Subscribe to the
Newsletter
Cody Schneider does early-stage growth. He recently tweeted what he'd done in a single day—40 Facebook ads, 100 landing pages, three guest blog posts, four podcast bookings, 25 scheduled tweets, and two pieces of software as lead magnets—and it got 315,000 views and a literal Polymarket bet on whether it was real. Five founders DM'd him in a week looking to hire someone with that skill set.
He's not a software engineer. He's a growth person who describes what he wants clearly enough that Claude Code builds it. The domain knowledge took years. The building part now takes minutes.
Below is his exact system—step by step, with copy-paste prompts you can drop straight into Claude Code. Let's get into it.
Use Cases
If you run paid ads and your creative testing cycle is measured in weeks instead of hours, this changes that. The head of growth at a Series A who's been begging for a designer to make ad variations — you don't need the designer anymore. You need 30 minutes and a template.
If you do cold outbound and you're manually finding leads, enriching them, validating emails, and loading them into campaigns one by one — this collapses that into a single pipeline triggered by a Slack command. The SDR who scrapes LinkedIn engagers from a competitor's CEO posts and has those leads in a cold email campaign before lunch? That's the new bar.
If you're a solo founder or a one-person marketing team and you've been telling yourself you can't compete with companies that have 20-person growth teams — that excuse just expired. The constraint now isn't headcount. It's knowing what to build and how to measure it.
I've done probably more work in the last two weeks than I had done since October. I can't even put a number on it — it's infinite. The problem now is no longer what can be done. It's knowing what should be done.
— Cody Schneider, Founder of Graph
01. Get Your AI Coding Tool Running
You don't need to know how to code. You need a tool that codes for you when you talk to it.
An AI coding tool is software that builds things for you when you describe what you want in plain English. It writes the code, runs it, fixes its own errors, and gives you the finished product. You talk, it builds. That's it.
You have a few options. Pick one:
Tool: Claude Code Desktop
Best for: Cody's setup. Most powerful for this workflow.
Cost: Pro plan ($20/mo)
Link: https://claude.ai/download
Tool: Antigravity
Best for: Great UI, beginner-friendly, runs Claude under the hood
Cost: Free tier available
Link: https://antigravity.dev
Tool: OpenAI Codex
Best for: If you're already in the OpenAI ecosystem
Cost: ChatGPT Plus ($20/mo)
Link: https://openai.com/codex
Download and install whichever one you pick. For the rest of this playbook, we'll show prompts for Claude Code since that's what Cody uses, but the same prompts work in any of these tools — you're just describing outcomes in English.
Create Your Project Folder
Once your tool is open, paste this:
That's your home base. Every workflow you build from this playbook will live in that folder. The .env file is like a keychain — it holds the passwords (called API keys) that let your AI tool talk to other software on your behalf.
What's an API key?
Quick definition so this doesn't trip you up later. An API key is a password that lets one piece of software talk to another. When you sign up for a tool like Apollo or Instantly, they give you a unique string of characters. You paste that string into your .env file. Now Claude Code can interact with that tool for you — pulling data, sending emails, uploading ads — without you ever logging into that tool's dashboard.
Every time this playbook says "you'll need the API key for X," it just means: go to X's website, sign up, find the API key in your account settings (usually under "Developer" or "Integrations"), copy it, and paste it into your .env file.
If you can't find it, paste this into Claude Code:
It'll give you step-by-step directions for whatever tool you're asking about.
Checkpoint: You have an AI coding tool installed, a "growth-agents" folder in your Documents, and a .env file ready to fill in as you go.
02. Mine Your Customer's Language
The best ad copy comes from Reddit threads, not brainstorming sessions.
Before you create anything, you need source material written in the words your customers actually use. Not your marketing team's words. Theirs.
Paste this into Claude Code (swap out the brackets with your actual product info):
What comes back is gold — real humans describing real frustrations in language that'll hit harder than anything a copywriter invents from scratch. Save that output. These pain points and outcomes become the raw inputs for every asset you build: ad headlines, landing page H1s, email subject lines, blog angles.
Checkpoint: You have a list of 15-20 pain points and desired outcomes written in your ICP's own language.
03. Build a Bulk Ad Generator
One template. Infinite variations. 20 minutes to build.
Here's the epiphany behind Cody's system: all design is just code. Figma, Canva, whatever — under the hood it's code rendering visuals. So skip the design tool entirely and have Claude Code build the ads as actual code that exports as downloadable images.
Find an ad format you like — screenshot a competitor's ad, grab a template from the Meta Ad Library, whatever catches your eye. Then paste this into Claude Code:
Go back and forth on the design until the template looks right. Then feed it your research from Step 02:
The whole build takes about 20 minutes. The first version will look rough. That's fine — you're testing messaging, not shipping final creative. Once you find a winner, you can always bring in a human to polish the design.
