Jevtown
Audience simulator where 10,000 computed personas read your post or listing and Jev decides who sees it and how each reacts.
Overview
Jevtown is a social network where people write and 10,000 AI personas read. You post a text, listing, product or headline and the crowd reacts within seconds: most scroll past, some like, repost, block, message the seller or buy. Every reaction comes from Jev. Deterministic code computes each persona from its id (work, age, interests, money, shopping, temperament), and one request asks Jev about 60 Score questions on who the text is for, which picks the first 600 readers, plus moderation yes/no checks that keep or block the text.
Batched Choice questions then return each persona's reaction, sampled with a fixed seed. The text spreads in waves of 600, 1,500 and 3,000 only while glad reactions outweigh sorry ones by at least a tenth of the wave, so spam and scams die early. You can describe a target audience in words, see a 100 by 100 map of who reacted, break down reactions by group and ask follow-up fixed-answer questions. List sources put a check at half a cent to ten cents and 3 seconds to a minute. Ukrainian and English crowds are supported.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | How much would people with this attribute care about this text? (about 60 audience attributes) | — |
| Yes / no | Does this text violate a moderation rule? | — |
| Choice | How does this persona react? | scrolled past · stopped · liked · disliked · reposted · followed · blocked |
How it works
- User posts text; Jev scores audience attributes and runs moderation checks
- Code selects the first wave of 600 personas
- Jev answers a reaction Choice per persona; code samples with a fixed seed
- Text spreads to the next wave only if glad reactions outweigh sorry ones
- Results shown as a map, group breakdowns and follow-up answers
Strengths
- Reproducible simulations from seeded sampling
- Moderation and audience targeting in the same request
- Clear cost range per check
- Bilingual crowds
Caveats
- Personas are synthetic and limited to known attributes
- Simulated reactions are not real audience data
- Audience descriptions outside known attributes are ignored
Getting started
Use the live site at https://jevtown.ivanhabor.com (no sign-in).At a glance
- Type
- Demo
- Status
- beta
- Language
- TypeScript
- Runtime
- web app
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Audience simulation: a town of 10,000 personas computed from their id reads a post, listing, product or headline; one request asks Jev about 60 `Score` questions on who would care plus seven `Noul` moderation checks (0.5 keeps the text out of the public feed, 0.85 blocks it), batched `Choice` questions then return each persona's reaction in waves of 600, 1,500 and 3,000, and code sends the text to the next wave only while glad reactions outweigh sorry ones by at least 0.1 of the wave.
yibie/awesome-jevA town of 10,000 computed personas reads your post, listing, product or headline. Jev scores who the text is for to pick the first 600 readers and answers one `Choice` per persona for its reaction; code sends the text to the next wave only while glad readers outnumber annoyed ones by at least a tenth of the wave. Live: [jevtown.ivanhabor.com](https://jevtown.ivanhabor.com).
AnotiaWang/awesome-decision-modelsA check costs from half a cent (a text that dies in the first wave) to ten cents (one that reaches all 10,000), and takes from 3 seconds to a minute. The interface comes in Ukrainian and English, and so do the personas: a text is read by the crowd that speaks its language, 10,000 Ukrainians or 10,000 English speakers, so there is nothing to choose.
logicrw/awesome-jev-projectsTown of 10,000 computed personas where Jev scores a post, listing, product, or headline against about 60 audience attributes to plan who sees it, then answers one batched Choice question per persona for the reaction (scrolled past, stopped, liked, disliked, reposted, followed, blocked) and a follow-up question for listings and products; deterministic code computes every persona from its id, picks each wave, samples the reaction from the returned probabilities with a fixed seed, and stops the text when sorry reactions outweigh glad ones.
cobanov/awesome-jevJevtown: a social network where people write and 10,000 AI personas react.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev decide?
Who the text is for, whether it passes moderation, and each persona's reaction.
Are the personas real people?
No. They are computed deterministically from ids with fixed attributes.
What does a check cost?
From about half a cent if it dies early to about ten cents if it reaches everyone, per the listing.
Related projects
Codex plugin that runs polls on simulated reader profiles with a System One model, recording typed answers for follow-up studies.
ProjectbetaTypeScriptReal-time bot battles where you write a fighter's strategy in plain English and a System One model pilots it several times a second.
ProjectbetaTypeScriptBot that plays Red Alert 2 web skirmishes, with code handling mechanics and a decision model making stance, threat and timing calls.
ServicebetaPythonMultiplayer 3D science-fantasy version of Houston's River Oaks District with Jev-powered NPC companion, chauffeur and smart drive.
ProjectbetaTypeScriptBrowser 3D planetary rover sandbox where a TypeSafe controller chooses the next action from complete candidates, versus a baseline.
ProjectbetaTypeScriptKorean RL tycoon game training a company decision model that answers situations with choice, probabilities and confidence.
ProjectprototypePython