# The Perfect Delivery — case notes

*Research file. The story lives in [the storyboard](comic-perfect-delivery.html) and
[the script](podcast-perfect-delivery.html) — this holds only the grounding neither surface can carry.
Nothing here restates the plot or references a panel number, so nothing here goes stale when a
draft is rewritten.*

**Lead:** Connie Chase. Delivery and intake are coordinator work.

**Feature:** the unanswered-question alert. Pulsar watches for a question that expected a reply
and didn't get one, and raises it after a few days — the *"no responses have been received"*
path in `DisplayAlert::Base`. Note what it is **not**: a delivery-status check. Every message in
this case sends and delivers perfectly.

**Trial:** the **SIGNAL** trial. A texting study where nothing arrives.

---

## What the case rests on

**The culprit is two digits and a piece of paper.** Recruitment ran on paper interest cards —
handed out at the clinic, IRB-approved, and changing them to a digital form needs an amendment
nobody has six months for. Sixty cards in sixty handwritings, typed in over one evening. Sixty
numbers, ten digits each: **six hundred chances to get one wrong**, and five hundred and
ninety-eight right. The two transposed digits produced a perfectly valid mobile number belonging
to a stranger.

> The Quiet Participant was *absence doesn't print*.
> The Invisible Line was *staleness doesn't print*.
> The Perfect Delivery is *sent was true, delivered was true, and both were useless*.

### Why nothing fired for six weeks

This is the part that has to be right, and it's the most interesting thing in the case.

**The system can only notice silence where it expected speech.** The nightly nudges were nudges —
nothing asked for a reply, so no reply was missing. The first message that *expected* an answer
was the six-week check-in, and three days after that went unanswered, the alert fired.

Six weeks isn't a delay. It's the first moment there was anything to detect.

---

## What the case argues

The intake was paper, the transcription was human, and the error was inevitable at that volume.
The alert didn't prevent the mistake — nothing prevents that mistake — it caught it in **three
days instead of never**.

That's a more honest claim than "our system stops typos," and it's the one that's true.

---

## Notes

**Nobody is a fool, again.** Connie made a transposition error in six hundred digits, which is
what happens to anyone who transcribes six hundred digits. The participant wrote her number
clearly. The stranger ignored texts that weren't for him. "Sent" and "delivered" were both
accurate. The failure is a paper front door.

**Don't let the stranger be a joke at his expense.** He's a man who got six weeks of unsolicited
step-count advice, which is funny once, from a distance. He isn't stupid and he isn't a nuisance.

**The alert must read as noticing, not diagnosing.** It cannot say "wrong number" — nothing
could know that. It says a question went unanswered, which is the only thing that was ever
detectable.

---

## Open

- Whether the stranger gets a line in the radio version, or stays offstage.
- Whether to name the participant (Mrs. Okonkwo in the current comic draft) or keep her a number.
- Two Connie cases in a row, which is correct by the assignment rule but worth watching.