Y Combinator CEO Garry Tan recently sparked a polarized debate at SXSW after revealing his intense fascination with AI coding agents. During an on-stage interview with investor Bill Gurley, Tan admitted to suffering from what he jokingly termed cyber psychosis, citing sleep deprivation as he works on multiple projects simultaneously. He compared the energy AI provides to taking anti-narcoleptic medication, noting that he no longer needs drugs to stay awake while overseeing his virtual team of ten agents. This excitement led Tan to publicly share his custom AI configuration, dubbed gstack, on GitHub on March 12. The gstack repository contains a set of opinionated prompts designed for Claude Code. These prompts, stored as reusable skills in specific files, instruct the AI to perform distinct roles within a software development workflow. Tan's setup simulates a full engineering team by assigning Claude different personas, including CEO, engineer, code reviewer, designer, and documentation specialist. The goal is to ensure that AI-generated code is reviewed for security and quality before deployment. Following the release, the project gained significant traction, trending on Product Hunt and accumulating nearly 20,000 stars on GitHub as users sought to replicate Tan's workflow. However, the reception was not universally positive. Tan's subsequent comments suggesting that gstack was god mode for instantly finding security flaws drew sharp criticism. Some industry figures argued that such claims were irresponsible, with one founder stating that a CTO making such a claim should be fired immediately. Critics, including vlogger Mo Bitar, dismissed the project as merely a collection of text prompts that experienced developers could already create themselves. Others noted that the project's visibility on platforms like Product Hunt might have been inflated by Tan's high-profile position at Y Combinator, suggesting the tool might not be as revolutionary if released by an unknown developer. To assess the validity of these claims, technical experts and rival AI models evaluated the setup. ChatGPT described gstack as a sophisticated prompt workflow rather than magic, emphasizing that the true value lies in simulating an organizational structure rather than a single request. Similarly, Gemini labeled the configuration as a professional-grade tool that prioritizes correctness over ease of use. Claude itself praised the system as a mature and well-designed example of Claude Code skill architecture, highlighting that it is built by someone who actively utilizes it daily. Ultimately, the controversy surrounding gstack highlights the rapid integration of AI into startup culture. While some view the tool as a breakthrough that democratizes high-level engineering practices, others see it as a hype-driven experiment. Tan maintains that the experience of coding with AI is unparalleled, describing it as a direct translation of thought into reality. His story reflects a broader trend where leaders are increasingly relying on AI agents to accelerate development, raising questions about the balance between innovation, safety, and realistic expectations in the tech industry.