Mirrortown1
Drop in the thing
Paste a post, a headline or an email, upload an image or a video, or describe an idea. Choose one of 17 content types, or start from a template.
Social postAdVideoLanding pagePrice ladder+12 more
Lifetime deal on AppSumoWindows · macOS
Show it to a town before you show it to the world.
Test any post, ad, landing page or price on up to 10,000 synthetic people before you publish. In minutes, Mirrortown tells you who stops, who likes it, who’s annoyed, why, and what they’d do next. By segment, by region and on a map.
One-time paymentBring your own Jev key~US$0.10 per 10,000-person run
Sample run · 2,000 peopleTest 1 / 3
Instagram ad“2×1 brunch at Café Lumbre. Saturdays only.”
Reacted 0 / 2,000Hover or tap a dot
Liked it0%
Stopped0%
Annoyed0%
Scrolled past0%
VerdictReading the room…
10,000synthetic people per run, each with an age, a place, an income, interests and habits.
~$0.10in Jev usage for a full 10,000-person run, paid straight to your provider.
17content types, from a one-line tagline to a price ladder.
LocalProjects and results live in SQLite on your own computer.
Plate I How it works
Three steps from “I think” to “they said.”
No panel to recruit, no survey to field. Mirrortown builds the town on your computer and calls Jev only for the reactions.
2
Choose who sees it
Use a ready-made population or your own. Run a quick sample, the whole town, or one segment by age, gender, region, income, interests or platform.
PopulationEnglish-speaking consumers
ModeFiltered segment
WhoWomen 25–40 on Instagram
3
Read the room
A one-sentence verdict first. Then the detail: reactions, a heat map by region, the segments that stopped or bounced, their objections and their own words.
Clear offer, weak photo. Under-30s would share it; many over 55 suspect a catch.
Plate II What you can test
Seventeen kinds of content. One town to try them on.
Each type asks the follow-up questions that fit it. A video gets a scene-by-scene retention curve. A price ladder gets a demand curve.
- 01Social postWho stops scrolling, and who would share it.
- 02AdAttention, clarity and purchase intent.
- 03Video Retention curveScene by scene: where people stay and where they leave.
- 04HeadlineWhich version earns the click.
- 05Landing pageDoes the promise land on the first screen?
- 06EmailWould they open it, read it and act on it?
- 07Product & price ladder Demand curveHow many would buy at each price.
- 08Idea or conceptKill it or keep it before you build it.
- 09Marketplace listingTitle, photos and price against the scroll.
- 10Public announcementIs it understood, believed, welcomed?
- 11Public-policy proposalSupport and opposition, region by region.
- 12Messaging-app broadcastRead, ignored or forwarded to the family group.
- 13Speech or presentationWhat sticks, what bores, what convinces.
- 14Tagline or sloganMemorable, clear, or forgettable?
- 15Radio or podcast spotHeard, remembered, acted on.
- 16FlyerThe ten-second street test.
- 17Custom survey questionAsk the town anything and get a distribution back.
Plate III What you get back
A verdict in one sentence. The why behind it.
Every run ends in plain language, then gives you everything you need to act on it, down to the person.
Example result · Landing page hero · 10,000 people
Strong with small-business owners, confusing for everyone else: 38% stopped, but only one in four could say what the product does.
5.2Clarity / 10
6.8Credibility / 10
CuriousMain emotion
11%Purchase intent
+14NPS
6%Would share
Heat map by region · liked it
FewerMore
- Business owners 35–54+21 pts stopped
- University district+18 pts liked
- Retirees 65++12 pts annoyed
If it really chases the clients who pay late, I’d try it this week.Marta, 47 · owns a print shop · Market
Chase them how? By email? I need to see it before I believe it.Omar, 38 · freelancer · Old Town · top objection
Sounds like it would nag my customers. Not the image I want.Hugo, 61 · retired accountant · Pinewood
Interview anyone
Turn any group into a persona panel, then run 1:1 interviews or a focus-group roundtable.
Follow-up objections
What holds back the people who almost said yes, ranked.
Reports
PDF, Markdown or JSON. With 3 codes, your brand on the PDF.
Claude Code & MCP
Export a run to Claude Code, or let AI agents run tests through the built-in MCP server.
Illustrative example. Your numbers come from your own runs.
Plate IV Templates
Start from a town that already exists.
Ready-made populations and one-click starter tests, so your first result is a few minutes away.
Populations
Spanish · real map
Colombia by department
A population spread across Colombia’s departments, drawn on a real map, with results department by department.
Spanish
Latin America consumers
Spanish-speaking consumers across the region.
English
English-speaking consumers
A broad English-speaking consumer population.
Persona panels
Ready-made panels
B2B SaaS buyersGen-Z social shoppersMiddle-class familiesLatAm SaaS foundersRestaurants & hotelsConsumers 25–40
Your own
Build your own population
Import people from a CSV, or start from a GeoJSON map plus statistics for each region. AI can draft the statistics for you to review.
Starter tests
SaaS landing pageLanding page
Restaurant promoSocial post
Course price ladderPrice ladder
Product launch emailEmail
Real-estate listingListing
App adAd
Funnel templates
Landing page (AIDA) 01
Signup 02
Onboarding 03
Checkout 04
Login 05
Plate V Testing modes
Seven ways to run the town.
