Zynga Founder: Consumer Is Not Investible Right Now - Thats Why You Should Build It

Mark Pincus argues that while consumer tech recently seemed “not investable,” the AI era now offers a rare chance to reinvent services and create a “new internet treasure,” building on his insight that true social networking began with Napster’s peer-to-peer connections and was completed only when Facebook added the missing “container of trust.” He distills his playbook from building five companies into a book for a broad audience, emphasizing that great products require engaging management, boards, and strategy—not just product. On AI, he describes the current models as genuinely “smart enough” peers that can do “relatively magical things” when given context and a harness, as shown in his live therapy-transcript example. His “Proven/Better/New” framework advises copying proven mechanics, making “better” something 10/10 users prefer, and treating “new” as speculative spikes that mostly won’t catch. He stresses that consumer success hinges on solving distribution, often through built-in viral hooks, and that founders should trust “true signal” or “heat” rather than metrics. Finally, he reframes “founder mode” as giving every founder permission to lead authentically, stay deeply involved in details, and remain the star player—while building team culture that tolerates changing direction every Monday to keep the mission alive.

• ▶ 0:00 Mark Pincus argues the consumer space is currently "not investable," yet the AI era creates an unprecedented opportunity to reinvent services and build a "new internet treasure."
• ▶ 1:14 The book distills a playbook from building five companies: to build great products you must engage with the whole company—management, boards, investors, and long-term strategy—not just product.
• ▶ 2:36 The book is intentionally written for a broad audience, from a stay-at-home mom with an idea to experienced founders and peers, as an invitation into conversation.

• ▶ 3:44 The guest frames this as the third major computing era—after the web (Freeloader, 1995) and social/mobile (Tribe, Zynga)—now entering the AI era, which he finds "really exciting."

• ▶ 4:31 He traces the true beginning of social networking to Napster: seeing "4.5 million machines connected to you" and people looking "through the network at each other" made it the first decentralized "people web," years before Friendster or Facebook.

• ▶ 5:41 He says his failed company Tribe missed the critical component of trust—people needed a "good container of trust" to put themselves online, something Facebook got right by launching inside .edu; even insiders like him, Reed, and Peter Thiel underestimated how huge social networking would become.

• ▶ 6:47 On AI's inflection point, he describes earlier systems as "toys" until a later model "changed everything"; he now treats the AI agent as "a peer"—trustworthy for certain things, still hallucinating and "not quite right," but genuinely "smart enough."

