Peter Yang (Product at Roblox): Asked the community about their preferred interface for switching between Claude and Codex — Composer, Cursor, or Terminal. Also wondering if “Project Hail Mary” is appropriate for his 7-year-old, and planning a trip to Shanghai/Hangzhou in early April after over a decade away.
Nan Yu (Head of Product @Linear): Engaged with discussions about startup survival rates, noting it’s remarkable that survival rates didn’t drop off a cliff despite many more startups being funded in recent years.
Thariq (Claude Code @AnthropicAI): Shared excitement about setting up new dev tools and figuring out how to stream the development process. Also promoted the agent SDK as the easiest way to build agents.
Ryo Lu (Design @Cursor_ai): Posted “still early. but clearer now.” — signaling growing conviction about the future of AI-powered coding.
Garry Tan (President & CEO @YCombinator): Open sourced his CLI toolset (gstack), calling it his personal toolset though “there are many like it but this one is mine.” Also commented on the importance of getting a good MacBook Pro.
Matt Turck (VC @FirstMarkCap): Reflected on the history of major platforms — LinkedIn (2003), YouTube (2005), Twitter/X (2006) — noting that network effects created massive long-term staying power, and raising the question of whether the AI era can build its own strong flywheels like data loops.
Zara Zhang (Builder, Harvard ‘17): Made a viral observation that almost every AI power user she knows is MORE stressed and busier after using AI. “What people thought AI would do: 10x productivity so we can finish work earlier and relax more. What it’s actually doing: 10x productivity so we end up with 20x more things to do because of the sheer possibilities.” Also posted about the importance of touching grass and reading books.
Nikunj Kothari (Partner @FPV Ventures): Posted about “2x humans, 10x mess, ∞ fun” — capturing the chaos and excitement of building startups.
Dan Shipper (CEO @Every): Bought $5k of Figma stock, expressing strong conviction: “Very bullish on SaaS adapting to AI, their stock is getting crushed right now, and [the CEO] isn’t gonna miss.”
No Priors — “Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI”
Andrej Karpathy (AI researcher, founder of Eureka Labs) described how the nature of coding has fundamentally shifted. “Code’s not even the right verb anymore,” he says — instead of writing code, he’s “manifesting” his will to AI agents for 16 hours a day. He recalls a moment when he realized something had flipped: “I just have to code for 16 hours a day or code’s not even the right verb anymore. I have to express my will to my agents for 16 hours a day. Manifest. Manifest.”
The conversation explored how AI agents are evolving from single sessions to multiple coordinated agents, how the “agent part” is now taken for granted, and how the next frontier is optimizing the instructions we give them. Karpathy described being in a state of “AI psychosis” — a perpetual sense that everything is now a “skill issue” because the tools have become so capable.