From a quick gut check to the whole population. A/B tests show every version to the same people, so the comparison is fair.
Quick sample
A slice of the town for a fast, cheap gut check.
Whole population
Everyone, up to 10,000 people, for the full picture.
Filtered segment
Only who matters: age, gender, region, income, interests, platforms and more.
Organic feed waves
Content spreads the way it would in a feed, wave after wave.
Word of mouth
Follow it through each person’s contact network.
A/B comparison
Two or more versions, the same people, a clear winner.
Price ladder
Several prices in one run, read as a demand curve.
Every mode ends the same way: a verdict, the map, the segments and the voices. Then interview whoever surprised you.
Plate VI Bring your own key
You bring the key. A full run costs about ten cents.
Mirrortown doesn’t resell AI. Reactions come from Jev through your own TypeSafe or OpenRouter account, and you pay the provider directly.
01
Your computer
Mirrortown, your populations, your content and every result, stored in SQLite on your machine.
02
Jev, with your key
A hosted decision model, not a chatbot. It answers with probabilities, which is why 10,000 people cost cents. Use a key from TypeSafe or OpenRouter.
03
Optional: an LLM key
OpenAI, Anthropic, Gemini or OpenRouter, only for 1:1 interviews, focus-group roundtables and written reports.
Jev cost estimator
$3.00in Jev usage
$0.10 per run · paid to your provider, not to us
Estimate based on about US$0.10 per 10,000 people. Actual cost depends on your provider’s current Jev price and the length of your content. Interviews and reports are billed by your LLM provider.
Plate VII Pricing
Pay once. Keep it.
Mirrortown is on AppSumo as a lifetime deal. Buy one code, or stack up to three for more seats and more room.
AppSumo lifetime dealStack up to 3 codes
1 code
For a solo marketer, founder or creator.
$49$299
One-time payment
- 1 user
- 3 workspaces
- 3 custom populations
- All 17 content types and all templates
- Populations up to 10,000 people
- A/B tests, interviews & reports
2 codes
For a small team or a growing agency.
$98$598
One-time payment
- Everything in 1 code
- 3 users
- 15 workspaces
- 15 custom populations
- Claude Code export
- MCP server for AI agents
3 codes
For agencies and teams who report to clients.
$147$897
One-time payment
- Everything in 2 codes
- 10 users
- Unlimited workspaces
- Unlimited custom populations
- White-label reports: your brand on the PDFs
Every tier
Lifetime accessAll future 1.x updatesBring your own Jev keyWindows & macOS60-day refund through AppSumo
Already bought? Redeem your codes. Bought more later? Redeem all of them together for an upgraded license.
A simulation, not a census.
Mirrortown’s people are synthetic and their reactions come from a model, not from real humans. Treat results as a directional signal: a fast, cheap way to kill weak ideas and decide what is worth testing with real people.
It doesn’t replace surveys, interviews or a real launch. It makes sure that what you take to them is your best shot.
Plate VIII FAQ
Questions, answered plainly.
Something missing? Write to brokenwatchceo@gmail.com. The founder answers.
Is this real research?
No, and we won’t pretend it is. The people are synthetic and their reactions come from Jev, a decision model. Results are directional: good for spotting what confuses, annoys or bores people, comparing versions and choosing what to test for real. They are not measurements of a real population and they don’t replace surveys, interviews or a real launch.
Which keys do I need?
One is required: a Jev key from TypeSafe or OpenRouter. Paste it in Settings → City.
Optional: an OpenAI, Anthropic, Gemini or OpenRouter key for 1:1 interviews, focus-group roundtables and written reports. You pay those providers directly; Mirrortown never marks up AI usage.
How much does a run cost?
About US$0.10 in Jev usage for a full 10,000-person run, and less for samples and segments. Your provider charges it to your own account. The lifetime license itself has no usage fees.
Where is my data?
On your computer. Projects, populations and results are stored locally in SQLite. The content you test goes only to the AI providers you set up, with your own keys. This website only sees your email and your AppSumo codes when you redeem them, to issue your license. Details in the privacy note.
Mac and Windows?
Both: Windows 10 and 11, and macOS on Apple Silicon or Intel. The macOS build isn’t signed with an Apple certificate yet, so the first time you open it, right-click the app and choose Open. Downloads.
How do I stack codes?
Buy up to three codes on AppSumo. On the redeem page, enter your email and all your codes together; you get one license for the matching tier. If you buy more codes later, come back, enter all of them (old and new) and paste the new license into the app.
What is Jev, exactly?
A hosted decision model built for choices, not conversation. Instead of writing an essay, it returns probabilities: will this person stop, like it, be annoyed, buy? That is what makes it fast and cheap enough to ask 10,000 people.
Can I build my own population?
Yes. Import people from a CSV, or start from a GeoJSON map plus statistics for each region; AI can draft those statistics for you to review. 1 code includes 3 custom populations, 2 codes include 15, and 3 codes are unlimited.
What happens after I buy?
AppSumo gives you your code or codes. Redeem them here, download the app, paste your license and add your Jev key. Refunds within AppSumo’s 60-day window are handled by AppSumo.
Does it work offline?
The app and your data are local, but runs need an internet connection to reach Jev and any other AI provider you use.