• ▶ 6:57 AI can do “relatively magical things” when given the right harness and context, as seen in hands-on use like walking and talking with a voice assistant.
• ▶ 8:16 A practical consumer AI example: pasting a live therapy transcript into an AI with full context (email, texts, conversations) can yield genuinely useful insights—showing the tech is already here.
• ▶ 10:20 The Proven/Better/New framework: legally copy proven parts from successful products like Granola, make “Better” mean 10/10 users prefer it, and focus “New” on removing friction—such as always-on listening AI.
• ▶ 13:14 Investors can be 180 degrees off—consumer companies like support.com had no interest until after the consumer.com wave failed, and investors often don't think from first principles.
• ▶ 13:51 Consumer is hard because there's no proven distribution; the guest would only start a consumer company today if he could solve distribution, often via a built-in viral hook like getting people to email friends.
• ▶ 15:44 Use a "proven / better / new" framework: proven gives a base, better gives direction, and new are speculative spikes—assume most won't catch, and stay dispassionate about specific failed features while staying passionate about the underlying vision.
• ▶ 17:32 Founders facing doubt should resolve the core question of alignment and style: have they given themselves permission to pivot, or are they trapped in one product path?
• ▶ 18:07 Look for “true signal” or “heat” — when the product is right, feedback loops are emphatic yeses and “the fish are running,” so you don’t need metrics or to push people to work harder.
• ▶ 20:20 Management isn’t about MBAs or playbooks; the only point is getting people to do the right thing when you’re not in the room, and the first lesson is to be in the room as long as you’re the best player there.
• ▶ 21:23 "Founder mode" is recognized as the framework for ensuring people make the right decisions when the leader isn't present.
• ▶ 21:29 Leadership is defined as "presence, not absence"—being actively involved rather than delegating from afar.
• ▶ 21:42 The "Master and Commander" example illustrates leading by example: the captain rows alongside the crew, embodying the standard he expects.
• ▶ 21:54 Founder mode requires deep involvement in details—team trust grows when they see the CEO knows and cares as much as they do, like Elon Musk sleeping on the factory floor.
• ▶ 22:06 A consumer company CEO must love their product and know it better than anyone else; the speaker can't understand how leaders without that obsession can succeed.
• ▶ 22:17 "Founder mode" resonates personally as giving ourselves permission to lead authentically—it's about self-permission and staying true to your own style.
• ▶ 22:26 Founder mode is about giving yourself permission to be yourself, but VCs debate it as only for a few "deserving" founders like Bezos.
• ▶ 22:47 Founder mode is for every founder — you became a founder to bet on yourself, so don't abdicate that bet to your board or investors, especially when no one else believes in you.
• ▶ 23:13 The "expert witness" concept: you're closest to the answer but furthest from the decision, and founder mode was made for that position, though you must avoid staying an expert witness in your own company.
• ▶ 23:27 Founders make many compromises to hire executives or win investors, contorting themselves until they build a company they don't want to live in.
• ▶ 23:44 The founder may think "this isn't the company for me," but the correction is that the founder is the star player the company needs.
• ▶ 23:52 The founder is both the owner and the star player, and the company is worth more with them there—so stop building an environment you don't belong in.
• ▶ 24:04 Founder mode is a weekly, instinct-driven practice, not just about governance; founders must stay in touch with the instincts that built the company.
• ▶ 24:49 Founders must create enough context so that changing direction or mind every Monday is understood and accepted; founder mode has two parts: following your own instincts and building a team culture comfortable with that.
• ▶ 25:20 The mission is fixed like a chosen continent, but the founder can shift altitude daily; if the current tack isn't heading toward the mission, the founder must be able to say, "We've got to go that way."
• ▶ 25:35 The speaker highlights a "Monday learning" ritual where team members openly share outside observations, such as noticing a competing product that does what they're discussing better.
• ▶ 25:52 The core cultural question is whether the company defines itself as humble, curious, and intellectually honest.
• ▶ 25:57 The opposite culture is an "execution machine" where there is no room for anyone to raise concerns, questions, or outside information.
• ▶ 26:04 An audience member admits "I don't get it," prompting the speaker to clarify a nuanced idea.
• ▶ 26:08 Every company operates in different modes at different times—there is no one-size-fits-all approach.
