Jevtown

Jevtown is a social network where people write and 10,000 AI personas read. You post a text, a listing, a product or a headline, and within seconds the town reacts: most scroll past, some like, repost, block, write to the seller or buy. The English-speaking town is described as having 10,002 residents, and the site also shows a saved example labelled "the Ukrainian residents", so posts are read by computed residents rather than by real people. Nothing about the product requires an account — the site states there is no sign-in — so a writer can move from typing a post to seeing reactions without creating a profile, connecting an account or publishing anything anywhere. The product positions itself as a way to find out how an audience responds to a piece of writing before that writing is released. Writing for an audience is usually a blind exercise. A headline, a second-hand listing or a product description goes live, and only afterwards does the writer learn whether people stopped, scrolled past, doubted it or bought — by which point the first impression has already been spent. Jevtown's stated purpose is to move that feedback loop in front of publication. The Product Hunt description puts the cost of finding out in concrete terms: a weak text dies for half a cent, so a failed test is cheap and private, while a text that earns approval reaches everyone in 14 seconds. For sellers in particular, the difference between a listing that reads as trustworthy and one that reads as a scam is often a matter of wording, and the site demonstrates that directly by showing one iPhone listing written two ways. The town is not a flat audience. A post starts with a small cohort — the description says the 600 residents it should matter to see it first. Only if more readers were glad than annoyed does it reach the next 1,500, and from there it can go on to the full town. That staged distribution is the core mechanic: the feed acts as a filter that must be earned, so a text that fails to please its first cohort stops there, while a text that lands keeps travelling. The site's own feed of sample runs shows the outcomes of that process as wave sizes — posts display 600, 2,100, 5,100, 10,001 and 10,002 people reached, with the largest waves attached to posts that collected far more glad reactions than sorry ones. The Product Hunt description summarises the fastest outcome: a good text reaches everyone in 14 seconds. Every run returns a breakdown that goes well beyond a like count. Jevtown reports who stopped, liked, reposted or blocked a post, sorted by interest, job, age, city and budget. Individual residents are shown with their first name and initial, age, job and city — for example a repost attributed to "Zoe, 20 · student, New York" or a purchase attributed to "Iryna, 37 · copywriter, Dnipro" — next to the action they took. The reaction vocabulary used across the site includes scrolled past, stopped, glad, reposted, sorry, wrote to the seller, smelled a scam, bought it and clicked, with headlines reported using stopped and clicked rather than glad and sorry. Because the numbers can be watched as they develop, the site offers a Replay control, and the public feed offers Latest and Travelled furthest views alongside a running total shown as 148 texts seen 260,005 times. For listings and products the feedback is qualitative as well as numeric. Jevtown surfaces the questions buyers would ask a listing first, and it separates out the residents who took commercial action. In the demonstration listing, 2,100 personas saw the post, 142 wrote to the seller and 6 smelled a scam, and the listing went into a second wave and reached 2,100 personas. A writer also chooses the visibility of a submission — to the public feed or by link only — and a post's text is capped at 2,000 characters, which keeps submissions the length of a real social post, a listing description or a headline. The same sample record shows how one iPhone listing was written two ways, the first version offering details and payment on inspection and the second demanding advance payment only, which makes the trust cost of wording visible in the reactions rather than in hindsight. Products can additionally be tested against price. The price ladder lets a writer enter several prices — the interface shows a set in ₴, $, € and £, allows naming a few and accepts another price — and everyone who stops at the product is asked for the highest of those prices they would pay. The result is a demand curve: how many buyers each price gets and which one earns the most. This is a distinct form of testing from the copy question above, because it tests the offer rather than the sentence, and it gives a product writer a way to compare price points using reactions from the same residents who read the post. The overall approach is simulation rather than publishing. Instead of putting a text in front of a real audience and waiting, Jevtown computes a fixed population of AI residents with their own interests, jobs, ages, cities and budgets, lets them read in waves, and reports their behaviour as a feed would. That design produces the two things the product leads with: a fast, cheap answer to whether a piece of writing works, and a demographic map of who it works for. The English-speaking town is given as 10,002 residents, a saved example is labelled "the Ukrainian residents", and persona cities shown on the site include Lviv, Dnipro, Zhytomyr, New York and Boston, with prices accepted in hryvnia, dollars, euros and pounds. Whether the submission is a text, a listing, a product or a headline, the reading, the wave logic and the reporting stay the same. The practical benefit is that a writer can see the shape of a reaction before it costs anything real. The Product Hunt description frames the value in terms of price and speed: a weak text dies for half a cent, and a good one reaches everyone in 14 seconds. Alongside that, the product answers questions a writer would otherwise have to guess at — who stopped rather than scrolled past, whether buyers would write to the seller, which residents were glad and which were sorry, which questions a listing invites first, and how demand shifts as a product's price changes. Because no sign-in is required, there is no setup cost to getting that answer, and because posts can be kept to a link, the test can stay private. Use cases are illustrated directly by the site's own feed. A seller testing a second-hand iPhone listing can compare a "details, pay on inspection" version against an "advance payment only" version and read how many personas wrote to the seller and how many smelled a scam. A marketer can post a headline and see how many residents stopped and how many clicked. A builder can post a product with a price ladder and see how many buyers each price attracts and which price earns the most. And any writer can post a plain text — the feed shows runs ranging from one-line posts to opinions, announcements and memes — and watch which wave it reaches, from 600 residents up to the full town. The product is aimed at people who write to be read: sellers and marketplace listers, marketers testing headlines and copy, builders describing products and price points, and writers posting text to a feed. Product Hunt lists Jevtown under Writing, Marketing and Artificial Intelligence, and the site's feed mixes English and Ukrainian content, with the English-speaking town singled out at 10,002 residents. The product's own pitch repeats the same core mechanics — computed residents, staged waves, reaction breakdowns, the questions buyers would ask first, and a demand curve over a price ladder — rather than any additional modules. Jevtown's core promise is simple to state and repeated throughout the site: publish a text, a listing, a product or a headline, and find out who in a town of 10,000 computed residents stopped, was glad, reposted, was sorry, wrote to the seller or bought, in seconds and without signing in. A weak text dies for half a cent; a good one reaches everyone in 14 seconds.