Jevtown

About Jevtown

Jevtown is a simulation tool that runs a post, listing, product description, or headline past 10,000 computed AI residents before it goes live. The text travels through waves of readers-starting with the 600 it should matter to most-and only spreads further if more residents were glad than annoyed by it. A weak text stops after the first wave, while a text that resonates reaches all 10,000 residents in about 14 seconds.

Review

Jevtown approaches pre-publication testing as a statistical sampling problem rather than a single-score judgment. Instead of asking one model for one verdict, it batches 10,000 persona-driven decisions and tracks who stopped, liked, reposted, or blocked. The output includes demographic and interest-based breakdowns, buyer questions for listings, and a demand curve over a product's price ladder-all without requiring a sign-in.

Key Features

• Wave-based propagation. A text first reaches 600 residents matched to its content. It advances to the next 1,500 only if the glad-to-annoyed ratio stays positive, simulating how content spreads or dies in actual feeds.
• Persona-driven reactions. Each of the 10,000 residents has a defined job, interest, age, city, and budget. The system returns what each resident did-scroll past, read, like, repost, follow, or block-along with their demographic filters.
• Buyer questions for listings. For product listings, the tool surfaces the questions buyers would ask first. In one test, a listing that required prepayment triggered scam-suspicion questions from 43% of stopped buyers, while an inspection-friendly version saw 24% asking about price negotiability.
• Custom audience targeting. The Audience tab lets users describe readers in plain words (e.g., "people who work in IT and are into startups"). Only residents matching that description read the text-238 out of 10,000 in the IT/startups example.
• Open-source and extensible. The code is MIT-licensed, and residents are defined in a single file (public/shared/personas.js). An MCP server lets agents like Claude write variants and run them through the town using compare_texts and check_text endpoints.

Pricing and Value

Jevtown is free to use. A text that dies in the first wave costs roughly half a cent, while one that reaches all 10,000 residents costs about ten cents. The site does not require sign-in or account creation. Users who want to run their own instance with a custom resident pool-such as a town of founders only-can self-host the open-source code and supply their own Jev API key.

Pros

• No sign-in or account setup required to run a test.
• Wave mechanics simulate how a text actually travels through a feed, rather than returning a single aggregate score.
• Output filters by interest, job, age, city, and budget let you isolate which segments responded and how.
• Buyer question generation shows what a listing is missing, giving concrete direction for the next rewrite.
• Open-source code and documented resident generation make self-hosted, niche-audience instances possible.

Cons

• The residents are simulated AI personas, not real people. The maker has not compared reaction patterns against live audience data, so the numbers are best used for A/B comparisons between text variants rather than absolute performance predictions.
• Jevtown does not write or suggest rewrites itself-it only answers questions about the text you supply. You must write the next version manually (though an MCP server now lets external agents handle variant generation).
• The tool is not well suited for users who need real human focus-group feedback or statistically validated market research. It is a simulation environment for comparing drafts, not a replacement for actual audience measurement.

Jevtown fits writers, marketers, and product sellers who regularly A/B test copy and want rapid, low-cost directional feedback before publishing. It's particularly useful for comparing two versions of the same text to see which one travels further and why. Users who need precise, real-world audience metrics or automated copy generation will find it less aligned with their workflow.

Open 'Jevtown' Website