Will AI be bigger than the internet?

1. Key Themes

The AI Platform Shift Is Real But Its Magnitude Remains Uncertain

The fundamental question isn't whether AI is transformative, but whether it's "just" as big as the internet and smartphones, or something more fundamental. Ben frames himself as a "centrist" believing this is as big as the internet or smartphones—"but only as big a deal as the internet or smartphones" [00:03:22]. The challenge is we don't know the physical limits of this technology, unlike previous platform shifts where we understood constraints. "We don't know the physical limits of this technology, and so we don't know how much better it can get" [00:00:37]. This uncertainty creates both opportunity and confusion, with competing claims from leaders like Sam Altman saying "we've got PhD level researchers right now" while Demis Hassabis responds "no, we don't. Shut up" [00:00:42].

The Usage-Capability Gap Reveals Fundamental Product Challenges

Despite ChatGPT having 800-900 million weekly active users, there's a striking disconnect: "If you're the kind of person who is using this for hours every day, ask yourself why five times more people look at it, get it, know what it is, have an account, know how to use it, and can't think of anything to do with it this week or next week" [00:00:03]. This gap reveals that while 10-15% of people in the developed world use AI daily, the majority who have tried it can't find consistent use cases. The challenge isn't just technical capability but product-market fit: "How do you map this against existing problems? But the other side of it is, how do you map this against new things that you couldn't have done before?" [00:27:33].

AI Will Follow Historical Software Patterns: From Raw Technology to Bundled Solutions

The evolution from general-purpose chatbots to specialized applications mirrors previous platform shifts. Ben notes that enterprises typically have "400 to 500 SaaS apps in the US, 400 to 500 SaaS applications. And they're all basically doing something you could do in Oracle or Excel or email" [00:24:20]. Similarly, AI will be unbundled into specific solutions: "Really? Isn't that what they're doing? They're unbundling chat GPT just as the enterprise software company, 10 years ago, is unbundling Oracle or Google or Excel" [00:25:03]. People don't want raw technology; they want solutions to specific problems, delivered with appropriate UI, workflow, and institutional knowledge built in.

2. Contrarian Perspectives

OpenAI's Consumer Dominance Is Surprisingly Fragile

Despite having 800-900 million weekly active users, OpenAI faces a unique vulnerability: "You've got these 800 or 900 million weakly active users, but that feels very fragile, because all you've really got is the power of the default and the brand. You don't have a network effect, you don't really have feature lock-in, you don't have a broader ecosystem, you also don't have your own infrastructure, you don't control your cost base" [00:41:47]. Unlike traditional platforms with network effects or ecosystem lock-in, OpenAI is "scrambling to get from this amazing technical breakthrough and these 800, 900 million wows to something that has like really sticky, defensible, sustainable business value" [00:43:01]. They must simultaneously build products, infrastructure, and defensibility across multiple directions while being billed monthly by competitors.

The Hyperscaler Investment Cycle May Be Rational Even If It Looks Like a Bubble

When questioned about potential overinvestment, Ben offers a counterintuitive defense: "You've got these kind of transformative capabilities that are already increasing the value of your existing products, if you're Google, or Meta, or Amazon. And you're going to be able to use them to build a bunch more stuff. And why would you want to let somebody else do that rather than you doing it as long as you're able to keep funding and selling what you're building" [00:18:18]. The calculus isn't about avoiding a bubble but about the asymmetry of risk: "The downside of not investing is bigger than the downside of overinvesting" [00:17:11]. This perspective suggests that what looks like irrational exuberance may actually be rational strategic positioning.

Apple's Lack of a Proprietary Chatbot Might Not Matter

Craig Federighi's argument that "we don't have our own chatbot fine. We also don't have YouTube or Uber" [00:47:45] represents a genuinely different strategic perspective. The question becomes whether AI fundamentally changes computing itself or is just another service layer. Ben explores the Microsoft parallel: "Microsoft loses a platform and sells an order of magnitude more PCs...an order of magnitude more Windows PCs as a result of this thing that Microsoft lost" [00:48:38]. If people still need a device to access AI—"it's probably going to have a nice big color screen and it's probably going to have like a one day battery life...Just that. Yeah. It kind of sounds like an iPhone" [00:49:44]—then Apple's position remains defensible even without owning the AI layer.

3. Companies Identified

Everlaw

Description: Cloud-based legal discovery software company

Why Mentioned: Used as an example of how AI capabilities get integrated into specialized vertical software rather than accessed directly through general-purpose tools. When machine learning enabled translation capabilities, Everlaw integrated it into their legal discovery platform rather than law firms accessing AWS APIs directly.

Quote: "Law firms want to buy a thing that solves this, want to buy legal discovery software management. They don't want to go out and write their own, buy a do API calls. I mean, very, very big law firms might, but typical law firm isn't going to do that" [00:31:32].

4. People Identified

Sam Altman

Description: CEO of OpenAI

Why Mentioned: Cited as representing one extreme in the AGI debate, claiming near-term achievement of PhD-level capabilities.

