Sep 29, 2023 · 46m · more-or-less
#16: Why Your AI Startup is Probably Already Dead · More or Less Podcast
Jessica Lessin · 13m spoken Sam Lessin · 12m spoken Scott Belsky · 8m spoken Dave Morin · 4m spoken Brit Morin · 3m spoken
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 Jessica Lessin, Sam Lessin, Brit Morin, and Dave Morin join Adobe executive Scott Belsky to analyze why generative AI overwhelmingly advantages tech incumbents over venture-backed startups. The group delves into foundation model commoditization, the commercial viability and architectural limits of AI agents, and the FTC's landmark antitrust lawsuit against Amazon.
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 78.3% of the talking time here. How this is scored →
The hosts as informed peer 6.3 Guest teaching 3.6 Guest disagreement 1.9 The hosts pushing back 3.9
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 24:58 Scott insists both agent models are valid
Scott firmly reframes the debate between his enterprise-side suggestion pyramid and Sam's personal digital twin, asserting that both exist concurrently and interact.
Hardest push from the hosts ▶ 31:00 Jessica calls out contradictory AI optimismJessica bluntly halts the speculative discussion to point out the contradiction between utopian agent futures and the reality that current LLMs require manual double-running just to produce acceptable results.
Biggest teaching moment ▶ 22:21 Scott breaks down the agent hierarchy pyramidScott systematically educates the hosts on what constitutes an AI agent, escalating from basic responsive help desks to proactive autocompletes and prescriptive strategic advisors.
The host holds their own ▶ 37:44 Jessica dismantles antitrust misconceptionsJessica displays deep domain expertise on antitrust law, correcting her co-hosts by articulating why consumer pricing benefits make the FTC's legal hurdles against Amazon far steeper than seller complaints suggest.
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 | 2 | The hosts welcome Scott Belsky as the fifth wheel and engage in playful roasting about corporate titles and fashion. Scott participates in the banter without tension. | |
| Weekly Tech Agenda: Meta Connect, Jony Ive, and Big Tech AI | 6 | 1 | 0 | 1 | Jessica lays out the tech news agenda, citing reporting on the OpenAI and Jony Ive hardware collaboration as well as Meta Connect announcements. Scott mostly listens as the hosts set up the topics. | |
| The Incumbent Advantage: Data, Distribution, and the 'D-Squared' Trap | 7 | 5 | 3 | 4 | Sam argues AI advantages belong strictly to incumbents with data and distribution. Scott introduces the D-squared concept while offering a counterpoint that small businesses also gain major leverage, prompting a technical exchange on proprietary datasets. | |
| Foundation Model Cloud Wars: Commoditization and Big Tech Equity Deals | 7 | 4 | 2 | 3 | Jessica questions model differentiation, prompting Scott to draw an analogy to cloud rivalries avoiding AWS. The hosts dive into enterprise equity plays and multi-cloud dynamics with specific reporting details. | |
| Defining AI Agents: Autocomplete Pyramids vs. Personal Digital Twins | 7 | 6 | 3 | 5 | After Dave admits confusion over the buzzword agent, Scott delineates a 3-tier capability pyramid. Sam counters with an alternative model based on pre-web digital twins, prompting Scott to reconcile enterprise versus consumer agents. | |
| The Agent Future: Scams, Agent Inception, and the Death of the Open Web | 7 | 4 | 2 | 3 | Scott suggests personal agents will safeguard against AI scams and warns of multi-layer agent inception across OS platforms. Sam synthesizes this into an argument about the collapse of the open web into private walled gardens. | |
| AI Reality Check: Inference Costs, Smaller Models, and OpenAI's Valuation | 8 | 4 | 3 | 7 | Jessica forcefully challenges the panel's agent optimism by contrasting it with current LLM failure rates and workarounds. Scott reframes the discussion around smaller localized models, which Jessica supports with reporting on Microsoft's in-house efforts. | |
| The Amazon Antitrust Case and Agent-Driven Commerce | 8 | 3 | 1 | 8 | Dave and Brit argue the FTC case against Amazon is clear-cut, but Jessica challenges their legal reasoning by distinguishing seller friction from legal consumer harm. Scott avoids legal debate and instead reframes the commerce shift around interface layers. | |
| AI in Government: Public Sector Workflows and the DMV | 5 | 4 | 1 | 2 | Scott highlights government bureaucracy and DMV operations as massive unaddressed TAMs for AI workflow overhaul. The hosts agree while humorously noting the execution hurdles faced by public agencies. |
Statements from this episode (21)
Dave Morin Says He Is No Longer Invited to Facebook's F8
“I started that conference and I don't even get invited to it anymore.”
