Key Moments
Zynga Founder: Consumer Is Not Investible Right Now - Thats Why You Should Build It
Y CombinatorKey Moments
Consumer tech is currently uninvestable due to lack of distribution and high AI costs, but massive new 'internet treasures' are inevitable as AI compute becomes virtually free.
Key Insights
The current state of consumer tech investment is poor due to a lack of distribution channels and the high cost of AI tokens, which can reach $1.1 million per month for some companies.
The framework 'Proven Better New' suggests copying proven elements, improving on them, and then innovating with a novel feature, though the 'new' aspect is often the part that fails and needs testing.
The AI revolution is still early, with many enterprises (around 90%) yet to see tangible benefits from their AI investments, suggesting a current 'skill issue' with adoption.
Mark Pincus believes the next wave of consumer innovation, akin to the early web or social media, will be enabled by AI and agents, creating 'internet treasures' that become indispensable.
Founder mode, advocated by Brian Chesky, is essential for all founders, allowing them to trust their instincts and lead with presence rather than absence, even when their ideas are unpopular.
The cost of AI compute is expected to decrease by orders of magnitude, potentially making unlimited AI features within consumer services viable and leading to a new 'meta' for applications.
The current uninvestable state of consumer tech
Despite the immense potential unleashed by AI and agents, the consumer tech sector is currently facing significant challenges, making it arguably 'uninvestable' right now. Mark Pincus highlights the lack of clear distribution channels and the prohibitive cost of AI compute as major roadblocks. Companies are spending astronomical amounts on AI tokens, with one example cited at $1.1 million per month. This high cost, coupled with a lack of proven pathways to reach consumers, creates an environment where investors are hesitant. Pincus contrasts this with earlier eras of the internet and social media, where distribution was more straightforward, allowing for the emergence of 'internet treasures' – services that become indispensable to users' lives. The current landscape requires founders to be highly inventive in finding distribution, possibly through viral hooks or compelling 'new' features that entice early adoption, even if those specific features are likely to fail.
The 'Proven Better New' framework for product development
Pincus introduces the 'Proven Better New' framework as a systematic approach to product innovation. It begins by identifying a core instinct or unmet need. Then, one legally copies all 'proven' elements from successful existing products to build a solid foundation. Next, the focus shifts to making it 'better' – an improvement that 10 out of 10 users would recognize, such as increased speed, reduced friction, or lower cost. Finally, the 'new' element is introduced as the primary innovation to test. Pincus emphasizes that this 'new' aspect is often the riskiest part; it's the hypothesis that might be wrong and is most likely to fail. The key is to isolate and test this 'new' element rigorously, without becoming overly attached to it, as it's more about getting a trial and learning what resonates than being the definitive winning feature. Even with successful AI tools like GPT, the 'new' aspect often requires human ingenuity, suggesting that while AI can prove and perfect, humans are still crucial for true novelty.
Generational shifts in technology and the emergence of AI
Mark Pincus has witnessed and contributed to multiple foundational shifts in technology, from the early web with Freeloader in 1995, through the rise of social and mobile with Tribe and Zynga, to the current AI era. He likens the current moment to entering a new epoch, comparing the ubiquity of AI and agents to the transition from a world without widespread internet access to one where it's taken for granted. He recalls the nascent days of social networking, tracing its roots to Napster, and the profound realization of people connecting peer-to-peer. This historical perspective provides a unique lens to understand what might 'rhyme' in the AI age. For instance, the initial vision for social networks differed from their eventual reality, and Pincus himself missed the critical 'trust' component in his own social network venture, Tribe, which Facebook later mastered. This historical context underscores the unpredictability of technological evolution and the difficulty in foreseeing the full impact of new paradigms like AI.
The role of AI agents and the future of consumer services
The advent of advanced AI models like GPT and Claude has fundamentally changed how Pincus interacts with technology. He describes treating AI agents as peers, capable of sophisticated tasks when provided with the right context and constraints, even acknowledging their current limitations like hallucination. A compelling use case he and Garry discuss is an always-on AI assistant that listens to conversations, capable of providing real-time input and insights. This conceptual 'smart other person at the table' illustrates the potential for AI to become deeply integrated into daily life, moving beyond mere tools to become conversational partners. Pincus expresses surprise that such a product hasn't been built yet, despite its apparent 10x or 100x billion-dollar potential. The current challenge is feasibility: while the technology exists, making it consumer-grade and affordable remains a hurdle, with current implementations costing significant sums, similar to enterprise solutions.
Reimagining consumer products in the AI era
The core argument is that despite consumer tech's current uninvestable status, the opportunity for creating new 'internet treasures' has never been greater. Pincus believes in a future where AI and agents enable the reinvention of existing services or the creation of entirely new ones that people can't imagine living without. This vision is fueled by the anticipated decrease in AI compute costs, potentially making advanced AI features free and unlimited for consumers. He draws parallels to the early internet, where massive investment in infrastructure preceded widespread adoption and the rise of companies like eBay and Amazon. Similarly, he suggests that while current AI is expensive, it's poised for dramatic cost reductions, creating a future where applications can offer unlimited AI capabilities. This potential for free, ubiquitous AI is seen as the catalyst for the next wave of consumer innovation, offering a significant advantage for those who start building now and work backward from this future state.
