Ask 100 AI personas
Ask one question; 100 synthetic personas answer at once, decided live by Jev. Watch the votes stream in and the count climb — a hundred real decisions in a couple of seconds, for under a cent. No key, just ask.
100 synthetic personas — made-up people (name, age, job, city, habits are real inputs to each decision), not real survey respondents.
The same demo is one Jev call
- Build 100 synthetic persona states (age, job, budget…)
- Ask every persona the *same* question
- Fire 100 Jev
noulcalls in parallel - Aggregate the yes/no votes
const results = await Promise.all(items.map((item) => // items = 100 persona states
fetch("https://jevtypesafeai.com/api/v1/decide", {
method: "POST",
headers: { Authorization: `Bearer ${process.env.JEV_API_KEY}` },
body: JSON.stringify({
state: item,
questions: {
vote: { type: "noul",
instructions: "Would this persona buy it?" },
},
}),
}).then((r) => r.json())
));
const yes = results.filter((r) => r.answers.vote.noul >= 0.5).length;See a real decision YES · 78% · 262ms
Persona: Maya, 29, product designer, SF, $$ budget, buys on aesthetics + reviews. Question: Would she buy a $99 AI note-taking app?
How it works
Type a question and Jev runs it past 100 synthetic personas at once. Each persona is a made-up person — a mix of age, role and outlook — and each answer is a real, independent Jev decision: a typed choice of yes, no or unsure. The votes stream back and the grid fills in live, usually in a couple of seconds, for well under a cent.
This is Jev's signature move made visible: not one clever answer, but a hundred fast, cheap, parallel decisions. The same fan-out is how you would score a thousand leads, classify a backlog of tickets, or moderate a feed — here it just happens to be opinions.
What the numbers show
- Decisions — how many of the 100 persona calls have returned.
- Elapsed & decisions/sec — how fast Jev fans out the parallel calls.
- Vote tally — a live yes / unsure / no breakdown as answers land.
- Jev cost — the whole 100-persona run at $0.042 per million input tokens.
FAQ
Are these real people or a real survey?
No. The 100 personas are synthetic — made-up profiles combining an age, a role and an outlook, so a question gets a spread of viewpoints. It is a demonstration of parallel AI decisions, not market research, and should not be treated as a real poll.
How can it answer 100 times so fast?
Each persona is an independent Jev API call, and they run in parallel. Jev is a System One model built for fast, typed decisions — tens to a few hundred milliseconds each — so 100 of them finish in about a second or two.
Is it free? Do I need a key?
It is free and needs no API key — the run is rate-limited on the live Jev decision API. To fan out decisions like this in your own product, get a hosted jv_live_ key on the pricing page.
What could I use this pattern for?
The same parallel fan-out is useful for scoring many leads, classifying or moderating a backlog, A/B judging variants, or getting a quick spread of takes — anywhere you need many small, fast decisions at once instead of one big model call.
More: All Jev tools · Batch Analyzer · Jev API docs · Watch Jev play 2048