• ▶ 26:10 In "new product mode," the team is trying something that has never happened before, requiring risk and experimentation.
• ▶ 26:16 In scaling/steady state mode, the product works well and the focus shifts to maintaining and growing what already exists.
• ▶ 26:20 The interviewer asks whether massive trillion-dollar consumer services and companies are still to be invented.
• ▶ 26:37 The speaker says "almost certainly" yes, because intelligence is now on tap, making new consumer services inevitable.
• ▶ 26:50 He cites that roughly 90% of enterprises that invested in AI haven't received any benefit yet, but argues this isn't surprising because it's still very early and adoption faces many barriers.
• ▶ 27:09 Token spending only matters if it actually changes how products are built and reaches users; burning tokens without user impact is just spending money with no effect.
• ▶ 27:28 2026 is the key checkpoint: if people still can't use tokens productively by then, it's a "skill issue," not a real technology limitation.
• ▶ 27:34 One speaker would only worry if unproductive token spending persists for three or four years, but the other questions whether the industry actually has that much time—leaving the timeline unresolved.
• ▶ 27:41 "Token maxing" is defined as spending aggressively on AI tokens to push the limits of what's possible.
• ▶ 27:54 Heavy spenders are common; the speaker personally spends ~$1M/year, and Peter Steinberger allegedly spends $1M+/month on tokens.
• ▶ 28:29 The "better version" is OpenClaw, built by Steinberger as a gift—spending a million dollars a year to create the world's best open-source platform.
• ▶ 28:53 AI adoption among the 90% is not homogeneous; the OpenClaw phenomenon reveals how usage is actually distributed.
• ▶ 28:55 Token spending is heavily concentrated—a million dollars a year on tokens may come from just one or two power users, not a thousand people.
• ▶ 29:08 The typical user in the 90% is on older, cost-sensitive models like GPT-3.5 on Copilot, not frontier systems, so headline adoption numbers overstate how advanced everyday AI usage truly is.
• ▶ 29:23 The speaker acknowledges that critics were right, despite initially facing backlash for his comments.
• ▶ 29:30 He admits that when he started coding again, he wrote half a million lines of Rails code — and that this was the wrong thing to do.
• ▶ 29:38 He uses his own experience as a candid self-correction, highlighting that massive code generation may no longer be the right instinct.
• ▶ 29:40 The old approach was writing human code that calls LLMs, treating the LLM as a peripheral tool.
• ▶ 29:46 The new approach is to have LLMs write the code you need on demand, cutting written code by 10–20x while making results more customizable.
• ▶ 30:01 Core principle: "Don't write code that calls LLMs. Write markdown that teaches LLMs to write code."
• ▶ 30:14 R&D is now about going out and "squandering tokens" because frontier models are so capable that token usage itself is a meaningful strategy.
• ▶ 30:31 Frontier model costs will drop by orders of magnitude in two years—roughly from 100,000 to 10,000 to 1,000—enabling a massive expansion of leverage.
• ▶ 30:48 Even if spending stays at a million dollars, you'd get the work of a million people instead of a thousand, pushing the limits of what pure software can do.
• ▶ 30:59 A pessimistic view is raised that society may already be at a point where a million people don't need to code, citing only 20 million coders worldwide.
• ▶ 31:10 The counterpoint argues the real number is likely much higher, depending on how "coder" is defined in an AI-assisted era.
• ▶ 31:23 The home screen is half empty and mostly filled with generic built-in apps, showing that even essential consumer surfaces are not saturated with amazing services.
• ▶ 31:40 Introduces "internet treasures" — indispensable services like Google or GPT — and notes very few exist so far; most truly essential consumer services have yet to be invented.
• ▶ 31:58 Though consumer may not be investable right now, AI and agents make the opportunity greater than ever, enabling both reinventions and entirely new services, making new consumer internet treasures highly likely.
• ▶ 32:52 The ideal consumer moment for AI is still about three orders of magnitude away due to high costs, with the founder agreeing that the consumer revolution is likely in 2029.
• ▶ 33:05 The founder draws an analogy to the early internet: heavy infrastructure investment (like dark fiber) and investor enthusiasm preceded a crash, but the real breakthrough came later.
• ▶ 33:22 Amazon’s numbers didn’t consistently climb until late 2002, when a big quarter proved the internet was finally happening—a full six years in, showing transformative consumer revolutions take much longer than hype suggests.
• ▶ 33:50 The speaker picks up an earlier thought before posing the core question.
• ▶ 33:51 They ask: "How do you think about timing for all this stuff?"