Jessica Lessin: Sam Altman and Jony Ive are discussing an AI device
“Myself and my amazing colleagues at the Information Broker story recently that Sam Altman, the CEO of OpenAI, and Johnny Ive, the former Apple designer and designer of the iPhone, Have been deep in conversations about building a hardware device together that k…”
Sam Lessin: Character AI will be displaced by platform incumbents
“Character AI, sayonara. Like, it's not like, you know, it's, these are cool ideas, but the reality is you have the data, you have the distribution. AI is an easy tack on for platforms. This is not disruptive. It's just extending what they can already do.”
Belsky: AI favors incumbents by embedding superpowers directly into existing workflows
“If you can just surface some of these superpowers in the workflows that customers have, so that tomorrow when they log in, suddenly they have it, then they don't have to, Necessarily go to an entirely new place and abs, you know, go through the friction of doi…”
Sam Lessin: VC-funded startups attacking incumbents with AI will not survive
“Anyone whose business ends in dot AI or who's taking loads of venture funding, right, to build, you know, a, some of the next generation money losing go for scale startup attacking an incumbent with AI as a differentiator. Those are the losers. Those are not g…”
Sam Lessin: Core LLM capabilities will be commoditized down to zero
“But I just think this stuff is all gonna, the core stuff will be all very commodity, whether it's decentralized or not. And commodity things end up getting competed mostly to zero.”
Morin Predicts Frontier AI Models Will Inevitably Train on Enterprise Data
“Like, there's no way that doesn't happen over time. It's too appetizing. The data asset is too, Like, the centralization of that much data, of that much value, like, you cannot keep your hands off of it.”
Belsky: Adobe builds proprietary creative models but partners on LLMs
“We've kind of come out publicly and said, hey, for all the creative models, we're going to build our own, because we feel like we have the data sets and the experience on that, but on the LLM side, we're partnering with a number of these folks, and we are just…”
Belsky Outlines an AI Agent Pyramid From Help Desk to Proactive Actions
“My view is that an agent is literally not only in the pyramid of things, right? It can start by answering your questions and it's a glorified help desk, but then going up the pyramid a little bit, it starts to proactively do things for you. Like it's sort of, …”
Brit Morin Says Bumble Is Deploying AI Agents for Dating Negotiation
“Bumble is already starting to do this, by the way. So props to Whitney. Bumble. Whitney over there is Is starting to.”
Belsky predicts AI agents will protect users from AI-driven harms
“Agents are going to save us from all the concerns we have about AI, because a lot of the concerns around AI.”
Lessin Predicts the Public Web Is Dead Due to AI Ingestion
“What's really happening, I think, with a lot of this AI stuff, is the public web is dead. People are not going to post shit publicly, right? Because why would you give away all the value, get nothing in return?”
Lessin: Retreat behind walled gardens will necessitate personal AI agents for discovery
“Everyone goes behind wall gardens, right? So what that means is that rather than having a public web you can look at and navigate, instead you're gonna have all these little pockets of wall gardens. And the only way that you're gonna get to know what's in them…”
Belsky: Many AI Applications Will Run Locally on Smaller Models
“I happen to believe by the way that a lot of AI powered things are going to be able to be run locally. And that, you know, and also with less powerful models that, you know, for good enough outcomes.”
Lessin: Microsoft Is Developing Cheaper In-House AI to Avoid OpenAI Costs
“So Microsoft, who's ten billion dollars into open AI, gave it exclusive access to their IP, has decided in part because it's too expensive to run, they're going to build a cheaper functional enough version to have to power some of these applications.”
Morin: Amazon antitrust lawsuit is the most textbook big tech case
“Of all of the antitrust cases to me, this is the one that is the most textbook, right? Like here you've got Amazon, I think in the most basic way you know, interfering with the market you know, releasing products to undercut the price of other products in thei…”
Lessin: It is impossible to argue consumers are harmed by Amazon
“There's no way you can argue that consumers are worse off because of Amazon. You just can't.”
Lessin: DOJ antitrust case against Google will not be earth-shattering
“The DOJ Google case, it's hard to know how it's going. I think the DOJ will land some blows, but it's not going to be. Earth shattering.”
Lessin Predicts AI Shopping Agents Will Obsolete Amazon in 15 Years
“When you look in 10 or 15 years in the future, I think a lot of this stuff will become less relevant. Because little mini AI Sam, Agent Sam looking for things and talking to all of their agents doesn't need to go to Amazon.com. It can be way more creative in t…”