Interview, Fireside Chat
Zynga Founder: Consumer Is Not Investible Right Now - Thats Why You Should Build It
- Mark Pincus argues that while consumer internet startups are currently difficult to fund ("not investable"), the opportunity to build "new internet treasures" has never been greater due to AI and agents.
- Pincus views the current moment as the third major computing era after the dawn of the web (1995) and social/mobile networks, suggesting we are transitioning to an age where intelligence is "on tap."
- He identifies the release of the "Opus 4.5" model as the specific turning point where AI shifted from being usable toys to acting as a "peer" capable of trust and complex tasks.
- Pincus describes his personal workflow of using an "always-on" AI agent that listens to conversations in real-time, noting that this specific use case represents a potential $10–$100 billion market that has not yet been built.
- He criticizes existing voice assistants like Siri and Alexa for failing to integrate modern Large Language Models (LLMs) effectively despite having massive engineering teams (e.g., 10,000 people on the Alexa team).
- Pincus introduces his "Proven Better New" framework for product development:
- Proven: Legally copy what is already successful (e.g., the note-taking mechanics of Granola) to avoid reinventing the wheel.
- Better: Optimize for metrics that existing users value (e.g., cost, speed, friction reduction).
- New: Isolate and test a single novel instinct (e.g., "always-on listening") while remaining dispassionate about its success.
- He warns founders that investors often incorrectly pivot successful consumer companies to enterprise, citing the 1999 dot-com bubble where enterprise software was favored while consumer sites were ignored, only for the consumer sector to dominate post-crash.
- Pincus suggests that current "prosumer" markets (e.g., 20 million developers) are the primary entry point for AI, using consumer-style tactics to acquire users before scaling.
- He defines the "Fisher Running" signal as a state of product-market fit where every feedback loop is a "yes," eliminating the need for extensive metrics or management intervention.
- Pincus outlines "Founder Mode" as the necessity for founders to remain the primary expert and decision-maker, especially when their vision is unpopular with boards or investors, citing Brian Chesky and Elon Musk as examples of leadership by presence.
- He estimates that a "Consumer Revolution" driven by AI will not materialize until 2029, as current inference costs remain too high for mass market adoption despite rapid improvements.
- Pincus notes that 90% of enterprises investing in AI currently see no benefit, attributing this to "skill issues" and the misuse of tokens rather than a failure of the technology itself.
- He advocates for a "squandering" strategy in R&D, where founders spend heavily on tokens now to build the infrastructure and workflows needed for when compute costs drop by orders of magnitude.
- Pincus predicts a 10,000x increase in AI inference capability (citing Jeff Dean), which will eventually allow services like video games to be generated "pixel by pixel by AI" at negligible cost.
- He recalls the 1995 launch of his first company, Freeloader, which used the "free" model for an interactive screensaver to compete with paid software, noting that "free" will likely redefine value in the AI era.
- Pincus advises founders to use "backward design" thinking: imagine a world where AI intelligence is free and unlimited today, then build the primitives that would exist in that future.
- He identifies the current "abyss" period between products as a necessary phase for expanding personal "taste zones" to discover the next generational opportunity.
- Pincus predicts that the gap between "what is possible now" and "what is affordable for consumers" will close within 3–4 years, making the current time the optimal moment to start building.