#558. The Personal Revolution in the AI Era: Garry Tan on Open-Source AI, Entrepreneurial Faith, and Trauma-Driven Creation

📝 Episode Summary

This episode is a clone of Rick Rubin’s long-form podcast *Tetragrammaton* **Garry Tan**.

Our guest, Garry Tan, is CEO of Y Combinator, as well as an engineer, entrepreneur, and investor. From a chaotic childhood where he sought order through computers and code, he has grown into the leader of one of Silicon Valley’s most influential startup institutions. In this conversation, Rick Rubin—approaching from a creator’s perspective—engages Garry in a deep dialogue spanning tech history, entrepreneurship, AI, open source, social media, psychological trauma, and life choices.

You’ll hear Garry explain YC’s core credo: “Make something people want”; why he believes AI is ushering in the next personal computing revolution; how personal AI, open-source agents, and vibe coding will transform ordinary people’s creative capacity; why entrepreneurs need sincerity, unique perception, and extraordinary agency; and Garry’s rare, candid reflections on his own startup failures, childhood trauma, family relationships, and inner fire.

This is not just a show about AI and entrepreneurship—it’s a profound conversation about how individuals find their origins, confront epochal change, and transmute pain into creativity.

👨‍💻 Guest

Garry Tan, CEO of Y Combinator, engineer, entrepreneur, and investor. Founder of Posterous and co-founder of Initialized Capital, with investments in numerous prominent tech companies. Today, he leads YC back to its roots in early-stage founder programs while actively advancing open-source AI, personal agents, Little Tech, and the broader startup ecosystem.

⏱️ Timestamps

00:00 Intro & Episode Overview

From Chaotic Childhood to the Code Cave

01:29 Technical启蒙: Approaching Science and Engineering Through Reading Far Beyond Comprehension

01:43 Childhood Escape: Order Found Through Computers, Code, and Math

02:16 Games and Narrative: From *Monkey Island* to *Red Dead Redemption* as Spiritual Sanctuaries

YC’s Credo: Make Something People Want

02:55 What Is Y Combinator: Twelve Questions, a One-Minute Video, and a Ten-Minute Interview

03:43 Garry’s YC Interview Memory: Generating a Blog via Email on iPhone

04:30 Paul Graham: Hacker, Painter, Lisp Pioneer, and Early Web Application Visionary

06:16 How YC Differs from Traditional VC: Money and Metrics Are Merely Outcomes

07:19 How YC Operates: $500K Investment—and Why Community Matters More Than Capital

The Personal Computing Revolution in the AI Era

08:12 Money Matters Less; Those Who Can Truly Build Matter More

09:13 Code Generation Boom: Claude Code, OpenClaw, and the Open-Source AI Movement

10:13 Why Personal AI Must Be Owned and Controlled by the Individual

10:59 G Brain: A Personal Knowledge Memory System That Reads Emails, Calendars, Contacts, and Notes

11:57 AI’s Apple II Moment: From Institutional AI to Everyone Owning Their Own Agent

12:25 From Personal Computers to the AI Homebrew Computer Club: The Next Revolution Is Underway

Open Source, Bias, and Model Competition

13:18 Everyone Running Their Own AI Agent: A Future Still Believed in Only by a Few

14:06 AI Bias and Control: Why Garry Sides with the “Pirates”

14:51 DeepSeek, Qwen, and Open-Source Models: Good News, Bad News, and Copyright Controversies

16:20 How Technology “Moves the Cheese”: The Death of Media and Advertising, and Public Anger Toward Tech

17:50 Creator Economy: Patreon, Patronage Models, and Artists’ Income Challenges

18:43 After AI Takes Jobs, Will Humans Create More Art, Music, and Experiences?

Internet Evolution and the Builder Spirit

19:55 Internet Revolution: High School Days of Laying Cables, Running Linux, and Building Websites

20:33 14-Year-Old Web Designer: Yellow Pages, Cold Calls, and First Job

22:27 The Importance of Mentors: Learning Design and Professional Habits from Adult Colleagues

22:48 Early Web Design: Making Sites as Visually Polished as Magazines

23:51 Apple’s First E-Commerce Store: Things We Take for Granted Today Were Once Radically New

24:51 Social Media Revolution and the Confusion After Web 1.0 Collapsed

26:17 Why the Best Early-Stage VCs Must Be Builders Themselves

How Startup Ideas Evolve

26:39 From Initial Pitch to Final Product: The Shifting Paths of DoorDash and Coinbase

27:50 Capabilities of Great Founders: Returning to First Principles and Continuously Reframing Problems

27:50 YC’s 13-Week Program: Removing Founders from Daily Life

29:14 Immersive Entrepreneurship Environment: Creating Space, Establishing Rhythm, Entering Real Competition

29:39 Peer Motivation and Competition: Why a Little Competition Helps Founders Accelerate

