Why Garry Tan’s Claude Code setup has gotten so much love, and hate

Y Combinator CEO Garry Tan has released an open-source AI coding framework, dubbed "gstack," sparking both intense interest and sharp criticism across the tech community. Speaking at SXSW, Tan claimed working with AI agents has induced a state of "cyber psychosis," an obsession so powerful it has replaced his past reliance on focus-enhancing drugs like modafinil. The public release of his personal setup for the Claude AI has ignited a debate over whether it represents a meaningful advance in developer workflows or is simply a common practice amplified by Tan's influential status.

Technically, gstack is a GitHub repository containing a set of reusable prompts, or "skills," that instruct Anthropic's Claude Code to perform specific roles within a simulated software team. The workflow involves using different skills to have the AI act as a CEO to validate an idea, an engineer to write the code, and a code reviewer to check for bugs and security flaws. Tan claims this approach allowed him to replicate a startup that previously required 10 people and $10 million in venture capital. The project quickly went viral, accumulating nearly 20,000 stars on GitHub and trending on Product Hunt.

The divided reaction to gstack highlights a core tension within the AI ecosystem regarding what constitutes genuine innovation. Critics have dismissed the project as "a bunch of prompts," arguing that experienced developers already build similar personal setups. However, other AI models, including ChatGPT and Gemini, assessed the framework as a "sophisticated" and professional configuration. The core value, they suggest, lies not in the prompts themselves but in demonstrating that AI coding is most effective when it simulates an established organizational structure. The incident brings into focus the ongoing discussion about the real value of prompt engineering versus the power of personal branding in the AI space.

The gstack debate is less about the technical novelty of prompt chaining and more about the value of codifying and sharing an opinionated, high-level workflow. It signals a shift from simply prompting an AI to strategically structuring human-AI interaction in a way that mimics proven organizational models for software development.