Garry Tan’s gstack, a set of prompt-based configurations for Claude Code, has become a lightning rod in AI circles. Shared on GitHub in mid-March 2026, the project promised to replicate the structure of a software engineering team—CEOs, engineers, code reviewers—via AI agents. Tan, Y Combinator’s CEO, described using gstack to simulate a $10 million startup with “10 people,” surviving on four hours of sleep and declaring himself afflicted with “cyber psychosis.” The tool now has 20,000 GitHub stars but has also triggered vitriol, with critics dismissing it as “a bunch of prompts” and accusing Tan of leveraging his YC credibility to hype a trivial trick.
This debate reflects a deeper tension in AI development. Proponents argue gstack’s value lies not in novelty but in structure: by simulating org charts, it forces AI to produce more coherent, correct code. Anthropic’s Claude, queried on the matter, called it “sophisticated,” noting it shifts from “build this feature” to “review this code for XSS vulns.” Competitors ChatGPT and Gemini were similarly gushy. Yet skeptics highlight a paradox: coding tools like Claude Code already enable role-based prompting. If gstack is just a configuration, why the mania?
The love and hate follow a predictable pattern. Enthusiasts frame gstack as the future—democratizing tech startups with AI. The rage stems from the same engineers who dismissed Elon Musk’s “neural laces” and Meta’s AI metaverse, fearing overhyped tools drowning in buzzword soup. The CTO who called it “god mode” for finding XSS flaws became a Rorschach test: to some, a testament of power; to others, proof of Silicon Valley’s delusion.
What’s missing from the discourse is data. No independent benchmarking exists to quantify gstack’s productivity gains or security improvements. Its GitHub metrics measure virality, not utility. Worse, the tool’s open-source license allows forking, but few have shared results showing real-world impact. The loudest claims—Tan’s modafinil comparisons, the CTO’s “90% adoption” prediction—remain untested.
Looking ahead, the fate of gstack hinges on two variables: whether open-source developers refine it into a robust framework or it drowns in the Sisyphean churn of AI hype. Watch for Y Combinator alumni to adopt gstack in their startups—a signal of institutional validation—or for major AI firms like Anthropic to fold its concepts into paid platforms, siphoning its open-source energy into proprietary pipelines.