Aug 22, 2025 · 1h 5m · more-or-less
#113 GPT-5 Disappoints: The End of the AI Gold Rush? · 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, co-hosts Dave Morin, Brit Morin, Sam Lessin, and Jessica Lessin critically examine the slowing capabilities and challenging economics of generative AI, the boundaries of Silicon Valley founder mythmaking, and Taylor Swift's masterclass in modern brand curation and operational scale.
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 99.4% of the talking time here. How this is scored →
The hosts as informed peer 4.6 Guest teaching 2.2 Guest disagreement 2.9 The hosts pushing back 3.3
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 26:26 Sam accuses top tech founders of lying
Sam rejects any benign interpretation of founder vision, explicitly asserting that Elon Musk and Sam Altman tell falsehoods they know are not true to manipulate capital.
Hardest push from the hosts ▶ 54:07 Jessica rejects social clip metrics for the podcast betJessica adamantly shuts down Brit's claim to have won the viewership bet, refusing the premise that aggregated short-form social video impressions equal complete episode views.
Biggest teaching moment ▶ 51:45 Sam explains why top creators avoid meme coinsSam provides a clear breakdown of tokenomics and creator brand incentives, demonstrating why real fan communities cannot be monetized via speculative crypto bonding curves.
The host holds their own ▶ 12:21 Dave exposes OpenAI's compute bottleneck and margin collapseDave demonstrates deep domain knowledge of model infrastructure, citing Q2 2024 compute configurations and explaining how frontier models hit hard hardware and margin limits.
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 Animated Title Sequence | 1 | 0 | 1 | 1 | The hosts open with light personal banter about their summer travels, Sam Lessin's home wine cellar setup, red light therapy beds, and Lyme disease treatment. The atmosphere is casual and collaborative with no substantive debate. | |
| Bret Taylor Interview: AI-Enabled SaaS Versus Foundation Models | 4 | 2 | 2 | 2 | Jessica Lessin recounts her interview with OpenAI board chair Bret Taylor, highlighting his distinction between high-margin AI-enabled SaaS and low-margin foundation models. The group quickly concurs that infrastructure is historically a tough business to monetize directly. | |
| Curating Dave Morin's Vintage Computer History Museum | 5 | 2 | 3 | 4 | After briefly touring Dave Morin's vintage computer museum, Jessica brings up whether AI IQ gains are plateauing. Dave directly challenges the tech industry's vague optimism, demanding a concrete definition for AGI. | |
| Compute Constraints and the Flawed Economics of GenAI | 6 | 3 | 4 | 5 | Dave argues that GPT-5's lack of a breakthrough marks a plateau driven by compute constraints and negative unit economics. Brit Morin pushes back, questioning whether enterprise productivity gains are merely early rather than stalled, while Sam and Jessica dissect real-world ROI. | |
| Silicon Valley Mythmaking: Visionary Storytelling Versus Deception | 5 | 3 | 6 | 6 | Sam forcefully argues that tech leaders like Elon Musk and Sam Altman knowingly lie to trick capital markets. Jessica and Brit push back strongly against conflating visionary storytelling and future-casting with outright fraud, prompting Jessica to issue a journalistic disclaimer. | |
| AI-Generated Pitch Decks and the Degradation of Information Signal | 6 | 3 | 3 | 3 | The conversation shifts to early-stage venture investing where pitch decks and rejection letters are increasingly contaminated by LLM-generated slop. The group agrees that automated content has dramatically degraded information signal and human communication. | |
| The Peak of Burning Man: Rave Culture and Polyamory Tales | 4 | 1 | 2 | 2 | Dave analyzes how Burning Man has shifted from diverse, community-driven art installations to a homogenized EDM rave culture. Brit and Dave share an anecdote about inadvertently camping with a polyamorous group. | |
| Dissecting 'The Hunting Wives' and Formulaic Streaming Television | 4 | 1 | 2 | 2 | The co-hosts critique the narrative breakdown of the series 'The Hunting Wives', comparing it to television series where primary showrunners hand off middle episodes to secondary writers or formulaic algorithms. | |
| Chinese Robotics Innovations and the Pitfalls of Creator Meme Coins | 7 | 5 | 3 | 3 | Examining rumors of an Emily Chang meme coin, Sam outlines the economic mechanics of pump.fun bonding curves and explains why legitimate creators cannot risk launching fan tokens due to severe reputational asymmetry. | |
| Settling the Viewership Bet and Sam's Riverside Exit | 4 | 2 | 5 | 6 | Brit claims victory in their Taylor Swift podcast viewership bet, citing third-party social media data. Jessica firmly pushes back, arguing that fragmentary clip impressions across social feeds do not satisfy the condition of full-episode viewership. | |
| Deciphering the Sourdough Easter Egg and Super Bowl Theory | 5 | 2 | 1 | 2 | The group breaks down Taylor Swift's appearance on 'New Heights', analyzing hidden clues like the sourdough references tied to the 49ers mascot for the upcoming Super Bowl, as well as the heavy handler infrastructure behind her unscripted image. |
Statements from this episode (24)
Jessica Lessin: Sierra and Harvey have roughly $5B private market valuations
“Both of them have about five billion dollar valuations.”
