Garry Tan is president and CEO of Y Combinator, the startup accelerator, and an engineer who publishes tools for working with AI agents. He co-founded Posterous and Initialized Capital. His agent projects, including gstack and GBrain, make his own development and information-management workflows available to other builders.[1][5][6]
Start with gstack →Background
Tan graduated from Stanford in computer systems engineering in 2003. Before founding companies, he worked on Windows Mobile at Microsoft and as an early designer and engineering manager at Palantir. He later co-founded Posterous, a blogging service acquired by Twitter. These experiences place his public advice in the context of building products as well as funding them.[1][4]
Career
Tan was a Y Combinator partner from 2011 to 2015 and built internal software including Bookface and the Demo Day website. He left to focus on Initialized Capital, which he co-founded, and returned to lead YC after Geoff Ralston announced he would leave at the end of 2022. His current YC biography lists him as president and CEO.[1][2]
Contributions
Structured roles for coding agents
gstack packages Tan’s Claude Code setup into tools for planning, design, engineering review, quality assurance and releases. The roles make different kinds of judgment explicit instead of treating every request as an implementation task. Developers can inspect those instructions and adapt a workflow to their repository. The repository documents Tan’s preferred process; its existence does not establish that an agent’s output is correct without verification.[5]
Memory with a source trail
GBrain stores facts together with their sources and supports correction and withdrawal. Its purpose is to let agents retrieve a durable record rather than repeatedly reconstructing context from old conversations. Tan describes building it for his own agents. The project distinguishes local memory from shared and hosted access, so permissions and deployment choices remain part of using it responsibly.[6]
Ideas
After Twitter acquired and later closed Posterous, Tan co-founded Posthaven with Brett Gibson. In his account, its subscription model supports a commitment to keeping the service available. This history gives context to his interest in durable software: a useful tool also needs an operating model that lets people continue relying on it.[3]
Tools, writing & talks
Tools
Start here
gstack
github.comA collection of explicit planning, review, QA and release workflows for Claude Code. Inspect a skill before using it in a real repository.
Read the gstack README and one review skill to see how Tan turns a development process into explicit agent instructions. You will get more from it if you already use a coding agent on a repository.
Creator / maintainer[5]Original source- GBrain
github.comAgent memory that retains source provenance and supports corrections, retrieval and shared access.
Creator[6]Original source
Apps
Writing, research & talks
Jepsen's Kyle Kingsbury, who spent a decade proving distributed databases broke their own consistency guarantees, turns the same methodical scepticism on language models.
Catalog workOriginal sourceLangSmith gives you trajectory evals, trace-to-dataset pipelines, model-as-judge and regression suites, but tools are not a practice: nothing tells you what to test, in what order, or when you are finished.
Catalog workOriginal sourceA follow-up on thin harnesses and fat skills, arguing that resolvers, the routing layer deciding which skill handles what, matter more than the skills that got all the attention, and are invisible precisely when they work.
Catalog workOriginal sourceA short essay arguing that productive AI coding agents come from a lightweight harness plus composable skills. Explains the context-management and workflow ideas behind the '10-100x' claim. An opinion piece, not a harness you can run.
Catalog workOriginal sourceFigma's Dylan Field and YC's Garry Tan discuss why current AI stays a 'tool' that widens exploration but hasn't closed the design-quality gap, and why craft and judgment still matter when models make prototyping nearly free.
Catalog workOriginal source- Unconventional Advice for Foundersecorner.stanford.edu
A Stanford talk in which Tan discusses decisions and assumptions that shape a founder’s work.
Speaker[8]