OpenCMO 开源首席营销官增长专家实战AI技能

获取基于16位顶级增长专家一手实战经验、贴合你业务场景的靠谱增长策略建议。

详细介绍

OpenCMO

The open-source CMO. Growth playbooks from 16 experts who actually ran the numbers — Cursor, Notion, Linear, Deel, Canva, Lovable, Grammarly, Wispr Flow — packaged as an installable skill for your AI agent.

Covers the whole growth stack : paid ads (Google · Meta · LinkedIn) · out-of-home · SEO & programmatic SEO · launches & Product Hunt · pricing & monetization · creator & influencer programs · PLG→enterprise · attribution & tracking · growth engineering with agents

The experts

Sixteen people who ran growth at companies you know, each distilled from first-hand sources — podcast transcripts, bylined essays, their own posts. Every claim carries a link back to where they said it.

Person
Background
One line to remember them by
Find them

Rex Gelb
Paid media, Cursor
Don't buy ads before PMF; after AI took over bidding, creative/tracking/landing pages are the three levers left
in

Matt Swulinski
Wispr Flow → Viktor
Ran $1M/month solo; "creative is the targeting"; caps AI-generated creative at 5%
in · X

Elena Verna
CPO, Lovable
"60–70% of traditional growth tactics don't apply anymore"; PMF is a treadmill, not a destination
in · X

Meltem Berkowitz
CGO, Deel
0→$300M in 3 years, profitably; the SEO traffic-light framework; "no successful paid ads on a 4-second site"
in · X

Yuriy Timen
Grammarly, 9 years
The oil-wells theory of channels; admits the best content doesn't always win — and why brands shouldn't chase the shortcut anyway
in · X

Karri Saarinen
CEO, Linear
$35K total marketing spend to a $1.25B valuation; the methodology is the marketing
in · X

Lena Waters
CMO, Notion
"We used to sell to humans who researched with tools. Now we sell to tools that report to humans"
in

Ben Lang
Early Notion / Cursor
Full-length creator videos over ad reads; evergreen signup flywheels
in · X

Morgane Palomares
ex-VP Marketing, Vercel
Replaced SDRs with technically-trained Product Advocates
in · X

Holly Chen
ExponentialX (ex-Slack, Google, Loom, Miro)
Channel choice follows ARR stage; PLG and sales-led converge past $20M
in · X

Phil Carter
Elemental Growth (ex-Quizlet)
80% of trials start on day zero — the first session is the whole game
in · X

Jared Liu (宗源)
Growth engineer, YouMind
Aggregation-site flywheels; "growth while you sleep" via agent automation
X

Frank
AI product, 5M users
The four-pillar system: SEM validates, SEO builds the asset, GEO rides along, social amplifies the brand
in

Austin Lau
Growth, Anthropic
Cut ad production from 30 minutes to 30 seconds by building his own tools, having never opened a terminal before
in · X

Grant Lee
CEO, Gamma
Won Product Hunt, still called it no-PMF; rebuilt the first 30 seconds and went to 20K signups/day with zero marketing
in · X

Chris Pedregal
CEO, Granola
A year in closed beta, ~150 hand-onboarded users, no built-in growth loops — it spread through VC circles anyway
in · X

Why this exists

Ask any AI assistant "should we start running ads?" and you get a balanced, reasonable, useless answer.

Ask an agent with OpenCMO installed, and you get four experts who disagree:

Expert
Position
Their precondition
Track record

Matt Swulinski (Wispr Flow)
Start immediately
Conversion tracking works + 50 conversions banked
$2M→$50M ARR in 12 months

Rex Gelb (Cursor)
Not before PMF
Week-4 retention >20%
Runs paid media at Cursor

Elena Verna (Lovable)
Year one is a death trap
Payback period under 3 months
$400M ARR in 18 months

Meltem Berkowitz (Deel)
Foundations first
Site loads <4s, findable, content to catch demand
0→$300M in 3 years, profitably

Paid ads are just the demo — every area listed at the top gets the same treatment.