Founder mode and maintaining vision amidst uncertainty
Pincus champions 'founder mode,' which he interprets as giving oneself permission to be authentic and follow intrinsic instincts, especially when facing doubt from others. He links this to Brian Chesky's concept of leadership as presence, requiring founders to be deeply involved and knowledgeable about their product. This doesn't mean being oblivious to team input; rather, it means creating an environment of intellectual honesty where the founder can pivot based on new information without demoralizing the team. He acknowledges that his own rapid shifts in focus can feel like 'third-grade soccer' but stresses that the goal is to remain aligned with the overarching mission while being flexible on tactics. Staying passionate requires finding things to fall in love with, which can be challenging during 'the abyss' – periods of uncertainty between projects. However, these moments are crucial for expanding one's taste and vision, ultimately leading to the next breakthrough. The key is to pursue the vision with passion but remain dispassionate about specific product variants that may not succeed.
The economic realities of AI compute and future potential
The conversation delves into the current high costs associated with AI, particularly 'token maxing,' where companies spend substantial amounts on AI models. While some individuals are spending upwards of $1.1 million per month, Pincus questions whether the output justifies the input. He notes that many enterprises investing in AI are not yet seeing benefits, attributing this to a 'skill issue' or a lack of strategic integration rather than a flaw in the AI itself. The projection is that AI compute costs will drop dramatically, possibly by 10,000x, making what is currently expensive financially inaccessible for broad consumer adoption a reality for everyone. This economic shift is seen as the key enabler for a new generation of consumer services, where unlimited AI capabilities can be offered for free. The current high costs are viewed as a temporary phase, and the path forward involves embracing this potential and building applications with the expectation of future abundance.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Books
●People Referenced
Mark Pincus's 'Proven Better New' Framework Evaluation
Data extracted from this episode
| Component | Evaluation | Timestamp (seconds) |
|---|---|---|
| Proven | A- (Faster and better than most; brilliant at proving) | 714 |
| Better | B- (Not always getting better; concept of 'better' inconsistent) | 736 |
| New | D (Human element still crucial for novelty) | 743 |
Common Questions
The consumer market is currently seen as less investable due to a lack of clear distribution channels and the current investor preference for enterprise solutions. Many investors are focusing on enterprise AI due to current trends and perceived fundability.
Topics
Mentioned in this video
Founder of Napster, who worked as an intern for Mark Pincus.
Founder of Zynga and author of 'Life at the Speed of Play'. Discussed his playbook for building companies and products.
Mentioned in the context of discussions about structurally setting up companies for the long game.
Founder of Facebook, who visited Pincus's office when starting out.
Co-founder of PayPal and Palantir Technologies, mentioned as someone who, like Pincus and Hastings, underestimated the scale of Facebook.
CEO of Airbnb, whose concept of 'Founder Mode' was discussed as a parallel to Pincus's ideas on leadership and presence.
Mentioned as an example of a founder exhibiting 'Founder Mode' by sleeping on the factory floor. Also mentioned regarding his predictions about AI and game development.
Founder of Amazon, cited as an example of a founder for whom 'Founder Mode' might be intended, though Pincus argues it's for all founders.
A Google AI researcher whose predictions about a 10,000x increase in inference capabilities were mentioned.
Co-founder of Netflix, mentioned in relation to the uncertain growth of early social networks.
Mentioned as someone spending significant amounts on AI tokens to create open-source projects like Open Claw.
Company founded by Mark Pincus, known for social and mobile games.
Mark Pincus's first company, launched in 1995, focused on the early web.
Considered the beginning of social networking, enabling peer-to-peer connections and file sharing.
Mentioned as a successful social network, with Mark Pincus regretting not getting the trust component right with his earlier company, Tribe.
Mark Pincus's failed social network company, where the key failure was the trust component.
Mentioned as a company whose numbers started to consistently climb in late 2002, signaling the revival of consumer internet companies after the dot-com crash.
Mentioned as a service that became indispensable, hinting at the potential for new AI-enabled 'internet treasures'.
Mentioned as a consumer internet company that took off early, preceding Amazon's sustained growth.
A company whose once-popular 'Flying Toasters' screensaver was directly competed against by Mark Pincus's free interactive screensaver, Freeloader.
An AI note-taking product that the speaker uses for real-time transcripts and analysis, but wishes had more evolved features like always-on listening.
An enterprise software company Mark Pincus built in 1999 that went public. It was overlooked during the consumer dot-com boom.
Mentioned as a company whose success is attributed to founder instincts which align with the discussed principles of founder mode.
Mentioned as an example of an 'internet treasure' service that has become indispensable.
The AI model that marked a significant shift, moving from usable tools to 'magical' agents that can be treated as peers.
An open-source AI voice plugin built on Gemini Live, mentioned as a competitor to Siri and potentially a future multi-billion dollar company.
A platform used as the base for an open-source voice plugin, indicating advancements in AI voice technology.
Apple's voice assistant, criticized for being terrible and performing poorly compared to newer AI solutions.
Mentioned as a significant AI model, used to analyze book content and apply the 'proven better new' framework.
A web application framework mentioned in the context of a past coding approach where writing many lines of code was common before LLMs.
Amazon's voice-controlled virtual assistant, the focus of a large development team with seemingly slow progress in LLM capabilities.
An AI model used alongside GPT to analyze book content and apply the 'proven better new' framework, though its 'better' aspect was rated lower.
An older version of GPT, mentioned as being used by companies worried about cost, which limits their AI adoption benefits.
A coding assistant that, when used with older models like GPT-3.5, reflects a cautious approach to AI adoption due to cost concerns.
More from Y Combinator
View all 601 summaries
43 minThe Age Of The 40-Year-Old Solo Founder Is Here
31 minGroww: If Your Customers Don't Love It or Hate It, You've Already Lost
77 min5 Papers That Show Where AI Research Is Heading Right Now
31 minHow Meesho Became India’s Biggest Shopping App
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free