endorsing Palo Alto's deal. How much business does Applied Compute say it's doing? |
“Stance” is how the speakers talked about the asset, not a recommendation.
Claims worth verifying
Nearly two-thirds of Generation Z respondents in one survey found a lead character played by artificial intelligence unacceptable.
Adina Specter ·0:24
How to verify: Locate the NRG survey cited in the transcript; check NRG's official reports or press releases for the specific question and demographic breakdown. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
39% of respondents said some artificially generated material would make them more interested in a movie.
Jamie Ross ·1:21
How to verify: Check the NRG survey report or press release for the specific question regarding interest in movies with AI material. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Palo Alto Networks agreed to buy Applied Compute for four and a half billion dollars in stock.
Adina Specter ·0:45
How to verify: Review Palo Alto Networks' official press release or SEC filing (e.g., 8-K) announcing the acquisition agreement. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Primary sourceThe acquisition price is about a third of Palo Alto's projected annual revenue.
Adina Specter ·0:45
How to verify: Compare the $4.5 billion deal value to Palo Alto Networks' projected annual revenue in their latest earnings guidance or investor presentation. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Filmmaker Eli Roth acknowledged that Ice Cream Man used generative artificial intelligence in some scenes.
Jamie Ross ·2:20
How to verify: Search for Eli Roth's public statements or interviews regarding the production of Ice Cream Man and the use of AI. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
NRG found that respondents were more open to AI helping with color correction or special effects than replacing actors and writers.
Jamie Ross ·3:11
How to verify: Review NRG's survey analysis or report detailing the comparative responses to AI in technical roles versus creative roles. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Primary sourceWrtn announced plans at G-Star to expand Crack into games, comics, novels, animation, and films.
Unidentified speaker (UNKNOWN) ·7:17
How to verify: Check Wrtn's official announcements or press releases from the G-Star Gaming Convention regarding Crack's expansion. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Wrtn says more than 5,500 creators have received payments from the Crack service.
Adina Specter ·8:46
How to verify: Review Wrtn's corporate disclosures or press materials citing creator payment statistics. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Overseas sales account for more than half of Wrtn's revenue.
Adina Specter ·8:46
How to verify: Check Wrtn's financial reports or investor presentations for revenue breakdown by geography. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Applied Compute claims an annualized revenue pace of $100 million.
Jamie Ross ·15:57
How to verify: Review Applied Compute's public statements, pitch decks, or press releases citing their revenue metrics. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Do your own research
- Verify NRG survey methodology and sample size for the Gen Z AI acceptance data.
- Review Palo Alto Networks' S-4 or proxy statement for details on the Applied Compute acquisition terms and valuation rationale.
- Check Wrtn's latest earnings report or press releases for confirmation of the 5,500 creator payment figure and overseas revenue split.
- Investigate Applied Compute's customer list, specifically verifying Microsoft and Shopify as clients.
- Research competitor AI character services like Scatter Lab's Zeta and Krafton's KIF for market share data.
- Examine Wrtn's content policy updates following the G-Star announcements to assess regulatory risks.
Full transcriptexpand
0:05Adina Specter: Keyshore, welcome to your daily morning brief. I'm Adina Specter.
0:08Jamie Ross: And I'm Jamie Ross. Glad you're here.
0:10Adina Specter: Moviegoer Nick Rampey spotted animation made with artificial intelligence in filmmaker Eli Roth's horror film Ice Cream Man.
0:18Jamie Ross: He found it repetitive and imitative and it put him off. That's one viewer's reaction.
0:24Adina Specter: The rap and an entertainment news outlet reported reports that nearly two-thirds of Generation Z respondents in one survey found a lead character played by artificial intelligence unacceptable. Cybersecurity company Pelo Auto Alto Networks has agreed to buy Applied Compute, which trains and runs artificial intelligence models for four and a half billion dollars in stock, about a third of Pelo Alto's projected annual revenue. A dose of inspiration, people can hear that a film used artificial intelligence and picture very different things. Jamie, what did viewers in the news survey say about seeing it on screen?
