Peregrine Phone — live session test plan (Beckett)
2 Oct 2026. Action item 10 of the 1 Oct follow-up meeting: the test plan for the live session, using the settled pass bar. Sources, later wins: the first Phone meeting's "Live session" section, then the follow-up meeting's pass bar and DAY 7 rulings, then docs/plans/2026-10-02-phone-build-spec.md.
1. Order of events
- Meeting two: the Empty Chair review of the revised stations, same experts, same four criteria. The real Beckett sees nothing until it passes.
- Readiness gate (§2) is met.
- Session A, evening: the night of DAY 8 (after midnight, DAY 9), on a TestFlight build.
- Session B, the next morning: DAY 9, if Beckett opens the phone on his own.
Beckett can be asked to test only a few times, so the session runs only once every gate item is true. A dead tap in front of him spends a session.
2. Readiness gate
Each line is true before Session A. State on 2 Oct is in the right column.
| Item | Owner | State 2 Oct |
|---|---|---|
| Meeting two passes | Sarah, the three experts | Not held |
| A launch flag hides every room but the Phone and the Desk card | Claude | Built: the Peregrine-PhoneSession scheme (PHONE_SESSION), or -phoneSession |
| Our status bar: written "now" ticking in real time, written battery, the iOS bar hidden | Claude | Built |
| Night lock screen (two noise notifications, one hairline on a catch) and morning lock screen (one tern receipt notification) | Claude | Built (wallpaper is a stand-in) |
| Home screen: ten tiles, red dots on Phone and Messages | Claude | Built (tiles are the r4 drawings) |
| Five live stations: Rides, Location History, Messages, Mail, Photos | Claude | Built, plus the Phone app, Browser, Notes and Calendar, so no home-screen tile is a dead tap |
| Desk card: opening line, "hold to put it away", the three closing lines in precedence, coins (10 a catch) | Claude | Built; the put-away logs the coins as a night through logsNight, which I have not yet seen land in the ledger |
| The morning session unlocks on the device's real date change; skips the picker and the Desk card; a same-day reopen lands on the night's end screen | Claude | Built; a real midnight is untested (B3 in the master list) |
Every station's copy is playable in the corpus | Sarah | 30 drafts, 0 playable |
| The four games pass a feel check on a real phone | Sarah | Coordinate Drop, Split Decision, Odd One Out pass (2 Oct). Red Flag: mechanics pass, visuals being refined by Sarah |
| The 280 hand-written game items are sculpted (240 before H6 added 40) | Sarah | Drafted (Red Flag 120, Split Decision 160) |
| Personal bests save and show | Claude | Built |
| No tap anywhere in the session's reach is dead, except the three Browser links the map makes dead on purpose | Claude, checked by Sarah | Open until the build exists |
3. Session A — the night (DAY 8)
Setup. A TestFlight build on Beckett's own phone or one he uses daily. A fresh profile. The session set to the night (now: DAY 9 · 1:14 A.M.). Sarah facilitates.
First minute (nothing is explained):
- The profile picker: one tap, and Beckett picks his photo.
- The Desk card alone: "You don't remember 10:30 to 12:10. Your phone was with you the whole time." The phone lies on the card.
- Tapping the phone slides to the DAY 8 lock screen with two noise notifications, a landlord Mail and a Photos memory.
- Swipe up to the home screen.
- His next tap is his first station.
Silence rule. Sarah may re-read the Desk card line or end the session warmly. She never explains a mechanic, never points at an app, and never answers "what do I do".
Ending. Beckett ends it himself with "hold to put it away", or Sarah ends it warmly. The Desk card shows its closing line. No line refers to a later session.
4. Session B — the morning after (DAY 9)
The morning session unlocks on the device's real date change. Its lock screen carries one notification: the tern receipt, 12.50 · DAY 9, 8:40 A.M. Nobody asks Beckett to open the phone. Burn night is not this morning; it lands at least a day later (ruled 2 Oct).
Record how he got to Rides: through the notification, through the home screen, or not at all. The first meeting measured replay as a return with no prompt, and the follow-up kept the tern notification, so the path is the only way to tell the two apart.
5. Pass bar
Signals, not proof. Each criterion passes or fails on what Beckett does, recorded as it happens.
| Criterion | Passes | Fails |
|---|---|---|
| Immersion | A first-person remark about the night ("where was I…") | He asks what to do, or flags something as not iOS |
| Engagement | He opens three case apps unprompted. The Games folder never counts | He stalls and looks to Sarah |
| Entertainment | An unprompted reaction to the ping stack or the 11:31 text, and a new personal best he screenshots | He sets the phone down silent, or calls it a quiz |
| Replay | He opens Rides the next morning and reads the fares before making a call | He never returns, or goes straight to the graded line |
The first meeting's replay fail read "goes straight to the surcharge line". The morning's graded line is now the tip (ruled 2 Oct), so the fail condition is written here as the graded line.
Not tested: the Phone's "be wrong about a person" stake, which needs several sessions and a person-reading station.
6. Observation sheet
Sarah fills this in during the sessions. Times are minutes from the first tap.
| Time | Where he is | What he did or said | Criterion it bears on |
|---|---|---|---|
After Session A
- Stations opened, in order:
- Graded calls: station, caught or wrong:
- First-person remarks, verbatim:
- Questions he asked, verbatim:
- Any tap that did nothing:
- Did he open the Games folder? Which game, and did a new best happen?
After Session B
- Opened the phone the next morning: yes / no, and at what time
- Reached Rides through: notification / home screen / not reached
- Read the fares before calling: yes / no
- Morning call (the tip): caught / wrong / none
7. Open for this plan
- Device. His phone or a household phone. Owner: Sarah.
- Session timing. Session A has to end before midnight on his device for the morning session to arrive the next day. Owner: Sarah.
- What counts as "reads the fares". Sarah's judgment in the room; the build does not log time on screen.