Give your AI agent a marketing job it can finish.
Each job is one file: find the questions buyers ask AI, rewrite the pages that should answer them, then do it again next week. Drop the file into Claude Code and it runs, checks its own work, and stops when the job is done.
Works with Claude Code, Cursor, and Codex · Every run publishes a receipt · Free to copy, MIT
VibeMarketers is an open marketing lab. We run these loops on this site first, publish what each one found, then ship the file so you can run it on yours. It is free and there is nothing to sign up for. It is an open lab by SuperMarketers, who build the same thing inside a company when you would rather not do it yourself.
The Lab Report. Every Monday: what ran on this site, what it found, and one loop to copy.
Confirmed opt-in, unsubscribe in one click. No other email, ever.
The department loop, as it ships. Three prompts and a stopping condition.
What you can hand an agent
- 01 Find where AI never mentions you AI visibility
- 02 Make every page answer the question it ranks for AI visibility
- 03 Turn each product release into a post Product marketing
- 04 Read what your market said this week Demand
- 05 Run all of it again next week, unattended Ops
How you use it
three steps, then it runs.
- 01
Download the file
One markdown file per loop: the goal prompt, the verifier prompt, the stopping condition. Open it and read it before you run anything.
- 02
Point it at your site
Your agent reads your pages, not ours. Nothing is uploaded anywhere and no account is created. The loop works on the files you already have.
- 03
Read what it did
Every run ends with a receipt: what it changed, what it cost, and what it could not prove. You approve the change or you throw it away.
Free to copy, MIT. See what a loop file contains → Or start with three agent skills → New to the vocabulary? the glossary → Want to write your own? the guides →
The lab
where this sits.
VibeMarketers.
The open lab. You run it yourself.
- Every loop published here has run on this site
- Every run publishes what it found, including the misses
- Free, MIT, no account, nothing to log into
- Never diagnoses your company
SuperMarketers
The implementation. Someone builds it with you.
- The same loop model, built inside your own stack
- Your keys, your repo, your team runs it after
- Starts by capturing what your team already knows
- Diagnoses your company, which is the part a lab cannot do
VibeMarketers is an open lab by SuperMarketers.
How the lab works, every gate, and every confounder: the Method page. What it costs and what it touches: the questions people ask first. Run by Gen Furukawa.
The catalog
copy the loop, not the advice.
A loop is a marketing job written as a file: what to do, how to check it was done right, and when to stop. You copy it into your own agent and run it on your own site. No account, no platform, nothing to log into.
- Your company in one file your agents read. Interview loop, gated, 20 minutes. auto+gate
- Every page answers the question it ranks for. Rewrites, then verifies. auto+gate
- Every question your buyers ask AI, mapped. Runs weekly, unattended. auto
- What your market said this week, in 200 words. Monday mornings, unattended. auto
- Every product update becomes a post your buyers can read. Gated before publish. auto+gate
- Where your buyers already comment, with a draft reply. Drafts only, never sends. assisted
- No page goes stale. Checks every page monthly, rewrites what slipped. auto+gate
- Every content gap becomes a brief. Reads the topic map, writes what is missing. auto+gate
- No page is an orphan. Maps the link graph, reports what nothing points at. auto
- Your llms.txt matches what the site ships. Checks it, reports what is missing. auto
- Every page carries valid structured data. Checks the build, reports what is missing. auto
- An email sequence that only says what you can prove. Drafts only, never sends. assisted
The whole library, filterable by stage and autonomy →
Agents onboard from vibemarketers.ai/llms.txt.
Proof
we run these on ourselves first.
No loop reaches the catalog until it has run on this site and passed its own check. Every one of those runs is published the day it happens, with what it decided, what it cost, and a link to the raw log. You can read how a loop behaves before you point it at your own site.
- engagement-mining Ran 4 queries against public discussion, kept the threads with real comment activity, and drafted one reply each.
- internal-links Ran against the three new guides and three new term pages to check the additions had not created orphans.
- market-intel Re-read the two watchlist sources and compared each against what was recorded on 2026-09-18.
- internal-links Ran against the new templates section to check the additions had not created orphans.
- llms-txt-build Compared the published llms.txt against the real page list, then added the 3 key pages it was missing and reran.
Control
Loop brain-build. One entry per run, append-only. Open the loop page →
- Baseline
- No Brain existed. Brain/ held one file, voice.md, written by hand during Goal 1. Every other loop started from an empty context and guessed at positioning.
- Hypothesis
- A Brain built by reading both properties in full lets downstream loops cite positioning, ICP and proven claims instead of inventing them.
- Intervention
- Read supermarketers.ai and vibemarketers.ai in full on 2026-09-21, plus the repo's own content. Wrote positioning.md, icp.md, buyer-questions.md, claims-and-proof.md, competitive-intel.md and content-history.md, and extended voice.md with voice evidence quoted from Gen's own writing rather than invented.
- Result
- 7 of 7 Brain files exist. 12 provable claims recorded, each with a source URL and a date, and 5 claims explicitly listed as unprovable until a named condition is met. The verifier failed the first pass: 3 claims cited a repo path rather than a resolvable URL. Fixed, then PASS.
- Commercial signal
- none
- Attribution limits
- This loop produces an input, not a marketing outcome. No citation or traffic movement is attributable to it. The Brain is built from two websites and the repo, not from sales calls or a CRM, so the buyer-questions file is inferred from published copy rather than observed in conversation.
- Learning
- The claims file is where the loop earns its keep. Separating what can be proven from what cannot forced 5 claims off the site that would otherwise have read as fact, including every citation count.
- Next intervention
- Rebuild the Brain from sales calls and a CRM export rather than published copy, so buyer-questions reflects what buyers actually ask rather than what the site answers.
Sources: run log, lab/runs/2026-09-21-brain-build.json, collected , window: single run ; supermarketers.ai, read in full, collected , window: homepage as published on the date read .
The Lab Report
what ran, what moved.
Weekly: what ran, what moved, what the loops decided next, and one loop to copy. Confirmed opt-in, and every number in it links to the run that produced it. Read a past issue.
Every Monday. One issue, built from the week's receipts.
Nothing moved? The issue says so and lists what ran.
Subscribe at the top of the page
Confirmed opt-in. Unsubscribe in one click.