Zynga Founder Mark Pincus: Why All New Fails + How to Copy to Millions

A Note from James:

Mark Pincus is one of the true OGs of the internet. You probably know him as the founder of Zynga, the company behind FarmVille, Zynga Poker, and Words With Friends. Zynga was eventually acquired by Take-Two in a transaction valued at approximately $12.7 billion. Before Zynga, Mark started Tribe, one of the first social networks—before MySpace and Facebook.

He has spent more than 25 years building, failing, and studying what gets millions of people to click, play, share, and come back. His new book, Life at the Speed of Play, inspired me to start coming up with new business ideas while we were still recording.

What I really love is how Mark teaches people to copy like a master without looking like a copycat. He has a framework called “Proven–Better–New.” Start with something that has already been proven. Make it obviously better. Then isolate the new idea you want to test. It’s one of the best systems I’ve heard for creating products people actually want.

We talk about the early days of Facebook and MySpace, the failure of Tribe, the gaming industry, consumer psychology, AI coding, and how agents could eventually network and work for us while we’re doing something else.

I loved talking with Mark. I was still thinking about this conversation afterward—and I’m literally building businesses based on what I learned. His new book is called Life at the Speed of Play. Listen to this episode, and then read the book.

Episode Description:

Most founders begin with an idea and then spend months—or years—trying to prove that people want it. Mark Pincus thinks that process is backward.

At Zynga, Mark’s teams built “failure machines”: simple systems that allowed them to test hundreds of concepts before writing the code. They put unfinished ideas in front of real users, watched what people clicked, and refused to build anything until the demand was obvious. The objective wasn’t to avoid failure. It was to make failure fast, cheap, and useful.

Mark explains the framework behind that process: Proven–Better–New. First, study an existing success down to every screen, click, and design decision. Then identify one improvement that current users would immediately recognize as better. Only after that should a team add the unproven idea—the part most likely to fail.

James and Mark also examine the problems facing today’s consumer entrepreneurs. AI has made software easier to build, but distribution has become harder. People aren’t searching for new apps, established platforms restrict organic growth, and algorithmic reach isn’t the same as users actively sharing something with friends.

Mark uses the failure of his early social network, Tribe, to explain why virality is not enough. Tribe grew quickly but lacked retention and trust. He ignored the communities users loved because they didn’t match the business model he had already chosen. That painful mistake became the foundation for much of his later product philosophy.

The conversation ends with Mark’s current experiments: personal AI agents modeled after members of his family, a proposed work network built specifically for agents, an enterprise AI company called Hivemind, and the difficult decision to end a four-year passion project without abandoning the instinct behind it.

This is a practical conversation about testing ideas, separating instinct from ego, learning from the past, and killing the wrong product before it consumes the right opportunity.

What You’ll Learn:

• How to build a failure machine: Test headlines, offers, videos, and fake doors before investing in a finished product.
• How to apply Proven–Better–New: Begin with a proven behavior, make one unmistakable improvement, and isolate the risky innovation.
• Why distribution is now harder than development: AI can generate a prototype quickly, but it cannot guarantee attention, trust, or adoption.
• Why Tribe failed despite rapid growth: Virality without retention, safety, and alignment with user behavior does not create a lasting network.
• How to copy without becoming a copycat: Study successful products at the pixel level, preserve what works, and innovate only where it matters.
• When to abandon an idea: Preserve the underlying instinct, but stop funding the particular expression of it when the evidence turns against you.
• How AI agents may change networking: Agents could eventually search for opportunities, exchange work, build reputations, and bring useful leads back to their users.

Timestamped Chapters:

• [02:00] Finding the “OMFG” Moment
• [02:58] A Note from James
• [05:00] Build a Failure Machine Before Building a Product
• [06:25] Testing Demand With Fake Doors and Broken Links
• [08:08] Writing Copy That People Actually Notice
• [10:52] Test More Ideas in a Week Than the Industry Tests in a Year
• [11:53] Why Neglected Products Become Innovation Labs
• [13:26] How Mobile Apps Slowed Product Experimentation
• [15:09] Can AI Bring Rapid Testing Back?
• [17:08] Why Consumer Technology Feels Uninvestable
• [18:38] The 90/10 Rule for Investable Platforms
• [20:08] Why Nobody Downloads New Apps Anymore
• [21:20] Franchises, “Spicy New,” and Healthy Platforms
• [23:21] The Internet’s Lost Cocktail Party
• [27:58] Why Tribe Failed While Facebook Won
• [30:26] Virality Without Trust or Retention
• [31:31] Ignoring What Tribe’s Users Actually Wanted
• [33:22] Facebook, Raya, and Designing for Trust
• [35:03] Social Networks as Lead-Generation Engines
• [37:12] Facebook, Instagram, and the App Nobody Knew It Wanted
• [37:51] Net Promoter Scores and the Feeling of Quitting a Drug
• [40:25] Algorithmic Virality vs. People Sharing With Friends
• [42:00] Building Products That Help People Create
• [43:47] What Entrepreneurs Should Build With AI
• [44:54] The Proven–Better–New Framework
• [47:12] What “Obviously Better” Actually Means
• [48:25] Why “All New Fails”
• [50:23] Zynga Poker and the Power of Removing One Click
• [52:00] What AI Does Well—and Where Humans Still Matter
• [54:25] Picasso, Slack, and Copying the Past
• [55:11] Adding Fun to Boring Enterprise Products
• [57:39] The Moral Arbitrage of Killing Your Ego
• [57:58] How to Copy Without Looking Like a Copy
• [59:10] Why Old Internet Mechanics Keep Returning
• [01:00:16] Anonymous Social Apps With an AI Twist
• [01:01:17] Don’t Invent a New Business—Reinvent a Big One
• [01:02:00] Test 20 Variants Before Building One
• [01:02:58] Mark’s Frustrating Experiments With AI Coding
• [01:05:29] Creating a Personal Team of AI Agents
• [01:07:57] Killing a Four-Year Passion Project
• [01:09:29] The “Social Membrane” of the Agentic Internet
• [01:09:57] Building a Work Network for AI Agents
• [01:12:16] Hivemind and the Human Side of Enterprise AI
• [01:13:52] Missing Twitch—and Knowing Your Zone
• [01:15:06] Why the Gaming Industry Still Isn’t Social Enough
• [01:16:30] Chess Ratings, Competition, and Mark’s Daughter
• [01:19:19] Writing Life at the Speed of Play
• [01:21:18] Don’t Chase Every New Technology Race
• [01:22:05] Final Thoughts

Additional Resources:

Mark Pincus and the Book

• Life at the Speed of Play — official website
• Life at the Speed of Play — HarperCollins — published June 23, 2026.
• Mark Pincus on X — the account Mark recommends for updates on his agent-network experiments.
• Mark Pincus on LinkedIn
• Mark’s interview about open-sourcing Stem Studio

Zynga, Games, and Product Examples

• Zynga’s company history — covers its launch as a Facebook poker project and the development of FarmVille, CityVille, and Words With Friends.
• Words With Friends
• FarmVille
• Take-Two and Zynga acquisition announcement — the transaction carried an enterprise value of approximately $12.7 billion.
• Tribe.net history — the early social network Mark analyzes as a major product failure.
• Raya — the private community Mark discusses as an example of building trust through curation.
• Grow a Garden on Roblox

See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.