Iran, protests, and sanctions

What is the cost of ignoring your own direct experience in favor of the crowd? Gary Tan, President and CEO of Y Combinator, reflects on this through the lens of his most expensive mistakes—like turning down an early opportunity at Palantir because he was chasing the “status” of a corporate promotion. He argues that the most successful founders aren’t those who follow the current “hot” trend, but those who possess the courage to be earnest and double down on what they uniquely know, even when the consensus says that field is dead.

The conversation shifts to the radical transformation of company building in the age of AI. Tan describes a shift toward “Vibe Coding” and “Agenda Coding,” where a single founder can now operate with the productivity of hundreds of people. He envisions a future where “markdown files are employees”—where a perfectly executed business process is documented once and then automated via agents into a reusable skill. This allows tiny teams to scale to millions in ARR with minimal headcount, effectively erasing the need for the traditional, lossy middle-management layer that has plagued large corporations for decades.

Beyond the technical, Tan emphasizes the importance of “founder psychology” and civic stewardship. He discusses the “white pill”—the optimistic belief that while AI will disrupt the world, the inherent slowness of government and corporate bureaucracy provides a grace period for thoughtful implementation. He also shares his personal journey into San Francisco local politics, arguing that the same builder’s mindset used to scale a startup—identifying a bottleneck, ignoring the noise, and executing on a solution—is exactly what is needed to fix crumbling urban institutions and improve local communities.

Surprising Insights

• The “Markdown Employee”: The idea that a business process, once perfected and documented in a markdown file, can function as a permanent, flawless employee through AI agents.
• The Management “Distraction”: The theory that the last hundred years of managerial science were essentially a distraction from “founder mode,” and AI will now automate the “middle” of organizations.
• The 7±2 Limit: Most global institutions are built around the human limitation of only being able to hold roughly seven pieces of information in active memory; agents break this bottleneck, offering a form of organizational “clairvoyance.”
• The Bureaucracy White Pill: The comforting realization that society, government, and legacy companies are far slower to change than the tech cycle, meaning the “AI apocalypse” for jobs will likely be a slow, manageable transition rather than an overnight crash.

Practical Takeaways

• Avoid “LARPing” as a Founder: Stop asking “what is hot” and start asking “what do I uniquely know?” Trust your direct experience over the consensus of Twitter or the Wall Street Journal.
• “Token Max” Your Workflow: For founders and CEOs, don’t be afraid to spend more on high-end models and massive context windows (token maxing) to gain a strategic advantage and “live in 2028” today.
• Skillify Your Success: Every time you perform a complex task perfectly, turn the process into a “skill file” (markdown + code + tests) so it can be automated via a cron job or agent.
• Build Trivial Things to Learn: To build intuition for AI, don’t wait for a “big idea.” Use every new model to build completely trivial tools simply to create a “chassis” for learning the technology.