01
Predict
Write the ask and commit a probability that it works. The prediction locks the moment you save it — there is no edit button, because a mutable prediction is worthless data.
A 30-day experiment for iOS 26
Find out how wrong you are about people.
You think people will say no. Research on how often others agree to direct requests suggests that expectation runs low — but you have never measured your own. NoJar tracks your predictions against reality for 30 days and shows you the gap, in your numbers.
Coming to the App StoreiOS 26 · $5.99 one-time · no account
The inversion
Your score is how many times you get told no.
The weekly target counts rejections collected, not asks made. A yes does not count. That single inversion removes the failure state — the one that makes people stop asking in the first place.
- A no moves the week forward.
- A yes is a miss. Said plainly, without a verdict attached.
- Backing out gets no colour and no comment. It still gets logged.
- Three no's is roughly seven asks a week. The app says so up front.
How it works
Predict, ask, resolve, calibrate.
02
Ask
Your own ask, or one from a library of around 120 curated prompts across four tiers. Nothing in it involves deception, pressure, or people who cannot decline freely.
03
Resolve
They said yes, they said no, or you backed out. Two taps. A no counts toward the week. Backing out is logged without comment — it is data too.
04
Calibrate
After 15 resolved asks the curve unlocks: what you predicted against what actually happened, bucketed, with intervals wide enough to be honest.
What you get at ask 15
Not a study. Your spreadsheet, reading your own numbers back to you.
The headline is plain language and it flips three ways — you may turn out to underestimate people, to be calibrated, or to overestimate. The app reports what your data says and skips the verdict.
It also states its own limitation, on the same screen: the asks you choose are not a random sample, so the observed yes rate is your rate under your choices, not a fact about people in general.
Example screen
You predicted 43% on average. People said yes63% of the time.
You underestimate people by 1.5×.
n = 19 resolved asks
Inside
An instrument, not a coach.
The reliability curve
Predicted probability against observed yes rate, with a reference diagonal. Points above the line are the asks you talked yourself out of expecting.
Dodge rate by prediction
The metric nobody else tracks. Not whether you were right — whether you asked at all, and at which odds you stop.
The Day 30 report
Totals, drift from week one to week four, and the one that mattered: the ask that worked which you rated lowest. Shareable as a single image.
Weekly, not daily
The quota runs on weeks and forgives a bad Tuesday. There is no daily streak to break and nothing to lose by missing a day.
Your data, exportable
Every ask, prediction, outcome and note goes out as CSV whenever you want it. The app holds your data, it does not hold it hostage.
Nothing leaves the phone
No account, no server, no analytics SDK. The App Store privacy label reads Data Not Collected, and that is the whole architecture, not a policy.
Price
One purchase. No subscription.
Full app$5.99
Free asks10
Subscriptions0
Trackers0
The free tier logs 10 asks. One purchase unlocks unlimited asks, the calibration screen, the Day 30 report, the full ask library and CSV export. It stays unlocked, on every device signed in to your Apple Account.
Coming to the App Store
Worth saying plainly
What this is not.
- It is not treatment. This is a curiosity experiment about your own predictions. If asking for things causes real distress, a therapist is the better tool.
- It is not a dare app. There are no stunts, no AI-generated challenges, and no prompts that push anyone into a corner.
- It is not a social network. No feed, no friends, no leaderboard, nothing to post.
- It is not forever. Day 30 is the finish line and the app says so. It keeps working after that, and it stops asking for your attention.
Questions before you buy? Support answers most of them.