DTC founders and media buyers get a creative brain for Meta ads in Claude. It remembers every ad and suggests the next 10 each Monday.

<aside> 🚀

🚀 FREE STRATEGY CALL Want to grow your product through viral reach? We help teams across iGaming, mobile apps, AI products, and SaaS build creator-led organic distribution generating tens of millions of views at CPMs starting from $0.03.

<aside> 🚀

🚀 Book a free strategy call

</aside>

On the call you'll receive:

Contents

In the post Count Where it lives on this page
Things the brain remembers 4 One section each, in the post's order
Every ad, with its angle, hook, creator and format 1 Section 1, the ad log and the Ad log update prompt
What won, what tired, and why 1 Section 2, the Verdicts prompt
The words buyers use in reviews and comments 1 Section 3, the words file and the Buyer words prompt
Which angles are still untested 1 Section 4, the angles file and the Angles prompt
Inputs the brain reads 8 The eight inputs table in Start here: Every ad, Angles, Hooks, Creators, Formats, Review words, Winners, Tired ads
Monday: your next 10 ads 10 Section 5, the Monday routine, with a brief for each ad
Sample ad names for the Monday list 10 Written into the Monday routine prompt and its example
Copy ready prompts 6 Five in the library, one in the bonus toggle

Also here: The brief template. The weekly loop as a text diagram. The Pebblecove run: 60 past ads, 400 reviews and two Mondays on one fictional brand. Five imperfect inputs and who decides.

Jump to: Start here. Every ad. What won and tired. Buyer words. Untested angles. Monday. The Pebblecove run. How to use it every week. How this was made.


Start here

  1. Download the four memory files and open them in a spreadsheet.
  2. Save your ad numbers as a CSV, one row for each ad. Include lifetime spend, impressions, link clicks and purchases, plus spend and purchases for the last 14 days.
  3. Open a Claude chat and paste the Ad log update prompt from Section 1.
  4. Paste your CSV below the last line and copy the reply into the ad log.
  5. Run the Verdicts prompt on the same rows and read which ads won and tired.

Five steps. Under twenty minutes. One honest read of your own ads.

The brain is four plain files. Claude reads them and suggests. A person picks what gets made.

Input Where it lives
Every ad The ad log, one row for each ad
Angles Angle column in the ad log, plus the angles file
Hooks Hook column in the ad log
Creators Creator column in the ad log, plus who can be booked in the rules file
Formats Format and Length columns in the ad log
Review words The words file
Winners Status Winner in the ad log, with the reason
Tired ads Status Tired in the ad log, with the reason

<aside>

How to run any prompt on this page. Open a Claude chat. Paste the prompt text. Paste your inputs below the last line. Claude Code works too: save the prompt pack as a file and ask Claude to follow one prompt on your files. Without Claude Code, paste the prompt text into a Claude chat.

These prompts are not a running service. Each one runs when you ask. Claude reads, compares and drafts from the files you give it. People film, spend, message and decide. Views and clicks are not purchases. A starting point on this page is yours to change. Before using these prompts you should omit any PII (personal identifiable information) and check with your compliance team and/or your AI policy. Remove customer names from reviews and comments before you paste them.

</aside>

<aside>

Downloads. The prompt pack holds all 6 prompts, one sample input for each, and the brief template. Four memory files: the ad log, the words file, the angles file and the rules file, each with a header row and sample rows. The zip holds those five files plus the 480 row Pebblecove reviews sample, six files in all.

creative-brain-prompts.md

creative-brain-pack.zip

ad-log.csv

buyer-words.csv

angles.csv

rules.csv

</aside>


Every ad, with its angle, hook, creator and format