Garry Tan: AI Will Replace Jobs, Create New Billionaires, and Change Everything
In a Nutshell
Garry Tan argues that AI agents will act as personal operating systems—costing $100k–$200k today but dropping dramatically in 2–3 years—that give individuals control over information and productivity, potentially creating 400× output gains and new billionaires while displacing jobs. He warns that without aggressive open-market policies, antitrust enforcement, and open-source alternatives, AI power will consolidate into five dominant “kings,” mirroring the 1890s robber-baron era. YC’s mission is therefore to fund talent, push regulation that keeps platforms open, and steer AI toward societal cohesion rather than division or oligarchy.
These notes were generated by AI and may contain inaccuracies.
Y Combinator receives 80,000 to 90,000 applications annually for approximately 800 spots, resulting in less than a 1% acceptance rate. Between 8 to 10% of applicants receive an interview. The final interview stage lasts only 10 minutes, a constraint that Paul Graham determined was sufficient after testing shorter durations. The process asks founders to clearly identify the user, the problem, and how the technology solves it, followed by why the founder is the right person to build the solution.
Approximately 10 to 15% of all self-made billionaires age 50 and under are YC alums. Examples include Elon Musk, Bill Gates, Steve Jobs, Drew Houston from Dropbox, Brian Chesky from Airbnb, Brian Armstrong from Coinbase, and Tony Xu from DoorDash.
YC tests its decision-making by asking whether it would fund transformative founders like Elon Musk, Bill Gates, Steve Jobs, Drew Houston, Brian Chesky, Brian Armstrong, or Tony Xu in a 10-minute interview. If the answer is no, the organization would need to fundamentally retool its process.
Garry Tan's startup Posterous was described as "Dead Simple Blogs by email." The company had a strong demo that Paul Graham, Jessica Livingston, and Trevor Blackwell immediately understood without extensive grilling. Tan was funded based on this straightforward presentation.
Tan cites a Jay-Z lyric about people wanting to give advice without having done the work themselves. The act of being a technology investor is described as bewildering because the industry sells money in a space where more capital seeks opportunities than good opportunities exist. When YC started in 2005, funding typically went to people with connections through Stanford Business School, golf buddies, or prep school networks like Harvard-Westlake, representing an access-based system.
YC aspires to be a meritocracy but acknowledges it will never fully achieve this ideal. The organization recognizes that meritocracy is a virtue worth attempting despite its impossibility. YC provides $500,000 to accepted companies, a significant increase from the original $18,000 when Tan participated. The increased funding aims to allow people without rich parents to participate in startup culture.
Tan was one of Palantir's earliest employees and had the opportunity to observe Peter Thiel and Alex Karp. Karp was described as a philosopher rather than a technologist, with extensive world knowledge. During a customer demo trip to Washington D.C., the 25-year-old Tan questioned why the team was staying at the Mandarin Oriental rather than a budget option like Motel 6. Karp explained that staying at a budget hotel would signal to customers that Palantir was "Motel 6 software," when the company needed to project the quality of "Mandarin Oriental Software."
Tan argues that true free market enterprise has never been fully implemented. He cites the DOJ consent decree against Microsoft as enabling Google's rise by preventing Microsoft from using its Windows platform advantages to dominate the browser market. Tan contends Google now engages in similar self-preferencing, with decreasing percentages of Google search results showing non-Google properties. The Delaware corporate structure legally requires maximizing shareholder value, which drives this self-preferencing behavior.
Tan supports antitrust enforcement to prevent self-preferencing and advocates for open markets. He cites Europe's DMA as a positive example requiring big tech to open platforms, provide data transparency and privacy, enable portability, and allow reverse engineering. He argues big tech is too powerful and that "little tech" serves as the antidote, comparing current dynamics to robber baron railroad monopolies.
Despite new AI companies like Anthropic and OpenAI emerging successfully, Tan expresses concern that AI companies may become the new big tech. Critics suggest these companies might gatekeep access to frontier models as AI competition intensifies.
YC functions with union-like characteristics, providing a voice for startup policy positions on issues like California's billionaire tax and antitrust laws. Tan discovered that millions of dollars in astroturf funding supported groups claiming to represent small software companies while actually serving big tech interests. YC maintains a minimal DC presence with one full-time hire, Luther Low, operating on a shoestring budget compared to big tech lobbying expenditures.
Approximately 50% of successful startups have at least one immigrant co-founder. Tan's own background includes his mother's family arriving as refugees from Burma and his father immigrating from Singapore. This immigrant-driven innovation is described as uniquely American.
Tan's father experienced periods of unemployment and alcoholism. As an 11-year-old in 1992, Tan developed heightened sensitivity to his father's state upon returning home, learning to read people's inner states and reactions. This "scanner" ability now serves as a valuable skill for understanding founders during YC interviews.
Tan advocates funding AI companies that treat agents as core operating systems rather than features. YC has developed an internal AI agent named Straylight with personality and access to the complete context of all founders, emails, and applications. Unlike humans limited to 7 plus or minus 3 items in working memory, AI agents can process and synthesize information equivalent to three Harry Potter books, with capacity expected to grow to 5, 10, or 20 books within a year.