30:55 Demo Day Pressure: Real Stakes and Funding Outcomes After 13 Weeks

31:02 Why YC Accelerates Teams: Changing Environment, Peers, and Mindset

Garry’s Return to YC and Lessons from Sam Altman

32:25 Paul Graham Asks Garry to Return and Run YC

33:05 Sam Altman Takes Over YC: From Controversy to Seeing the AGI Future

34:38 Garry’s Reflection: Why He Failed to Grasp Sam’s AGI Insight Back Then

35:30 “Yes, And”: Staying Open with Founders Instead of Rushing to Dismiss

36:52 Re-Becoming a Builder: How Open-Source Projects Let Garry Keep Creating Hands-On

YC’s Open Knowledge and Founder Judgment

37:22 “Do Things That Don’t Scale”: Why YC Freely Shares Its Methods

38:03 Homogenized Pitches: Social Networks, Tutor Marketplaces, and Common Pitfalls

39:10 Fringe Ideas: Why a Space-Based Nuclear Fusion Company Deserves a Bet

40:30 What Really Matters in a Ten-Minute Interview: Ideas Are Functions of People

40:59 People > Ideas: Founders Must Genuinely Care About Something

41:39 San Francisco’s Uniqueness: Where Eccentrics No Longer Feel Like Outsiders

42:10 Origins of Silicon Valley: Semiconductors, Stanford, DARPA, and Government-Funded Basic Research

43:14 Where Is the Next Silicon Valley? Startup Probability Differences Between NYC and SF

43:40 Founder Traits: Sincerity, Boundaries, and AI-Enabled Blue Oceans

How AI Is Transforming Entrepreneurship and Knowledge Transfer

45:27 Pace of AI Development: Fast, but Not Yet So Fast That Startups Can’t Exist

46:33 Inventing the Wheel Isn’t Enough: You Must Bring It to People

47:00 Starting a Startup from Personal Problems: As Long as You Don’t Die, There’s Always a Chance to Succeed

47:39 Demo Day and Fundraising: What Happens After a YC Batch Ends

48:03 Pitching Skills and AI Training: From Speech Coaching to Office Hours Agents

49:07 Granola, Circleback, and Meeting-Note Agents

49:44 Risks of the Summary Economy: Could AI Summarization Cause Us to Lose Something Essential?

50:04 YC’s Bookface Agent: Offload Foundational Knowledge to AI, Free Human Time for Novel Problems

51:07 Prompts Must Be Written by Users Themselves: Why Enterprise AI and Closed-Source Models Face Fundamental Issues

Open Source Business and Social Media

52:51 How Open Source Projects Grow into Companies: Supabase, Managed Services, and the Future of Data Infrastructure

53:49 Linux and Linus Torvalds: An Open-Source Contribution Humanity Ought to Thank

53:49 How Social Media Changed the World: Not Creating Problems, But Making Them Visible

54:30 Teaching Kids to Evaluate Information: What on YouTube Helps You—and What Deceives You

55:13 Reorganizing YouTube with AI: Letting Agents Filter Beneficial Content

55:37 Garry and X: Influence, Responsibility, and Risk from 750,000 Followers

AI Has Only Reached 1%

56:18 How AI Has Already Changed the World: One Billion Users, Yet Only a Few Have Experienced Truly Cutting-Edge Capabilities

57:24 Why Frontier Models Are Still Too Expensive: Productivity Behind the $1M Annual Token Cost

58:13 YC’s Outrageous Success Stories: Coinbase, DoorDash, and Airbnb

59:17 What Changed After Garry Took Over: Returning YC to Its Roots—Supporting Early-Stage Founders

59:57 YC’s Future: In the AI Era, Small Teams Can Generate Hundreds of Millions in Revenue

Techno-Optimism and Little Tech

01:02:35 Human-Machine Relationship: Fire, Wheels, Books, PCs, the Internet, and AI

01:03:47 From Desktop Internet to iPhone, Then to Smart Computers

01:04:46 Techno-Optimism: Technology Should Serve Humans; Often, It’s the Market That Breaks

01:05:28 Little Tech: Beeper, Unified Messaging, and Breaking Big Tech’s Walled Gardens

01:06:10 Walled Gardens in the AGI Era: Why Access to Models Will Become Critically Important

The Multi-Model Era and Antitrust

01:07:20 AI Model Personalities: The Distinct “Vibes” of Claude, Codex, and DeepSeek

01:08:34 Can Grok Catch Up? More Competition Is Better

01:09:17 Should All Companies Go Public? Private Equity, Secondary Markets, and Opportunities for Ordinary People

01:10:06 Will Big Tech Block Little Tech? Markets, Governments, and Antitrust

01:11:24 Zero-Based Budgeting: How YC 3.0 Reexamines Everything from Scratch

Entrepreneurship, Vibe Coding, and the Hardware Renaissance

01:12:22 Does Entrepreneurship Require Historical Knowledge? Schlep Blindness and How Ignorance of Difficulty Can Help You Begin

01:13:18 What Vibe Coding Changes: The Engineers Most Opposed to AI May Be the Ones Who Benefit Most

01:14:04 YC’s Hardware Startups: Stoke Space, Astranis, and Reusable Rockets

01:14:54 Not Just Software: Empowering Engineers to Build the Coolest Things in the World

01:15:37 Robotics, 3D Printing, and U.S. Manufacturing Supply Chains

01:16:07 Nox Metals: Markets, Oligopolies, and How Government Antitrust Efforts Can Help Startups

01:17:00 Immigration, Entrepreneurship, and America: Why Starting a New Company Is a Noble Act

Personal Trauma, Startup Failure, and Inner Fire

01:17:54 If I Were 16 Today: Would I Still Go to Stanford?

