The 1,000x Developer Isn't a Myth. gstack Is the Evidence.
Garry Tan claims he ships at 810× his 2013 pace while running YC full-time. The math is in the tooling. Here's how gstack's 35 skills make that number possible.
The "10x developer" has been a fixture of engineering culture for decades — usually as a vague compliment for unusually productive engineers, occasionally as a target for ridicule, almost never as something precisely measurable.
Garry Tan changed the conversation this year.
In building gstack, his open-source AI-powered software factory, he documented something more specific: he's shipping at 810× his 2013 pace in terms of logical code changes — while running Y Combinator as President and CEO.
That number is not a typo. Let's look at what produces it.
The Architecture of Leverage
gstack isn't a single AI tool. It's an interconnected system of 35 specialist skills organized around a complete software development lifecycle:
Think (office-hours, autoplan) Plan (plan-ceo-review, plan-eng-review, plan-design-review, plan-devex-review) Build (design-consultation, design-review, design-html, browse) Review (review, codex, health) Test (qa, qa-only, benchmark) Ship (ship, land-and-deploy, canary) Reflect (retro, document-release, investigate)
Each skill represents a specialist role. Not a tool that helps a developer do their job faster — a specialist that does a job that previously required a different human.
The code reviewer. The QA lead. The security officer. The design critic. The release engineer. The documentation writer. The data-driven retro facilitator.
In a traditional team, assembling these specialists costs hiring budgets, coordination overhead, meeting time, and organizational complexity. In gstack, each specialist is a skill invocation.
What "810×" Actually Means
Let me try to make that number concrete.
In 2013, a solo developer shipping a feature would typically:
- Write code
- Manually test it
- Push to staging
- Do a rough self-review
- Deploy to production
- Monitor manually for a while
There was no automated review pipeline, no consistent design system enforcement, no adversarial security audit, no benchmark comparison against baseline. The velocity ceiling was set by one person's ability to wear every hat sequentially.
With gstack, the same developer runs:
/office-hours— validates the idea before writing a line/plan-ceo-review+/plan-eng-review— stress-tests the approach- Build phase (actual coding)
/design-review— visual audit against design system/review— multi-specialist code review (security, testing, performance, API contracts)/qa— systematic bug finding and fixing/ship— automated release pipeline/canary— post-deploy monitoring/retro— pattern analysis for continuous improvement
The quality floor of what ships is dramatically higher. The time spent on coordination, context-switching, and manual processes is dramatically lower.
The 810× isn't a productivity hack. It's the compound effect of removing every bottleneck simultaneously.
The Roles That Used to Require Headcount
The deepest insight in gstack's design is what it says about the nature of specialist knowledge.
Most of what a senior code reviewer does is apply a known framework to a new piece of code. They check for common vulnerabilities. They look for test coverage. They notice when an API change is breaking. They flag when a function is doing too many things.
This knowledge is codifiable. It's not replaceable — a human reviewer catches things no tool catches yet. But it's composable — a system that runs 6 specialist checks simultaneously covers more surface area than most human reviewers have time for.
Same logic applies to design review, QA, security audit, documentation, and retrospectives. These aren't creative problems that require unique human judgment. They're the application of known frameworks to specific artifacts. AI is very good at that.
What changes when you have all these specialists available on demand is not just speed. It's what you choose to build.
Projects that previously required a team are now within reach of a single builder with taste, judgment, and the right toolkit. The constraint shifts from "do I have enough people?" to "do I have a clear enough idea?"
The Uncomfortable Implication
If gstack's approach is directionally correct — and the evidence suggests it is — then the question it raises for organizations is uncomfortable.
What is a five-person engineering team optimized for if one person with good AI tooling can do 810× the work of a 2013 developer?
The honest answer is: the highest-value activities that don't reduce to applying known frameworks to artifacts. Novel architectural judgment. Creative problem definition. Relationship-dependent decisions. User research that requires genuine human empathy.
Everything else is being automated. Not in a decade — right now, today, with tools anyone can install.
The 1,000x developer isn't a myth. It's a choice — about whether you develop the taste and habits to use leverage well, or whether you continue working the old way while wondering why some people seem to move so much faster.
gstack is the evidence. The choice is yours.
AI Skill Daily 001. Part of the gstack series — 35 specialist skills from garrytan/gstack.