Oct 3, 2025 · 49m · more-or-less
#119 OpenAI Sora vs. TikTok: Can “AI Entertainment” Fund the Compute Bill? · More or Less Podcast
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of the More or Less podcast, hosts Dave Morin, Sam Lessin, Jessica Lessin, and Brit Morin critically examine OpenAI's video app Sora, questioning whether consumer AI entertainment can monetize sufficiently to justify massive infrastructure capex. The panel also explores the broader AI financial bubble, the spread of synthetic 'AI slop', and the fundamental architectural limitations of large language models.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 99.8% of the talking time here. How this is scored →
The hosts as informed peer 5.7 Guest teaching 0.2 Guest disagreement 0.3 The hosts pushing back 3.8
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 2:10 SPEAKER_10 Blunt Rejection of Super Bowl Rumor
When Jessica Lessin speculates about Taylor Swift Super Bowl Easter eggs, SPEAKER_10 bluntly cuts across with 'No, that's done,' representing the sole direct disagreement from any guest speaker.
Hardest push from the hosts ▶ 44:18 Dave Morin Rejects Sam Lessin's Framing of AI ConsensusWhen Sam insists no credible technologist thinks expanding LLMs alone will solve fundamental problems, Dave directly pushes back, asserting a significant camp views LLMs as the foundational kernel of intelligence.
Biggest teaching moment ▶ 2:08 Clarifying NFL Halftime Show CompensationSPEAKER_10 and Brit Morin correct Jessica's speculation regarding a Super Bowl performance by explaining that talks broke down because the NFL does not pay performers.
The host holds their own ▶ 18:10 Sam Lessin Deconstructs Dead Internet EconomicsSam demonstrates deep analytical domain expertise by reframing Dead Internet Theory around the collapse of human economic incentives rather than mere technical generation of slop.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| More or Less Podcast Opening Sequence | 2 | 1 | 2 | 1 | The opening segment consists of light personal banter regarding Taylor Swift, concerts, and LP meetings. The only guest involvement is brief interjections, including SPEAKER_10 bluntly cutting off Super Bowl speculation. | |
| OpenAI Sora Launch: The MySpace Analogy and Creator Tools | 7 | 0 | 0 | 5 | The hosts delve into OpenAI's Sora release, analyzing social graphs, copyright risks, and modern mercantilism versus traditional venture ownership. The listed guests contribute only passive affirmative utterances like 'Yeah' and 'Yes.' | |
| Defining AI Slop, Dead Internet Theory, and Compute Economics | 8 | 0 | 0 | 6 | The panel debates AI slop, venture pitch degradation, and Dead Internet Theory, with Sam Lessin arguing high gloss cannot substitute for human economic incentives. No designated guest speaks during this segment. | |
| Hardware Necessity, Context Windows, and LLM Reasoning Limits | 7 | 0 | 0 | 4 | Jessica describes hardware OS limitations with voice assistants, prompting Dave Morin to discuss context windows and Richard Sutton's critique of LLMs versus world models. The nominal guests only provide minor one-word affirmations. | |
| Macro AI Bubbles: Infrastructure Capex and Solo Superintelligence | 8 | 0 | 0 | 6 | The hosts debate whether current AI capex resembles historical telecom bubbles, during which Dave Morin challenges Sam Lessin's assertion that scaling LLMs has lost credibility. No guest participates. | |
| Government Shutdown Fallout: Silicon Valley Taxes and Fleet Week | 2 | 0 | 0 | 1 | A quick wrap-up discussing the local San Francisco fallout of the federal government shutdown, such as tax deadlines and Fleet Week cancellations. There is no guest presence. |
Statements from this episode (14)
Brit Morin: Sam Altman only has 10,000 followers on OpenAI Sora
“Sam Altman only has 10,000 followers right now, which, so it feels like it's like very not viral from a friending perspective.”
Brit Morin: OpenAI Sora blows Meta's video tools out of the water
“I think compared to like Meta AI's video stuff they just launched recently, it, Kind of blows them out of the water, and so I think the AI social wars are definitely heating up.”
Sam Lessin: Sora is loose on copyright, strict on content moderation
“I'm just saying like that they've got, it's they've got a very interesting model going on right now where it's very loose on copyright and extremely tight on content moderation.”
Sam Lessin: Cartoon-like Sora video generations will be a short-lived fad
“The, what's cool about it is like the, just like the, what is it, the glibly, like the cartoons people are doing is it's clearly going to be a fad.”
Sam Lessin: AI video will devalue talking-head content until nobody cares
“I think it'll just devalue the content format to the point that no one cares of like the talking head, right?”
Sam Lessin: The internet is valuable as an economy, not technology
“The reason the internet is valuable is because it's an economy. It's not valuable because of technology. It's valuable as an economy. And that economy was set up in such a way to pump interesting information and unique commentary into it and like products and …”
Dave Morin: Founders are pitching AI-generated startups they don't even understand
“We have founders that show up and they will start showing me their idea and they'll clearly have vibe coded, whatever it is they're showing me. And they'll be trying to explain the idea. And it's clear they don't actually understand the idea and that the AI ac…”
Sam Lessin: Public prompts prevent creators from building moats on Sora
“Those prompts are all public, Brit. So no one can create a competitive advantage in the content of the prompts. Like if anything, the whole design is then to rip them.”
Dave Morin: OpenAI's obvious social play was group chat, not Sora
“I always assumed that one day I would wake up and I would open ChatGPT and it would ask me to upload my contacts and then I would be able to use ChatGPT with my friends, right? Like the four of us could use ChatGPT and have ChatGPT in the middle of the convers…”
Sam Lessin: Selling proprietary data to LLM providers for pennies is stupid
“All of the people who are, you know, some version of like, how do I sell data or give data or give context to chat GPT and get paid cents on the dollar are stupid, right? And like, it should be exactly the way around, which is you just use them as commodity ar…”
Jessica Lessin: Frontier AI models remain fundamentally next-token text predictors
“Today's like frontier models are like LLM pluses, right? They're LLMs and then they've got different capabilities kind of glommed onto them, but they're still fundamentally Text tradition kind of things”
Sam Lessin: Current LLM applications are not viable standalone businesses
“Like when you back it all out, you're like, none of these are actually good businesses. They're kind of like product and growth shims towards something, but no one can really tell you what that something exactly is, right? Or how it's going to work as a busine…”
Sam Lessin: The narrative of scaling LLMs with more compute is dead
“Like, I don't think anyone thinks that just expanding LLMs and the current technology was just like, there was an era of people were just like, just throw two orders of magnitude, more compute at it, and the game is over. And like, I just, I think that narrati…”
Sam Lessin: Multiple successful tech figures are building solo artificial superintelligence
“I have talked to three people in the last month that are all very, very smart, successful people. Who, when you're like, what are you doing? They're like, oh, I'm working on super intelligence. I'm like, with what team? I'm like, just by myself.”