Quote: "You've got Sam Altman saying, we've got PhD level researchers right now. And Demis Hassabis says, no, we don't. Shut up" [00:00:42].

Demis Hassabis

Description: CEO of Google DeepMind

Why Mentioned: Represents the more conservative view on current AI capabilities, publicly contradicting claims of AGI achievement.

Quote: Referenced in the same exchange: "You've got Sam Altman saying, we've got PhD level researchers right now. And Demis Hassabis says, no, we don't. Shut up" [00:00:42].

Craig Federighi

Description: Apple's Senior Vice President of Software Engineering

Why Mentioned: Articulated Apple's strategic rationale for not building a proprietary chatbot.

Quote: "Craig Federighi made this point, which is like, we don't have our own chatbot fine. We also don't have YouTube or Uber" [00:47:45].

Andrej Karpathy

Description: AI researcher and former director of AI at Tesla

Why Mentioned: Referenced as offering timeline predictions for AGI capabilities.

Quote: "Karpathy goes into Dwarkesh's podcast and says, I feel like, you know, it's a decade out" [00:14:12].

5. Operating Insights

Validation Efficiency Determines AI's Practical Utility

The critical operational question isn't whether AI can perform a task, but whether validating its output is more efficient than doing the task manually. "By the marketing use case, it's a lot more efficient to get a machine to make you 200 pictures and then have a person look at them and pick 10 that are good than to have people make 10 good images" [00:26:23]. However, "if you're going to do data entry, if I'm going to ask a machine to copy 200 numbers out of 200 PDFs, and then I'm going to have to check all 200 of those numbers, I'm going to do this to myself" [00:27:22]. This validation calculus should drive deployment decisions: high-volume creative work with human review scales well; precision tasks requiring full verification don't.

GUIs Encode Institutional Knowledge That Raw Prompts Cannot Replace

An overlooked insight about user interfaces: "There are 600 buttons on the screen. There's seven buttons on the screen, because a bunch of people at that company have sat down and thought, what is it that the users should be asked here? What questions should we give them?" [00:33:00]. This represents "a lot of institutional knowledge, and a lot of learning, and a lot of testing, and a lot of really careful thought about how this should work" [00:33:20]. When you give users a raw prompt, "you're kind of got to shut your eyes, screw your eyes up, and think from first principles, how does this all of this work?" [00:33:35]. This explains why enterprise software companies building AI-powered vertical solutions will continue to have value—they're not just wrapping APIs, they're encoding domain expertise into constrained, optimized interfaces.

Distribution Beats Technical Excellence in Consumer AI

Despite models being technically equivalent on benchmarks, usage patterns diverge dramatically: "Cloud has basically no consumer usage, even though on the benchmark score it's the same. And then it's ChatGPT, and then halfway down the chart, it's Meta and Google" [00:41:21]. The explanation is pure distribution: "If the model for a casual consumer user certainly is a commodity, and there's no network effects, or winner takes all effects yet...How is it that you compete? Do you just compete on being the recognized brand?" [00:41:00]. For operators, this suggests that in commodity technical environments, brand recognition and default positioning matter more than incremental technical superiority—a lesson applicable beyond AI.

6. Overlooked Insights

The "AI" Label Is Ephemeral—Technology Becomes Invisible When It Works

A subtle but profound observation about terminology: "The term AI is a little bit like the term technology or automation. It's only kind of applies when something's new. Once something's been around in you for a while, it's not AI anymore" [00:04:34]. Ben's LinkedIn polls reveal this: "Is machine learning still AI? I don't know...in actual general usage, AI seems to mean new stuff" [00:04:46]. This parallels his example of 1950s elevator automation marketed as "electronic politeness"—today "you don't say, I'm using an electronic elevator. It's automatic. It's just a lift" [00:04:03]. The implication: as AI capabilities become reliable and ubiquitous, they'll fade into infrastructure. Companies building "AI companies" should prepare for the moment when their differentiator becomes invisible, requiring them to compete on actual business value rather than technological novelty.

LLMs Force Businesses to Discover What They Actually Sell

Perhaps the most profound strategic insight, briefly mentioned: "Do I just want the answer? Do I want that school or do I want to hear Stanley Tucci talking about cooking in Italy?" [00:55:23]. This question—do customers want the information artifact or the experience of expertise—will unbundle entire industries. Ben extends this: "Do I just want money or do I want to work with A16Z's operating groups? Like, what is it that I'm doing here?" [00:55:51]. The AI moment mirrors what the internet did to newspapers, which "didn't really say, well, we're a light manufacturing company and a local distribution and trucking company. And that was the bit that was the problem" [00:56:46]. Every company should be asking: what parts of our value proposition are really just friction that customers tolerate, versus genuine differentiated value? "What industries are protected by having that [boring, time-consuming processes] and they didn't realize that" [00:57:40]. The companies that survive will be those that honestly identify their irreplaceable value before competitors do.