01:18:13 Late Bloomers: Why Garry Didn’t Truly Start a Company Until Age 27

01:18:54 Posterous, Instagram, and the Pain of Missed Opportunities

01:19:30 Psychological Causes Behind Startup Failures: Co-founder Conflicts, Hero Complexes, and Unprocessed Trauma

01:21:01 Rick’s Counter-Observation: Trauma Can Sometimes Fuel the Breakthrough Fire

01:21:49 Garry’s Inner Room: His 7-Year-Old Self, Fear, and Liquid-Plutonium-Level Energy

01:22:47 How to Relate to That Inner Child: Use Him—Don’t Let Him Fully Control You

01:23:16 Self-Destruction and Extreme Control: Sobriety, Integration, and Finding the Middle Ground

01:24:27 Family, Work, and the Difficulty of Switching Off: A Reminder from Warren Buffett’s Regrets

01:25:18 Purpose and Connection: How to Take Off the Fire Suit and Come Home for Dinner

Things I Disbelieved When Young—but Later Accepted

01:26:31 Turning Down an Early Palantir Opportunity: How Security Can Make You Miss Huge Potential

01:27:30 Reunderstanding Self-Worth: Don’t Treat Your Role in a Large Organization as the Sole Source of Value

01:28:20 Getting Closer to the Source: Don’t Treat Media, Authorities, or Mentors as Final Answers

01:29:54 Killing the Mentor in Your Head: Respect Advice—but Walk Your Own Path

01:30:31 Hit the Core and Keep Moving Forward: When Opposition Lacks First Principles and Relies Only on Emotion, You May Be Nearing the Truth

🌟 Highlights

💡 “Make Something People Want” Remains the Core of Entrepreneurship

Garry Tan repeatedly stresses that YC’s core has never been fundraising, valuation, or financial tricks—but that classic motto: *Make something people want*. Money and metrics only follow once real value is created. In the AI era, building things is easier than ever, making it even more critical to ask: *Do people truly want this?*

“Money and numbers come only after you’ve made something people want.”

🧠 Personal AI and Open-Source Agents: The Next PC Revolution

Garry believes we’re standing on the eve of a moment akin to the birth of the Apple II. Today’s AI remains highly corporate, expensive, and hard to control. The truly transformative future lies in empowering everyone to own, run, and program their own AI agents. Projects like G Brain, OpenClaw, and the open-source model movement represent the early stages of personal AI.

“Anyone can actually join—and become a player and participant in this next revolution.”

🛠️ AI Amplifies Builders’ Value by 100x

In Garry's view, code gen and vibe coding enable someone who truly knows how to build things to accomplish dozens or even a hundred times more work than in the past. He himself has started coding again, releasing open-source projects, and believes that the engineers who might benefit the most from AI are precisely those who resist it the most.

"I'm now writing code, creating software that people can use, and I'm about to reach a hundred times the workload of the entire year of 2013."

🧭 Truly Outstanding Founders: Genuine, Unique, Connected to the Source

Garry says that ideas are important, but in the earliest stages, ideas are essentially a function of the person. YC looks for people who truly care about something, have strong initiative, and can perceive things that others cannot see. He describes such founders as having "a nuclear reactor inside them."

"You must really care about something. There must be something inside you."

🔓 Prompts Should Belong to Users, Not Companies

When discussing AI products, Garry makes a very critical point: prompts must be modifiable by users themselves. An AI prompt written by engineers and PMs at a large company cannot truly understand your life, your goals, or your judgment. This is one of the fundamental reasons why he supports open source, personal AI, and OpenClaw.

"Some Google engineer and PM wrote that prompt. That person hasn't lived your life and doesn't know you."

❤️ Trauma, Ambition, and Integration: Garry's Most Honest Self-Disclosure

In the second half of the program, Garry rarely talks about his childhood trauma, startup failures, psychotherapy, and family relationships. He admits that the "fire" driven by fear and anger gave him immense energy, but also led to self-destruction, difficulty shutting down, and difficulty truly connecting with others. This conversation makes the source of motivation for entrepreneurs complex and real.

"If you don't know who you are, or if you haven't dealt with what's truly happening inside you, these things will resurface."

🌐 Podcast Information Supplement

This podcast uses the original vocal tone for audio production, so some parts may sound a bit strange.

AI is used for translation, so there may be some parts that are not smooth;

If you would like to listen to other foreign-language podcasts in Chinese, feel free to contact WeChat: iEvenight.