Garry Tan System Map

Funnel and CTA Teardown

How OSS adoption, YC hiring, and founder demand connect.

gstack = methodgbrain = continuitygbrain-evals = proofYC = network

Funnel and CTA Teardown

Product Frame

The gstack/gbrain system is open-source product marketing, founder education, recruiting, and YC brand reinforcement in one surface. The CTAs are not random. They move from "try this tool" to "build this way" to "work at YC."

CTA Inventory

SurfaceCTAJob
gstack README opening"Fork it. Improve it. Make it yours."Convert criticism or curiosity into hands-on adoption
Quick startInstall, run /office-hours, /plan-ceo-review, /review, /qaGet first value in minutes
Team modecommit gstack into a repo for teammatesTurn individual usage into team/process adoption
OpenClaw sectioninstall gstack and route coding tasks through Claude Code sessionsConnect into multi-agent workflows
Native OpenClaw skillsinstall office-hours, ceo-review, investigate, retro via ClawHubLet OpenClaw users sample the methodology without Claude Code
Other AI agentssetup supports Codex, OpenCode, Cursor, Factory, Slate, Kiro, Hermes, GBrainMake gstack a portable workflow standard
License/footer"Free, MIT licensed, open source. No premium tier, no waitlist."Remove buyer friction and maximize distribution
Hiring block"Come work at YC" with ycombinator.com/softwareConvert inspired builders into candidates

Funnel Stages

1. Belief Shift

Claim: one person with the right tooling can ship like a team.

Evidence used:

User state: skeptical but curious.

Conversion target: accept that the workflow might be real enough to try.

2. Fast Trial

CTA: clone and run setup, then run /office-hours or /qa.

Why it works: the user does not need to understand the whole system. The fastest magical moment is a skill that pushes back on a product idea or tests a real URL.

User state: wants proof without commitment.

Conversion target: first meaningful agent output.

3. Workflow Adoption

CTA: team mode, repo bootstrap, auto-update, CLAUDE.md guidance.

Why it works: after individual value, the product asks for process adoption. This is where gstack becomes team infrastructure, not a personal prompt pack.

User state: "this helped me; how do I make it standard?"

Conversion target: repo-level install.

4. Ecosystem Expansion

CTA: use with OpenClaw, native ClawHub skills, other coding agents, GBrain.

Why it works: gstack positions itself as a workflow layer above model/client choice. That increases surface area and reduces lock-in objections.

User state: "I already use another agent."

Conversion target: methodology adoption, even if host differs.

5. Recruiting and YC Affinity

CTA: "Come work at YC" and the YC software page.

Why it works: the tool is the proof. A builder who wants to harden this way of working is already self-selecting for YC software's desired persona.

User state: "I want to work on this frontier."

Conversion target: YC software lead or applicant.

YC Layer

The YC software page asks interested builders to get in touch at software@ycombinator.com with a brief note about themselves and what they would want to build. The broader YC application page frames YC as full access to YC, $500k investment, San Francisco in-person batch, late applications considered, interviews, and ongoing alumni support.

Product implication: the OSS funnel has two exits:

Messaging Strategy

The voice has four traits:

What Ren/OpenClaw Should Learn

Exact CTA Inventory, Deep Pass

SurfaceExact CTA / CopyFunnel Job
gstack opening proof"How does one person ship like a team of twenty?"Establish the category problem: individual leverage through agents.
Founder credibility"I'm Garry Tan, President & CEO of Y Combinator..."Borrow YC authority and personal builder credibility.
Proof claim"In the last 60 days: 3 production services, 40+ shipped features..."Turn the repo from prompt pack into proof-of-work.
Main OSS CTA"Fork it. Improve it. Make it yours."Convert curiosity or criticism into hands-on adoption.
Try-first CTA"I'd rather you just try it first."Lower ideological resistance; do not debate AI coding, run the workflow.
Audience CTA"Founders and CEOs", "First-time Claude Code users", "Tech leads and staff engineers"Segment the reader by job, not by tool preference.
Quick startInstall gstack, run /office-hours, /plan-ceo-review, /review, /qa, then "Stop there. You'll know if this is for you."Force a short path to first value.
Install CTA"Open Claude Code and paste this. Claude does the rest."Make install feel agent-native, not developer toil.
Team adoption"Team mode - auto-update for shared repos (recommended)"Convert personal workflow into org standard.
OpenClaw CTA"Paste this to your OpenClaw agent"Make OpenClaw a distribution surface for gstack methodology.
Native OpenClaw CTAclawhub install gstack-openclaw-office-hours...Let users sample gstack without Claude Code.
Other agents"gstack works on 10 AI coding agents, not just Claude"Reframe gstack as workflow standard, not Claude dependency.
Host expansion"Want to add support for another agent?"Recruit contributors into adapter growth.
See it workDaily briefing app transcript ending in shipped PRDemonstrate end-to-end sprint, not a feature list.
Competitive redirect"If you already use Karpathy-style CLAUDE.md rules, gstack is the workflow enforcement layer..."Absorb adjacent practice into gstack.
Parallel sprint claim"10-15 parallel sprints"Sell the next-level operating model after one sprint works.
ReversibilityFull uninstall sectionReduce install fear; mature OSS posture.
Trust footer"Free, MIT licensed, open source. No premium tier, no waitlist."Remove buyer/sales friction.
Hiring CTA"We're hiring. Want to ship real products at AI-coding speed and help harden gstack? Come work at YC"Convert strongest-fit builders into candidates.
gbrain upgrade"GBrain - persistent knowledge for your coding agent" and /setup-gbrainMove user from process to continuity.
gbrain install"Your AI agent is smart but forgetful. GBrain gives it a brain."Name the memory pain sharply.
gbrain agent install"Paste this into your agent"Position install as delegated work.
gbrain eval CTABrainBench scorecards and sibling gbrain-evals repoRoute skeptics to proof.
YC software page"Build the software that runs YC"; email software@ycombinator.com with a brief note about yourself and what you want to buildTalent capture for engineers attracted by the system.
YC apply pagesubmit the application online; apply as soon as readyFounder capture for builders who want the broader YC path.

Expanded Funnel

  1. Belief shift: "Same person. Different era. The difference is the tooling." Proof stack: Karpathy quote, OpenClaw example, Garry's YC/software background, contribution graphs, LOC caveats.
  2. First value: install, run /office-hours, /plan-ceo-review, /review, /qa. The first experience is not generate-code; it is product judgment, review, and browser QA.
  3. Workflow standardization: team mode, repo bootstrap, auto-update, CLAUDE.md / AGENTS.md integration.
  4. Agent ecosystem expansion: OpenClaw, Codex, Cursor, OpenCode, Factory, Slate, Kiro, Hermes, GBrain.
  5. Memory upgrade: /setup-gbrain, /sync-gbrain, MCP registration.
  6. Proof layer: gbrain-evals, LongMemEval, BrainBench, scorecards.
  7. YC capture: work at YC, apply to YC, contribute, or join the YC-aligned audience.

Message Architecture

The strongest pattern is a four-part stack:

  1. Status proof: "I am YC's CEO and I use this myself."
  2. Behavior proof: shipped products, shipped features, contribution graphs.
  3. Mechanism proof: commands, skills, browser tools, memory tools, evals.
  4. Invitation: fork it, install it, use it with your agent, work at YC.

This avoids a classic SaaS funnel. There is no pricing page, demo request, or waitlist. The conversion asset is the repo itself.

Funnel Risks