Garry Tan's Claude Code setup: thousands tried it, opinions split

Garry Tan, CEO of Y Combinator, published his personal Claude Code configuration on GitHub, and the developer community did not stay indifferent. Thousands of people downloaded, studied and tested his settings, splitting roughly evenly: some were delighted, others sharply criticized the approach.

Y Combinator is one of the most influential accelerators in the world, with Airbnb, Stripe, OpenAI and hundreds of other companies behind it. When its CEO publicly shares his working toolkit, it automatically draws the attention of the professional community. Tan is not a typical venture investor: he actively writes code, experiments openly with AI tools and documents the process, which makes his configuration especially interesting as a reference point.

Claude Code is Anthropic's command-line assistant that lets developers work with code directly from the terminal through a dialogue with a language model. The tool is gaining popularity among developers who want to build AI into their usual workflow without switching between applications. The configuration defines the assistant's behavior: its style, constraints, context and automations. That is why someone else's config is not just a set of settings but a snapshot of their philosophy of working with code.

Reactions in comments and threads were mixed. Some developers saw the post as a valuable template and began adapting the approach to their own tasks. Others criticized it as excessive or out of step with the real work scenarios of an ordinary engineer. Some noted that the configuration is optimized for someone who evaluates startups, not someone who writes production code for hours every day.

A separate talking point was that the language models themselves, Claude, ChatGPT and Gemini, gave their own views on Tan's settings. That gave the discussion an unexpected and telling character: AI systems were assessing how people configure them. This twist reflects a new reality: users increasingly treat models not as a tool with fixed behavior but as a co-author whose opinion they can ask about their own workflow.

In a broader context, the Tan configuration story illustrates a shift in development culture. Programmers used to share dotfiles, snippets and vim configs. Now system prompts, CLAUDE.md files and AI agent settings have joined them. GitHub is gradually becoming a place where developers show off not only their code but also their relationship with artificial intelligence.

The polarized reactions are easy to explain: the market for AI development tools has not yet settled, and every specialist builds their own workflow. A configuration that is ideal in one context can be useless or even a hindrance in another. This is not a weakness of Tan's approach; it is simply a sign that tools like Claude Code are in a phase of active exploration, where there are no best practices yet but plenty of live experiments.

The post's viral success points to the main takeaway: developers are actively looking for guidance on how to build AI into their coding work sensibly, and they are ready to learn from others' experience, even if they then argue over the details. In that sense, any config that makes the community think and debate has already done its job.