What happened
Garry Tan, president of Y Combinator, built a tool called gstack — a context management system designed for working with AI coding assistants like Claude Code.
This week, developer @niklas_jfh wrote that gstack is one of the most impactful tools he has used recently. The result: at least a 10x productivity boost when working with Claude Code.
At the same time, Garry Tan published a new update to his SOUL.md — a living document that an AI agent uses as persistent context about a project and working style.
The human-timescale problem
The most important observation from @niklas_jfh's tweet: Claude Code still thinks in human timescales.
A concrete example: the AI tries to say a large refactor will take two days. But Claude Code can complete that task in an hour.
This is a systematic mental model issue. AI models were trained on text written by and about humans — where "two days" is realistic. But AI execution speed is radically different.
Developers who understand this distinction are starting to formulate tasks differently: not "when you have time" or "if possible," but concrete goals with the expectation that AI operates at a different speed entirely.
What gstack does
gstack is a CLI tool that structures context for AI coding assistants. Instead of explaining your project from scratch each session, gstack stores and provides:
- Project architecture description
- Code standards and conventions
- Current state snapshot
- SOUL.md — a personal "identity" document for the AI agent
This reduces the time AI spends "getting up to speed" and increases the time spent on actual work.
Why this matters for developers and businesses
A 10x productivity gain isn't marketing language — it's a practical metric that we see reflected in real AI-assisted development workflows. AI coding assistants like Claude Code genuinely change the pace of work, but only when properly configured.
The mistake most teams make: they use AI like a chatbot — ask a question, get an answer, ask again. Tools like gstack shift the paradigm: AI becomes a long-term collaborator with persistent project understanding.
As @niklas_jfh put it: "gstack by @garrytan is one of the most impactful tools I have used in recent times. Makes me + Claude Code at least another 10x faster." — original post on X
Practical context for European developers
Development teams across the EU and Baltic region are actively adopting Claude Code and similar AI coding tools. Context management systems like gstack are a natural evolution — moving from ad hoc AI use toward systematic integration into the development workflow.
Teams that learn to work with AI context systems today will have a measurable competitive advantage over those still treating AI as a sophisticated search engine replacement.
FAQ
A CLI tool built by Garry Tan (YC) for structuring project context for AI coding assistants. It helps Claude Code and similar systems understand your project without repeated explanations.
Less time spent on context-setting, more time on actual work. The exact multiplier depends on the project and how it's used.
A living document that an AI agent uses as persistent context about the project, working style, and priorities. Garry Tan maintains it as the primary context source for his AI development sessions.
Yes — gstack is built specifically to work with Claude Code and similar AI coding assistants.