This conversation features three Y Combinator partners: Gary Tan (CEO), Jared Friedman (Partner), and Harj Taggar (Partner), discussing how YC has evolved from 2006 to 2026. Gary did YC in Summer 2006 (third batch ever), Jared in Winter 2007, and Harj in Summer 2008, giving them unique perspectives as both founders and partners.

The discussion covers YC's core transformation process - taking earnest, technical people and making them formidable founders. They explore how AI coding tools like Claude Code are revolutionizing both startup building and YC's evaluation process, with Gary demonstrating by recreating his entire 2008 startup in just 90 hours.

The conversation spans current market dynamics, from prediction markets inspired by Kalshi to the consolidation of venture capital, while examining what types of companies remain safe from AI disruption. They also discuss YC's expansion to four batches per year, their outreach to college campuses, and Gary's political involvement in San Francisco reform efforts.

AI Coding Revolution Changes Everything for Startups

Gary recreated his entire 2008 startup (70,000 lines of code) in 90 hours using Claude Code, compared to the original two years with five engineers and anti-narcoleptics

"Something happened in November, end of November when Opus 4.5 came out... AGI is here, guys, for code" - Gary describing the inflection point

YC now accepts Claude Code transcripts in applications, evaluating founders based on how they prompt agents, use plan mode, and think about systems architecture

The advice has evolved from "write code, talk to users" to "prompt and talk to users" with dramatically less time spent on the prompting phase

Identifying Talent in the Age of AI Agents

"How you do anything is how you do everything" - partners can assess founder quality through their prompting transcripts, looking for systems thinking and edge case consideration

AI coding expands the net for founders like Parker Conrad, who wasn't traditionally technical but deeply understood products and could now build sophisticated applications

The "game recognize game" principle remains crucial - would other great founders want to work for this person, demonstrated through craft and attention to detail

Apoorv Mehta's smooth-scrolling iPhone demo app revealed exceptional craft when most iOS apps couldn't achieve that performance in the early App Store era

Market Dynamics and Competition in AI Era

"Capital as a bludgeon" works less effectively now - small teams like Giga (under 20 people from IIT India) can beat well-funded incumbents with superior technology

Companies are raising bigger rounds than ever despite needing fewer people, with founders feeling constrained by the expanded surface area of what products should accomplish

YC advises founders to ignore competition and focus on execution: "Make something people want. It doesn't say do a market map"

Prediction markets represent a major trend following Kalshi's regulatory breakthrough, with capital flowing to newly legal consumer businesses

What's Safe and Unsafe from AI Disruption

Marketplaces aggregating people (Airbnb, DoorDash) and systems touching money/regulatory frameworks (Rippling payroll) remain protected

"Salesforce is probably screwed" - CRMs lack the money-touching stickiness and integration moats that can now be built with Claude Code in 30 minutes

Hard tech remains protected because "the AI hasn't exactly come for atoms yet" - physical constraints still create natural moats

Sales remains unautomated unlike support, requiring human salespeople even as engineering productivity dramatically increases

YC's Structural Evolution and Scale

YC now operates as "seven or eight simultaneous 2008 batches" with 15 partners each running autonomous pods of ~30 companies

The expansion from two to four batches per year was enabled by decentralizing operations - each partner picks their own companies and runs mini-YCs

"The bottleneck for us now is getting more great founders to want to do startups" - operations and capital are no longer constraints

YC visited over 30 college campuses and expanded to grad students and mid-20s founders, countering the trend toward younger founders in AI

Building Culture and Political Engagement

Breakneck by Dan Wayne illustrates how America built extensively through the 1950s but became a litigious culture unable to build high-speed rail or major infrastructure

Gary's political involvement stems from seeing "Asian American grandpas and grandmas couldn't walk down the street without being assaulted and killed"

Matt Mahan built 1,400+ homes in San Jose and reduced homelessness by 20% through execution rather than virtue signaling, representing the model for effective governance

"We need markets that allow new startups and new entrants to come in" - fighting for little tech against big tech dominance through regulatory advocacy