Y Combinator CEO Garry Tan says he stayed up till 5 AM using Claude Code

Claude AI

• Y Combinator CEO Garry Tan praised Anthropic's new AI coding tool, Claude Code
• Claude Code operates in the terminal, managing codebase tasks using natural language.
• AI tools like Claude Code are moving software engineering to higher-level roles.

Did our AI summary help?

Y Combinator President and CEO Garry Tan has publicly praised a new AI coding tool from Anthropic, saying it kept him awake almost all night. In a post shared on X, Tan said he was “so addicted to Claude Code” that he worked for 19 straight hours and did not sleep until 5 AM.

Claude Code is an AI-powered coding tool developed by Anthropic and made available more widely in October 2025. Unlike traditional AI coding assistants that live inside an integrated development environment, Claude Code runs directly in the terminal. Developers can use natural language commands to write, fix, explain, and manage code across an entire project.

One of Claude Code’s key strengths is its ability to understand a full codebase rather than isolated files. This allows it to add features, fix bugs, write tests, manage commits, and even help with pull requests. By integrating directly into existing command-line workflows, the tool effectively behaves like an assistant programmer that does not require developers to change how they work.

Tan’s comment arrives at a moment when the software development industry is undergoing a major shift. Once considered one of the most labour-intensive areas of the tech sector, software engineering is increasingly being shaped by automation and AI. By early 2026, many expect AI to move from a supporting role into a central position in how code is written, maintained, and scaled.

This transition has led to growing debate about whether traditional, line-by-line coding is becoming obsolete. The idea is not new, but it is gaining momentum as tools like Claude Code demonstrate how quickly complex tasks can now be completed. Code that once took weeks or months to implement manually can often be generated, tested, and refined in seconds.

A similar view was recently shared by Ryan Dahl, the creator of Node.js. In a post on X, Dahl argued that the era of humans writing syntax directly is effectively over. He acknowledged that this shift may be unsettling for software engineers who strongly identify with hands-on coding, but maintained that the change is unavoidable.

According to Dahl, developers are not becoming irrelevant. Instead, their responsibilities are evolving. While AI can handle repetitive implementation and boilerplate code, humans remain essential for design decisions, system architecture, problem-solving, and oversight. In an AI-driven development environment, higher-level thinking is becoming more valuable than raw typing speed or memorising syntax.

This evolution represents a significant cultural change within software engineering. Many developers built their careers around writing and optimising code by hand, and moving away from that identity can feel uncomfortable. Even Dahl has described the transition as “disturbing” for some engineers.

However, the broader picture suggests opportunity rather than decline. As AI tools take over routine tasks, developers can focus more on strategy, creativity, and building systems that solve real-world problems. Adapting to new tools, learning continuously, and guiding AI rather than competing with it are becoming core skills.

Garry Tan’s late-night experience with Claude Code may sound extreme, but it reflects a wider reality. AI-powered coding tools are no longer experimental curiosities. They are becoming central to how modern software is built, and they are reshaping what it means to be a software engineer in the process.