Tan references Peter Steinberger of Adept, noting that current AI implementations prioritize repeatability and safety over the more human-like qualities of effective AI agents. Tan uses AI agents to read all his email, Dropbox files, and Apple Notes including 20 years of therapy notes. For interview preparation, an AI agent downloaded and transcribed the top 10 podcasts of his interviewer, then performed diarization with line-by-line, timestamp analysis to extract major themes.
The current AI agent experience costs Tan approximately $100,000 to $200,000 annually in tokens. He predicts that within 2-3 years, everyone will use AI agents as "Jiminy Cricket on your shoulder" - autonomous assistants that know users personally, help see around corners, and run ahead with suggestions. This capability remains expensive and represents a privilege not available to everyone.
Tan believes technology is overvilified and that AI safety concerns should be addressed by targeting misuse rather than stopping development. While acknowledging short-term job displacement, he advocates for retraining and tools like his open source project GStack, which helps YC founders achieve 400 times more output using Claude Code. He envisions medical billers and other displaced workers pairing domain expertise with AI coding capabilities to start companies in their fields.
YC evaluates companies for ethical components beyond profit potential. Gustaf Alströmer, formerly head of growth at Airbnb and now a YC partner, reviewed a company focused on ADHD medication distribution for children. Despite strong founders and clear profit potential, the company raised concerns about societal need versus profit motives. The company was eventually sanctioned by the FDA and shut down.
Y Combinator operates as an institution where accountability matters. The organization bears responsibility not only to generate returns but also to fund the futures that align with desired societal outcomes.
Garry Tan proposes funding social media platforms that strengthen society rather than divide and enrage users. He identifies personal AI agents as a direct mechanism to achieve this outcome. Rather than downloading Facebook or Instagram apps, users would interact with social networks through APIs mediated by their own cognitive filters. This approach returns control of information consumption to individuals.
Programming Personal Algorithms
AI agents that can code will understand user goals and preferences. For example, a user focused on self-improvement could program their agent to filter out national politics while retaining local city politics. Family members could also contribute to programming these filters. This capability was previously impossible because individuals lacked both programming skills and the technical infrastructure to create beneficial algorithms.
Current social media algorithms prioritize engagement through outrage and division because this model generates maximum revenue for hosting companies. Personal AI represents a shift toward user-controlled systems similar to the personal computer revolution Steve Jobs initiated in 1984. The 1984 Macintosh commercial positioned personal computing as resistance against centralized control by IBM.
The next year will feature a defining "browser war" centered on AI agents. The competitive landscape will likely consolidate to two or three dominant platforms, analogous to operating system competition between Mac and PC. Current major AI labs including Anthropic, OpenAI, and Gemini may consolidate into what industry insiders refer to as the "five kings" - a scenario where all commercial value collapses into five dominant entities.
Y Combinator works daily to prevent this consolidation through policy advocacy, funding next-generation talent, ensuring open markets, and supporting open source alternatives. Tan advocates for an open-source operating system for personal AI, comparable to Linux, that remains accessible and affordable enough for widespread adoption.
Smart regulation should focus on creating open app markets rather than asset seizure approaches proposed by figures like Bernie Sanders. Tan opposes asset seizure, citing his background as an immigrant whose family fled such policies. Effective regulation would prevent companies like Apple from claiming exclusive safety authority over iMessage, which currently blocks AI integration and stifles market competition.
Preventing Market Closure
Allowing trillion-dollar companies to maintain closed ecosystems prevents new startups from emerging. Open source AI development faces particular challenges in the United States, where no open-source AI companies currently exist, while Chinese companies benefit from distillation capabilities unavailable to American and European firms.
The current situation parallels the 1890s technology explosion and subsequent concentration of power among robber barons. Historical responses by Theodore Roosevelt and Louis Brandeis against Standard Oil and railway monopolies demonstrate that markets can be opened to competition. Brandeis's trust-busting philosophy remains relevant for addressing self-preferencing and market concentration in technology.
Two concerning scenarios exist: populist left movements implementing socialist systems, or technology becoming controlled by too few entities without democratizing effects. Both outcomes require opposition. The preferred middle path follows American historical precedent of creating markets and enabling prosperity without making wealth generation illegal.
Despite criticism from figures like AOC, billionaires and large companies like Google serve valuable functions in organizing information and creating prosperity. OpenAI emerged as a Y Combinator spinout, demonstrating how established platforms enable new ventures. The goal should be more billionaires, more choice, and more prosperity rather than eliminating successful companies or implementing mass asset seizure.
A middle American approach exists between big tech oligarchy and asset seizure. This path emphasizes creating markets, enabling new business formation, and maintaining the defining American trait of allowing prosperity generation while preventing excessive market concentration.
Keep The Builders in your library
Save the videos and channels worth coming back to, and find them again in one place.