Real growth knowledge looks like this: conditional, contextual, and contradictory. Average the disagreements away and you get soup.

Three rules govern everything in here:

Named people, direct quotes. No anonymous "experts say." Every claim has a person, a date, and a link.

Conflicts stay visible. When experts disagree, we build a comparison table instead of picking a winner. The disagreement is the information.

Evidence gets graded. First-hand quote, media coverage, company self-report, or third-party estimate — every number is labeled. We also debunk: the file on Jasper exists mostly to show why the "Google penalized their AI content" story doesn't survive contact with the evidence.

Install

curl -fsSL https://raw.githubusercontent.com/About-Intelligence/OpenCMO/main/install.sh | bash

Add -s — –project to scope it to the current repo instead of your home directory. Re-run it any time to update. Prefer to see what you're running? Read the script or just git clone this repo into ~/.claude/skills/ .

Then ask it real questions — "should we turn on Google Ads?", "how many Meta creatives per month at $100k?", "is a billboard worth it, and how would we attribute it?" — and it answers by citing specific people and the conditions they set, not by averaging advice into mush.

What's inside

open-cmo/
├── SKILL.md # entry point: expert index + scenario navigation
└── references/
├── experts/ 16 experts # theses, verbatim quotes, contrarian takes, conflicts
├── cases/ 4 teardowns (10+ companies) # PSEO canon, Canva deep-dive, OpenRouter, Jasper debunk
├── channels/ 4 channels # Google, Meta, LinkedIn, out-of-home
├── topics/ 9 topics # timing, creative, attribution, SEO, pricing, launches…
└── guides/ 4 guides # growth 101, PLG→enterprise, meta-patterns, why experts disagree

The meta-patterns

The most useful file in the repo is deep-insights.md : eleven patterns you can only see with all the experts side by side, ending in decision tables — what works by company type, by stage, and the single variable that settles each famous expert fight. A sample of the patterns:

The Amplifier Law. Every paid channel amplifies; none of them generate. The four "when to start ads" positions above are one answer wearing four hats.

The Moat Migration Law. Content volume stopped being a moat. Proprietary data replaced it. Extractability — whether an AI can quote you — is replacing that.

The Asymmetric Loss Function Law. Your strategy space is set by your downside risk, not by what's technically possible. This is why brands can't copy spammer tactics, in SEO, on Product Hunt, or with bought GitHub stars.

Attribution Humility. Direct traffic is lagged paid spend. Self-reported attribution flatters. The goal of measurement isn't precision, it's preventing self-deception.

How this compares

Growth books & courses
Asking a generic AI
OpenCMO

Freshness
Dated at print
Training cutoff
Updated continuously, everything timestamped

Sources
Sometimes
Untraceable
Every claim linked

Disagreements
One author's view
Averaged into mush
Preserved, with comparison tables

How you use it
You read it
You chat with it
Your agent cites it inside your actual decisions

Contributing

Found a good first-hand interview or teardown? See CONTRIBUTING.md for the evidence standards. The biggest gaps right now: enterprise B2B (CMO-level experts), consumer products, and non-US markets.

Roadmap

[ ] More experts: Bill Macaitis (ex-Slack CMO), Kipp Bodnar (HubSpot), consumer growth leads

[ ] Retention & lifecycle topic file

[ ] MCP server so agents can query the library by expert × scenario

[ ] Chinese translation

License

MIT for the repo structure and original analysis — see LICENSE .

Quotes belong to their original authors and publications. We quote briefly, for research and education, and link every source in full. If you're a rights holder and want something changed or removed, open an issue and we'll handle it quickly. Third-party traffic figures (Ahrefs, Semrush, SimilarWeb) are labeled as estimates and are not official company data.

Maintained by Soku , the agentic workspace for marketing teams. We built OpenCMO because our own agents needed it.