Checkpoint: You have a working bulk ad generator and a zip file of 40+ ad variations ready to upload.
04. Test Cheap, Kill Fast
$100 and three days tells you more than a month of gut instinct.
Upload all your ad variations into a single click campaign on Meta. Not a conversion campaign — a click campaign. You're optimizing for the cheapest cost per click, which is a leading indicator of what'll convert once you move it to a proper campaign.
Run $100 over three days. Look at CPCs. Cody's seen winners come in at $0.30 while losers sit at $0.85 from the same batch. The spread tells you everything about which messaging resonates.
Take the winners (cheapest CPC), spin them into their own conversion campaigns with dedicated budgets. Kill the rest. Then take that winning messaging and remix it — have Claude Code generate UGC-style video scripts with the same angle, build landing pages around that positioning, write email sequences using those pain points. The concept is what you're testing. The format can evolve.
This is the same principle behind how Morgan Housel's Psychology of Money started as a blog post and Malcolm Gladwell's books started as magazine articles. Test the idea as cheaply as possible. Put resources behind what works.
Checkpoint: You've identified your top-performing ad angles by CPC and have a plan to remix winning messaging across formats.
05. Generate Landing Pages at Scale
Every winning ad deserves a landing page that matches it.
When the ad headline and the landing page H1 say the same thing, conversion rates go up. This isn't new — it's why landing page builders are a billion-dollar industry. What's new is you can generate hundreds of these in minutes.
Cody uses Strapi (strapi.io — open-source CMS with an API) for this. You could do it with WordPress, Webflow, or anything that lets you create pages programmatically. The idea: build one template, then have Claude Code stamp out variations.
Paste this into Claude Code:
Cody demoed this live — fed Claude Code a keyword list for "AI dashboard generator" and it created seven landing pages in about 30 seconds.
Once the pages are live, get them indexed:
Checkpoint: You have dedicated landing pages deployed for your top-performing ad angles or target keywords.
06. Measure Everything or Loose
You can create infinitely now. The only advantage left is knowing what's working.
At this point you've got ads running, landing pages deployed, and messaging being tested across formats. This is where most people doing "AI marketing" fall apart. They generate a thousand assets and have no system for understanding which ones matter. All the AI slop tooling in the world doesn't help if you can't analyze the output at scale.
The data problem is real. Facebook Ads alone generates 25 million rows of data per month for a single customer spending $180K. You can't just hit the API through an MCP and expect to make sense of that.
Cody's approach: have Claude Code build you dashboards for each campaign. He demoed this live — pasted in an ad set ID and had a full dashboard in seconds.
The KPI dashboard serves two purposes: it tells you what to double down on, and it gives you proof to show internally. "I decreased CPA by 20% using this ad velocity testing process" hits different when there's a live dashboard backing it up.
Checkpoint: You have a measurement system tracking performance across every asset you've deployed, and you can articulate which angles are winning and why.
FAQS
Do I need to know how to code to do any of this?
No. Cody isn't a software engineer. You describe the outcome you want, Claude Code builds the infrastructure. If it asks for something you don't understand — like an API key or a specific permission — you ask it to explain how to get that thing, and it walks you through every click. The skill that matters is clear thinking, not programming.
Won't this all just be AI slop that nobody engages with?
Some of the best-performing Meta ads right now are AI-generated UGC avatars. The concept is what gets tested — the format can always be upgraded. Find a winning angle with AI-generated creative, then reshoot it with a human if you want. But most of the time, the AI version performs well enough that the reshoot never happens.
How do I convince my company to let me test like this?
Start with burner accounts or a small test budget. Run the ads, find a winner, show the CPC data. Nobody argues with a 20% CPA decrease. The dashboard from Step 07 is your proof.
The Takeover
The role that used to require a 20-person team now requires one person who understands the system. Not one person who works 20x harder — one person running 20 agents in parallel while they jockey between desktops.
The mindset shift: you're not hiring tools to help you do marketing. You're building a personal software stack that you bring to every company you work at. When you negotiate salary, you're not just selling your experience — you're selling the system you've built that makes you a one-person growth team.
The uncomfortable part nobody wants to say out loud: the people who aren't adopting this aren't falling behind slowly. They're falling behind at the rate this technology improves, which right now is roughly every two weeks. Everything Cody showed in this playbook didn't exist in his workflow a month ago.
HIRE US AS YOUR CHIEF AI OFFICER
We have the muscle — and the reps. Want us to run this exact playbook inside your org? Talk to us today.
Link: https://www.tenex.co/get-started
.jpg)
.jpg)
.png)
.png)
.avif)
-min-min.png)
.png)