• ▶ 33:53 They reference a compelling example from the conversation, noting "that's a great example, right?" before beginning to frame their own point.
• ▶ 33:56 Being early to a major trend means enduring years of seeming "wrong" and lacking external validation before the world finally recognizes you were right.
• ▶ 34:06 By the time an idea breaks through, everyone feels late—so to capture an opportunity that takes off in 3-4 years, you must start building right now, with no shortcut around the early unvalidated work.
• ▶ 34:16 The unresolved challenge is staying power: how to persist through that long, unvalidated stretch when you can barely manage daily logistics, leaving endurance as the essential test for founders betting years ahead of the market.
• ▶ 34:27 The speaker describes himself as naturally being "18 months ahead of the early mass market majority," based on his own user behavior and taste rather than grand vision.
• ▶ 34:44 He observes that half his phone is empty, with only GPT and Claude holding his attention — and "definitely not games," signaling a void in consumer apps.
• ▶ 34:52 This gap implies a market opportunity: with a seasoned builder finding almost nothing worth using, the space is open for the next generation of consumer services.
• ▶ 34:54 Staying power is hard, especially when nothing currently excites you; seeking out new things to fall in love with is key to finding inspiration.
• ▶ 35:09 “The abyss” is the uncertain in-between period after one product passion fades, when you doubt you'll ever feel that passion again.
• ▶ 35:29 The abyss is not a dead end—it lets you expand your taste zones, and what you encounter there may hold unrecognized clues to your way out.
• ▶ 35:41 The core barrier to ambitious products is a cost problem: magical experiences exist but are too expensive to deliver broadly.
• ▶ 35:49 The industry needs a Jevons paradox point—compute so cheap and abundant it can be squandered, especially in games where every cent counts.
• ▶ 36:18 The path forward is building enough GPUs and achieving massive inference gains (e.g., Jeff Dean's 10,000x projection), making tokens cheap enough to enable wasteful, real-time AI everywhere.
• ▶ 37:09 The key consumer insight is a "time machine moment": realizing we didn't understand that AI would become intelligence on tap, like water—shaping how to build consumer products.
• ▶ 37:28 Design for free compute: ask "How would I use it if I could use unlimited amounts for everything?"—e.g., an always-listening app would cost ~$1,000/month, making it enterprise, not consumer.
• ▶ 37:49 It's just a matter of time before expensive capabilities become cheap for mass adoption; today's premium users are effectively "time travelers" previewing the future consumer experience.
• ▶ 38:01 Being early to AI is painful, but reframe it as inspiration: start building now and work backward from the future everyone will eventually have.
• ▶ 38:15 Use technology cost curves to predict breakthroughs—just as falling memory, display, and compute costs made the iPhone inevitable, dropping intelligence costs will enable new products.
• ▶ 38:37 Adopt the “business plan of free”: anything that can be free will be free—proven by Freeloader and Zynga—and think about what free will mean in the age of AI.
• ▶ 39:41 The future will bring free, unlimited AI inside consumer services, despite current pessimism.
• ▶ 39:50 The market mood is that "consumer is not investable," but that will change—"it's always darkest before dawn."
• ▶ 40:14 This shift will give rise to the next major consumer platforms, comparable to "the new meta" and "the new snap."
• ▶ 40:21 The host wraps up the conversation, referencing the earlier discussion about future timelines.
• ▶ 40:22 The host thanks Mark for the interview, calling it "amazing."
• ▶ 40:27 The host promotes Mark's new book, noting it is available "anywhere books are sold."

• ▶ 0:00 Opening: The AI Consumer Era and Mark Pincus’s Book (0:00 - 2:49) - Mark Pincus joins to discuss the AI consumer opportunity, “make something people love,” and the full-stack founder playbook behind his new book.
• ▶ 2:49 Eras of Computing and the Early Social Web (2:49 - 6:57) - First-principles lessons from computing’s waves, early social networks, Napster, Tribe, and the AI inflection point.
• ▶ 6:57 Practical AI and the Proven/Better/New Framework (6:57 - 13:16) - Hands-on AI use cases, why Siri fails, and applying Proven/Better/New to always-on AI under consumer and enterprise pressure.
• ▶ 13:16 Consumer Lessons and Validating New Ideas (13:16 - 17:11) - Post-consumer.com reality, investor blind spots, distribution, prosumer tools, voice AI, and staying dispassionate when new ideas fail.
• ▶ 17:11 Founder Doubt, True Signal, and the Fish Running (17:11 - 21:23) - Recognizing real product heat, knowing when the fish are running, and staying in the room while scaling.
• ▶ 21:23 Founder Mode and Staying True to Your Bet (21:23 - 40:37) - Founder mode as presence and detail, the Master and Commander rowing scene, and not contorting yourself or abdicating your bet.