1:21Jamie Ross: The Rap, an entertainment news outlet, reported that nearly two-thirds of Generation Z respondents found a lead character played by artificial intelligence unacceptable. In a separate question, 39% said some artificially generated material would make them more interested in a movie. Those questions
1:44Adina Specter: And we don't know how many people answered or when they were asked. What can one moviegoer tell us that those figures can't?
1:52Jamie Ross: Nick Rampey was watching Ice Cream Man when the film shifted into animation. He said the generated scene felt repetitive and imitative, and he found it off-putting. off putting. He was discussing
2:02Unidentified speaker (UNKNOWN): the film.
2:02Jamie Ross: describing his experience of that scene. We can't extend his reaction to the other people who saw the film.
2:09Adina Specter: I keep thinking about the moment he noticed it. He was there to watch a horror film, and his attention went to the animation. Do we know what the filmmakers used artificial intelligence to do?
2:20Jamie Ross: Filmmaker Eli Ruth Roth acknowledged that the film used generative artificial intelligence in some scenes. We don't have a full account of the decisions behind this one. What we do have is a viewer reacting to the finished images.
2:39Adina Specter: Which is whether the scene works for the person watching it. I care about who did the creative work, too, but if I decide what I think of a film before seeing it, I miss what the filmmakers actually put on.
2:51Jamie Ross: I care about both. NRG, the survey firm, found that respondents answered differently when asked about an artificial intelligence lead character and about some generated material appearing in a film. Kevin Tran, NRG's senior manager of Trends and Futures, said respondents appeared more open to help with color correction or special effects than to replacing actors and writers. That's his interpretation of their answers.
3:17Adina Specter: Let's make those jobs plain. A person adjusting the color can help a scene look the way the filmmaker wants. An actor gives a character a performance. A writer decides what that character says. I can understand someone feeling differently about those uses even before they've seen the result.
3:34Jamie Ross: The answer to the lead character question describes those respondents. It doesn't describe every young moviegoer, and it can't predict ticket sales. A filmmaker still has to decide what people will make and perform, then put the finished scene in front of an audience.
3:50Adina Specter: The other answer deserves the same care. Respondents were asked whether some generated material would increase their interest. Some leaves a lot unresolved. We don't know whether each person imagined a small visual. effect, a whole animated scene, or something else.
4:06Jamie Ross: Right. I wouldn't ask a filmmaker to treat either answer as permission for a particular scene. The filmmaker still has to choose what people make and perform, what the computer helps with, and what the audience sees.
4:22Adina Specter: You're putting more weight on the production decision than I am. I want to know who wrote and performed the character, but I also want to leave room for someone to watch the work and disagree with me about it. Where's your line?
4:34Jamie Ross: For me, replacing a person's performance changes who made the character we're watching. That's a consequential choice for the filmmaker and the performer. I can say that plainly while accepting that a viewer might care most about whether the scene holds their attention. The survey captures an objection to an artificial intelligence lead. Rampay described his response to a finished scene.
4:56Adina Specter: That's fair, and I don't think a studio can settle the concern by saying a human gave the computer instructions. I'd want to know which human decisions survived into the character's words and actions.
5:08Jamie Ross: Disney's Anzelans And Zelans give us a concrete comparison. Their They're Star Wars robot prototype. prototypes from Disney, the entertainment company. Artificial intelligence selects among dialogue and movements created by its theme park designers. The designers set the characters' available responses. We have no audience reaction that tells us people prefer those robots to an artificial intelligence film performance. performance.
5:33Adina Specter: I like how clearly you can describe the designer's part in that. It still leaves the work to be judged when someone encounters the characters. A studio could explain its process perfectly and then make a scene I don't enjoy.
