The method

A rep is three minutes
and it does not care
how it ends.

Everything below is a rule the product actually enforces, not a description of an intention. The format is fixed on purpose: a training set you can change the length of is a training set you will quietly make easier.

Anatomy

What happens in
the three minutes.

  1. Before
    Sixty words of brief

    Where you are, who she is, what she is in the middle of, and what would make this go badly. Then the brief goes away.

  2. 0:00
    You speak first

    Nothing prompts you and nothing rescues you. She is not waiting to be helpful — she has her own reason for being in the room and it is not you.

  3. Anywhere
    The room happens to both of you

    Her friend comes back. The train is announced. She gets a message. Recovering from an interruption you did not cause is most of what actually happens.

  4. 2:30
    She starts to wind down

    One direction, once, thirty seconds out — and what she does with the last half minute was decided by everything before it. You are not told which way it went.

  5. 3:00
    The clock stops, she finishes her sentence

    Cutting a character off mid-line to save twelve seconds would truncate the best moment in the product. The clock reads zero; she is simply finishing, the way people do.

  6. After
    The scorecard

    Composite, six dimensions, evidence quoted from your own lines, and the two dimensions to take into the next rep.

She can also end it early. A character who has decided this is not worth having says one curt goodbye and goes, which is the outcome the simulator exists to make survivable.

The number

Sixty per cent measured.
Forty per cent judged.

The measured half is arithmetic on the transcript and cannot be argued with. The judged half runs at temperature zero against a fixed rubric and has to quote you to justify itself — if it cannot find the evidence, it returns nothing rather than inventing it.

Measured from the transcript
  • How much of the three minutes was you talking
  • How many of your questions were real ones rather than tags
  • Whether you followed what she said or changed the subject
  • How long you were willing to let a silence sit
  • Whether the ending was warm or trailed off
Judged against the rubric
  • Opening, curiosity, listening, signal reading, composure, close
  • Each one quoting at most twelve words of yours
  • One thing you did well, named before anything critical
  • Nothing about whether it worked, at any point

Transcription errors are never held against you. Speech recognition mangles words, and if her reply shows she understood you, you were understood.

The law

Why the ending
is worth nothing.

Score the outcome and you teach the outcome. A user who is rewarded for a yes learns to push for one, and pushing is both the wrong skill and the thing that makes somebody else’s evening worse. So the grader is told, in the rubric itself, that whether she gave a number or walked away contributes zero — and that if it finds itself rewarding a rep because it went well, to stop and score how it was played instead.

The practical consequence is the one that matters: a clean rep that ends in rejection can score 92, and a sloppy one that got lucky scores 54. Tiers open on scores, not on wins, so there is nothing to be gained by chasing an ending.

Opens a tierScore 70+ in 2 reps below it
Opens nothingWinning
Costs youMissing a clear no
The characters

They remember one line,
and it is not about you.

A character carries at most one line from a previous encounter, and it has to be about the encounter or about her own situation — the book she could not find, where she was going. Never how you did, never whether she would like to see you again.

Allowed

“Still looking for the blue one. Sister’s birthday is Thursday.”

Refused by the code

“I’ve been hoping you’d come back.”

The second one is not filtered out by a style note. It is rejected by a function that refuses rather than sanitises, because the failure it prevents is this becoming a companion app — and that is a boundary worth enforcing in code. You can clear every line every character is holding, in one tap, whenever you like.

The field

The half that
happens outside.

Four tiers of graded exposure, in order, because going too hard too early sensitises rather than habituates. Nothing at any tier depends on somebody saying yes.

  1. Tier 1 · In the appEnd a rep yourself.

    No social risk at all. Day one work — hearing yourself do the thing once.

  2. Tier 2 · TransactionalAsk a stranger for the time.

    A person, a sentence, an answer. The point is the asking, not the information.

  3. Tier 3 · SocialCompliment a choice, then leave.

    Say your piece and go. If they are working, in headphones or alone at night, this is not the moment — walk on and find another one.

  4. Tier 4 · The real thingTake a no well.

    Make an ask, and when it comes back no, say something warm and leave inside ten seconds. No bargaining, no joke to soften it, no second attempt.

Before each one you say how nervous you expect to be. Afterwards you say how nervous you were. The gap between those two lines, drawn over weeks, is the most useful chart in the product — and for almost everybody it closes.

The count

You collect the refusals
on purpose.

  • 10
    Ten no

    Ten times you asked for something and did not get it. Nothing happened. No one told anyone. You went home and had dinner.

    That is the entire lesson. The rest is repetition.
  • 25
    Twenty-five no

    Twenty-five is past the point where you can call any of them a fluke. You have a habit now, and the habit is asking.

    Look at what you predicted for the first few. Then at what they cost.
  • 50
    Fifty no

    Fifty refusals is more than most people collect in a decade, because most people arrange their lives so they never have to hear one.

    You went and got them on purpose. That is the difference.
  • 100
    One hundred no

    A hundred. At this point being turned down is a thing that happens on a Tuesday, not a thing that happens to you.

    Nothing left to prove here. Go and ask for something that matters.
Standing

Four ranks, and none of them
are a badge shelf.

  1. Rookie

    You have started. That is the part most people never do.

  2. Regular

    You can hold a conversation with someone who is pleased to see you.

  3. Contender

    You can keep one going with someone who is not helping you.

  4. Closer

    You read a person who gave you nothing to read, twice over.

On Sunday you get one letter: what moved, what did not, and the single thing to take into next week. It is written from your reps, not from a template.

Ready

Three minutes,
and then you know.

The first tier is nearly impossible to fail. That is deliberate, and it is stated in the brief, in those words.