Sam Lessin: Frontier foundation AI model companies are terrible businesses
“These are terrible business.”
Dave Morin: Only three foundation AI models have any reasonable size
“There's three foundation models of any reasonable size.”
Sam Lessin: Tech infrastructure doesn't make money; even railroads relied on real estate
“You don't make money in infrastructure. Even the railroads made money through real estate.”
Dave Morin: Capability and intelligence gains in frontier AI models are plateauing
“The answer has to be yes.”
Jessica Lessin: Bret Taylor's AGI optimism is just OpenAI fiduciary duty
“This is the part of the conversation where he does his fiduciary duty as the board chair of open AI. That's what that sentence is, right?”
Jessica Lessin: AI leaders shift public narrative from AGI to narrow domains
“That's what these AI leaders have been saying for months and months and months. And now they're saying, well, it doesn't matter or they'll get better at specific things. I think this is a big deal.”
Dave Morin: OpenAI's GPT-5 is only a marginal improvement over predecessors
“GPT-V launches. It's not that much of an improvement. Everybody knows it. Nobody wants to talk about it, but anyone using the thing can Just feel it. Like it's not actually that much better.”
Dave Morin: OpenAI peaked in Q2 2024 and has declined since
“Whatever the compute configuration plus the software configuration that was going on in Q two of 20, 24 at open AI was the most powerful, most awesome version. And everything's been downhill from here.”
Dave Morin: OpenAI has no profit margin and is entirely underwater
“It's not currently working, right? Like the margin is, there's no margin. They're underwater. They have, you know, they're still funding the thing on large scale venture capital.”
Jessica Lessin: Average customer service interaction costs companies $20 to $30
“The average customer service costs cost the company between 20 and 30 dollars.”
Dave Morin: LLMs won't match internet impact without major new breakthroughs
“I think we have a lot more research to do to find a breakthrough of similar scale to this LLM one. It's not discounting how big of a breakthrough it is. It's just that the ecosystem has gone to like the max, I mean, to like LLM equals internet in its impact. R…”
Lessin: AI value will concentrate in vertical applications over general platforms
“So I do think you end up with like more vertical applications, leveraging this versus like platformy type stuff and open-ended type stuff, which again, we've also been talking about forever.”
Sam Lessin: Thin AI API wrappers are now viable venture investments
“There's actually things that Two years ago, I would have kicked and screamed about it and be like, that's just an API wrapper. Don't invest in that. I'm like, ah, actually we're at the point now we're like plugging this stuff into very verticalized application…”
Sam Lessin: Tesla and OpenAI only exist because of Elon Musk's marketing
“And so I'd say that like, I would say that just in the spectrum of like, Tesla only exists as Tesla, because Elon is a great marketer, right? Like, I would say that the same thing is true of OpenAI is like, it was a very explicit marketing heavy strategy.”
Jessica Lessin: Every good Silicon Valley founder sells a 10-act future narrative
“I have never met a founder in the history of Silicon Valley, a good one, who doesn't have a narrative that's 10 acts ahead of where they are.”
Sam Lessin: Elon Musk and Sam Altman deliberately lie to motivate followers
“I think that Elon and Sam Altman lie, right? I think they say things they know to not be true, to Get people to do things. Right. And like, go on like with them. I just believe that from what I've seen, like, I think they're both way too smart to believe what …”
Dave Morin: Early-stage startup pitch decks are now filled with LLM content
“Our pitches are actually filled with LLM now. And how do you figure out if an idea is an authentic idea or whether it was, you know, harvested out of the LLMs instead?”
Brit Morin: Using ChatGPT to evaluate follow-on capital allocation math
“There's a company in our portfolio that is doing really well. I was talking to GPT about like, how much follow on capital do you think this company should like, what is the potential outcome here? Like I was doing all of this math with GPT just to try to like,…”
Sam Lessin: I spend roughly $1,000 a month on OpenAI APIs
“I actually probably spend On the order of a thousand dollars a month on OpenAI APIs right now. I don't really calculate it, but something like that, right?”
Sam Lessin: Much of generative AI revenue is empty machine-to-machine noise
“I actually bet that like a lot of the use of, and this is why this is so scary from a revenue perspective, is like, with the machines talking to machines doing random things, there's just a lot of like, A lot of movement and no output right from a lot of what'…”
Sam Lessin: None of my personal creator currency investments have worked
“I have invested in a ton of them. They have, none of them have worked.”
Sam Lessin: Top influencers avoid creator tokens fearing they will lose fans' money
“No actual influencers or people who have brands want that because it's such an asymmetric risk. If they don't want to lose their fans money, they're terrified of losing their fans money because they like want to be loved”
Brit Morin: Taylor Swift's New Heights podcast drove 500 million views rapidly
“According to measurement company Mondo Metrics, Taylor Swift and New Heights reached half a billion views across all of the podcast episode across social media in under 24 hours. Three hundred and seventy nine million on Instagram, sixty one million on X, fift…”