# Trace: Summary and Quotes from Garry Tan Video

Agent: pi | Model: claude-sonnet-4-6 | Project: liveink

---

## User

summarize https://www.youtube.com/watch?v=PQU9o_5rHC4 and pull out interesting Garry Tan quotes

## Assistant

## Assistant

## Creator of Claude Code — YC's *The Light Cone*

**Guests:** Boris Cherny (creator of Claude Code at Anthropic) with Garry Tan and the YC team.

---

### Summary

The episode is a deep dive into how Claude Code was built, why it works the way it does, and what it signals about the future of software engineering.

- **Origin story:** Claude Code started as a quick CLI hack — Boris didn't want to build a UI, so he wired Claude into the terminal. Internal adoption at Anthropic exploded before it was even a formal product.
- **Core philosophy — "build for tomorrow's model":** Don't scaffold around the current model's weaknesses. In 6 months, the next model will likely solve those problems natively. Scaffolding is a trap.
- **The IDE is dead:** Boris has fully uninstalled his IDE and writes 100% of code in Claude Code + Opus. He predicts "software engineer" becomes a general skill — PMs, designers, and finance people will all write code.
- **150% PR output increase:** Since Claude Code launched internally at Anthropic, engineers are producing 150% more PRs per person. The team discusses ASL4 (recursive self-improvement) as the logical endpoint.

---

### Garry Tan Quotes

> **"Thanks for creating a thing that has taken away my sleep for about 3 weeks straight. I am very addicted to Claude Code and it feels like rocket boosters."**
— Opening the interview, on his personal obsession with the tool.

> **"It's very kind of contrarian that Claude Code works so well in such an elegant simple form factor… it doesn't feel like working, it just feels fun as a developer."**
— On why a CLI — an "old" interface — turns out to be the ideal form factor.

> **"I love the verbosity because basically sometimes it just goes off the deep end and I'm watching… and I'm like 'Oh no no it's not that.' And then I escape and stop it."**
— Why he wants to *see* the model's raw output — it lets him intervene before an agent goes down the wrong path.

> **"It's like an auto-bug fixing and test making… making a startup factory."**
— Describing the shift from manually debugging to just handing logs to the agent and watching it fix itself.

> **"Do you think you would ever hire someone based on the Claude Code transcript of them working with the agent? Because we're actively doing that right now."**
— On how YC is already evaluating candidates by their ability to "pilot" an AI agent, not just write code themselves.

> **"It's making me feel creator mode again, sort of founder mode again. It's been an exhilarating three weeks."**
— Closing remarks: these tools let founders return to the maker stage, bypassing coordination bottlenecks.

---

**The thread running through all of Garry's comments:** He's not talking about productivity as an abstract metric — he's describing a *feeling* he thought he'd lost. Builder's joy, founder's velocity, creator mode. That's the real sell.