5:47Jamie Ross: Absolutely. And it could make a compelling scene while leaving actors or writers with good reason to question their place in the production. Viewers and workers can be concerned about different parts of the same film. The figures we have don't tell us how to weigh those concerns for everyone.
6:04Adina Specter: Kishur, suppose a studio asks you to make a character consistent across several scenes. I'd want to hear who approves what the character says, what a performer contributes, and how the studio plans to tell people about the character. Those are questions you could bring into the room without claiming to know how the audience will react.
6:25Jamie Ross: Yes. Before anyone talks about making that character more convincing, the studio can decide which parts of the performance belong to people and which parts it's asking the model to handle. Then it can show the finished work. We should listen to viewers about what they actually watched, with the same care we use when we hear from the people who made it.
6:46Adina Specter: The generated animation in Ice Cream Man made Rampey care less about the scene.
6:51Jamie Ross: Next, we're moving to Wear RTN's crack and its plan to take creators' character worlds beyond chat. Client calls UARATN, a South Korean consumer artificial intelligence company, wants to take its creators' character worlds beyond chat. Adena, what did it announce?
7:17Unidentified speaker (UNKNOWN): At
7:17Adina Specter: the G-Star Gaming Convention in Pusan, URTN said it wants successful characters and worlds from its crack service to become games, comics, novels, animation, and films. Digital Today, a South Korean technology news outlet, reported the announcement. Chief Executive Lee Seyoung Seiyung also described plans to expand abroad and add video and interactive features, though he gave no launch dates for those editions.
7:43Jamie Ross: Before we get to any possible fit with your work, what does a person do on Crack?
7:48Adina Specter: A creator designs a character and the world around them. A user enters that world by talking with the character and the exchange shapes the story. Crack's own description also includes party chat, where up to five people can take part together. UARATN is trying to carry the characters and worlds people enjoy in chat into other kinds of entertainment.
8:09Jamie Ross: Give me the company behind it. Where did UARATN start?
8:13Adina Specter: URTN says it was founded in Seoul in April 2021. It began with tools for writing. Its URTN copywriting service launched the following year. The company later moved into Character Chat, which became Crack. It also has an in-house artificial intelligence organization and works on business and government projects. Today's announcement is about growing its consumer entertainment business in Korea and abroad.
8:40Jamie Ross: So a creator makes the starting world and users help make the story through conversation. Who gets paid?
8:46Adina Specter: RATN says creators receive payments from the service. The company says more than 5,500 creators have received them, though it hasn't said how many are active now or what a typical creator earns. IRN also says overseas sales account for more than half its revenue. Those figures tell us it has paying activity beyond Korea. Don't tell us whether a particular character succeeds when RN takes it into another format.
9:14Jamie Ross: What's Lee's argument for doing that? A popular chat character still has to work in a film or a game.
9:28Adina Specter: UITN plans to take its platform into more countries as well as distribute content made by Korean creators overseas. The company already operates entertainment services and in Japan and North America under local names.
9:56Jamie Ross: What does a crack creator need before putting more work into one of these worlds?
10:00Adina Specter: Kagito had a pointed answer at the convention. That's the public name of a crack creator who studies philosophy at university. Kagito asked URTN to say clearly in advance what kinds of expression and images its rules permit. A creator can spend time making a character and its story before finding out whether War Tn will allow the material. Kaguito asked for the rules before doing that work. And
10:24Jamie Ross: what did Ware Tien say back?
10:26Adina Specter: Chief Product Officer Li Dongje said the company would make its guidance clearer using legal review and consultation with government agencies. Li gave no date for the clearer rules. Kogido didn't say War Tien had removed their work. They're trying to decide what to make while they wait for clearer rules.
10:44Jamie Ross: If UAA wants a character to move from a chat into a comic and then a game, that uncertainty could follow the career of the
10:50Unidentified speaker (UNKNOWN): game.
10:50Jamie Ross: creator into each new project. I'd want to know which decisions belong to the creator before UAA spends money developing the character elsewhere.
10:59Adina Specter: And I'd want to know which rules the creator can actually use. A general promise to be careful with content doesn't tell Kagito whether a particular image is allowed. The planned expansion gives creators more places their work might appear. Uraya Yurayatian's Tien's answer says clearer guidance is. coming. Creators don't have it yet.
11:17Jamie Ross: There's competition for those creators and their audiences too. Scatter Lab, a South Korean artificial intelligence company, runs Zeta, another character chat service. We've also discussed KIF, the conversational music fan character from game company Krafton. Crafton. Crafton shaped an official character with an artist. Crack draws on worlds built by
11:38Unidentified speaker (UNKNOWN): the first time.
11:39Jamie Ross: its creators. That difference affects whose approval your ATN needs as a character grows.
11:45Adina Specter: It affects what people come to play too. On Crack, someone can enter a creator's world with friends and see where their conversation takes it. I can picture us disagreeing about what our characters should do next. The creator still needs to know how far that shared story is allowed to go.
12:02Jamie Ross: Here's the possible fit I'd examine, and it's a question. If UARATN and a creator wanted help, could a creator-approved persona stay consistent in chat and in an interactive story? You could evaluate whether it remembers the character's history, keeps relationships straight, and respects the creator's limits. We have no indication UATN is seeking a vendor.
12:26Adina Specter: And we don't know how CRACS models are trained. Before proposing any outside work, I'd ask how UATN's own team trains and evaluates its characters today. They already have an artificial intelligence organization and a working product. A useful conversation would have to begin with something the creator and URTN both say needs improving.
12:47Jamie Ross: If UARATN wants more from Crack's creator. creator-made worlds, creators need its content rules before they build.
12:54Adina Specter: Next, we're moving to Palo Alto Networks and its agreement to buy applied compute. Palo Alto Networks. I think its agreement to buy applied compute has a clear strategic idea behind it. Build artificial intelligence for businesses and help them control what it does. Jamie, what has Palo Alto agreed to buy?
13:20Jamie Ross: Palo Alto, a cybersecurity platform company, has agreed to acquire applied compute for four point five billion dollars in stock, roughly a third of Palo Alto's projected annual revenue. The price can be adjusted. Regulators still have to clear the deal, and Palo Alto expects it to close in its fiscal. third quarter of 2027.
13:41Adina Specter: I'm taking that pending deal seriously because Palo Alto says it assessed applied compute while considering becoming a customer. What does applied compute actually make?
13:51Jamie Ross: It refines artificial intelligence models after their initial training, which is called post-training. It also runs those models to produce answers, which is called inference. Applied Compute works with open weight models. Companies can obtain and adapt the model's learned settings, though the code and training data may have separate restrictions. Palo Alto has agreed to buy a company that helps shape an artificial intelligence model's behavior and then run that model.
14:20Adina Specter: Palo Alto says its Unit 42 security team assessed applied compute and found strong security foundations. Chief Executive Nikesh Aurora goes further. He says a company should train the artificial intelligence in an application alongside the artificial intelligence supervising it. What does the supervisor do?
14:39Jamie Ross: It checks or guides another model's work.
14:42Unidentified speaker (UNKNOWN): If
14:42Jamie Ross: a company trains a model to handle a particular job, Palo Alto argues the model watching it should understand that job too. I can see the logic. Palo Alto has given us its assessment and its reason for buying. It hasn't shown that customers get safer outcomes from combining the two.
14:58Adina Specter: Fair. I think having evaluated applied compute as a prospective customer gives Palo Alto a more concrete starting point than a purchase based on a presentation alone. It has looked at the technology it wants to put beside its security products. Why are you still uneasy?
15:15Jamie Ross: The purchase price buys a business Palo Alto still has to integrate. Applied Compute's people have to stay, its technology has to work with Palo Alto's products, and customers have to want the combined offering. Palo Alto says it can bring applied compute to its existing customers. That's a plan for future sales. The agreement itself doesn't make those sales happen.
15:37Adina Specter: Applied compute does have customers of its own. It names software company Microsoft, online store platform Shopify, legal technology company Harvey, and delivery company DoorDash among them. Those customers aren't being presented as endorsing Palo Alto's deal. How much business does Applied Compute say it's doing?
15:57Jamie Ross: Applied Compute claims an annualized revenue pace of $100 million. a small fraction of Palo Alto's projected annual revenue. Annualized means taking a current monthly pace and projecting it across a year. It isn't a completed year of sales, and it tells us nothing about profit. Applied Compute says it reached that pace less than 18 months after its founding.
16:20Adina Specter: That's early, and it's still a company with paying customers by its account. Palo Alto says it serves more than six million. 70,000 customers. If even some of those businesses want tailored models, I understand why Aurora wants the capability inside the company. I'm not saying they've signed up for it.
16:37Jamie Ross: They'll have other choices. Palo Alto names CrowdStrike, Fortinet, and Zscaler all security vendors among its competitors. Large cloud companies compete for enterprise artificial intelligence work too. Owning applied compute may help Palo Alto sell a combined service, but a customer can ask whether it wants the same supplier, training an application and checking its behavior.
17:00Adina Specter: I'd ask that question. I'd also ask whether two suppliers would understand the customer's application as well as a team building both pieces together. That's the case Palo Alto is making. You're right that it has to win the customer's trust after cloud.
17:33Jamie Ross: 14.52 and 14.57 billion dollars, roughly four times its latest quarter's sales. Those are Palo Alto's results and forecast. Applied Compute's future contribution remains unknown.
17:46Adina Specter: Applied Compute gives one example of what its work could mean to a customer. It says a model trained for Harvey improved answer and citation quality at less than half the cost of running a leading general-purpose general purpose model. For someone reviewing a legal answer, a usable citation matters. Does that prove Palo Alto's customers will see the same result?
18:08Jamie Ross: No. Applied Compute is describing its own customer project. Palo Alto will have to show customers what its combined products achieve. You know I like the idea of a model trained for a real job. I'm asking what happens when Palo Alto has to deliver it across many customers?
18:24Adina Specter: You're allowed to like the idea while arguing over the bill. Palo Alto has expanded by acquisition before. It completed its purchase of CyberArc, a company whose tools control who can access systems earlier this year. Does that make this purchase easier to judge?
18:41Unidentified speaker (UNKNOWN): It
18:41Jamie Ross: shows Palo Alto has used acquisitions to expand its platform. Applied Compute is a different technical layer. Its team refines and runs models, and Palo Alto wants to connect that work to security supervision. A past acquisition can explain management's approach. It can't tell us whether this team will stay or whether these products will fit together.
19:02Adina Specter: Applied Compute co founder and chief executive Yash Patil is expected to become Palo Alto's co chief artificial intelligence officer after closing. So even that leadership plan depends on the deal clearing. For your work building post-trained persona models, this shows a large buyer taking model behavior seriously. It says nothing about Palo Alto looking for an outside persona vendor.
19:25Jamie Ross: Exactly. Palo Alto has agreed to pay for applied computes people and technology because it believes training applications and the models supervising them belong together. Closing, integration and customer demand, still stand between that belief and a working business.
19:41Adina Specter: Palo Alto has made an expensive case for building and supervising artificial intelligence models for businesses.
19:57Jamie Ross: Look ahead. G G-Star, Star, the game convention in Busan, runs through Sunday, November 22nd. Goodbye,
20:05Adina Specter: Kishore.
20:05Jamie Ross: I'm Jamie Ross.
20:06Unidentified speaker (UNKNOWN): I'm
20:07Adina Specter: Adina Spector, and this is Kishore's nomenclature. Your next brief arrives tomorrow at 1 p.m. Eastern