ClaudeQACoverage has carried "long-horizon accuracy is tested only at the
unit level" as a standing gap for four rounds. The simulated-history half
is now closed by LearningCurveTest, and the entry says so -- along with
what it found, since a harness that changed nothing would be a weaker
claim than one that caught two real faults.
The other half stays open and is stated plainly: a simulation exercises
the engine, not the app. Nothing here proves a real install accumulates
those cycles correctly over a year, that snapshots survive updates and
time-zone changes, or that the accuracy card shows what the engine
computed. That needs elapsed time on a device.
Development log entry per WORK_CYCLE step 6, including the two
corrections this batch made to its own plan: the coverage dip was a
40-seed artifact, and a swept constant shipped one value away from the
measured one.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
closes#36
The deletion has existed since Batch 01 with an instrumented test and no way to
reach it. This adds the Privacy & Security section, a confirmation that says
what goes and what stays, and PrivacyViewModelTest.
A PRIVACY CONTROL WHOSE LABEL DID NOTHING
The confirmation promises "your reminder settings are unchanged". Proving that
meant changing a setting first, so I tapped "Maximum privacy" on a device and
nothing happened: PrivacyRow and onboarding's PrivacyOption both put onClick on
the RadioButton and left the row inert. The option that decides what a lock
screen shows could only be changed by hitting a 20dp circle — in both places a
user ever chooses it.
Modifier.selectable on the row, onClick = null on the radio. That is Material's
documented pattern and it also merges the semantics, so TalkBack announces one
selectable option instead of a radio button and two loose strings.
Found by trying to verify a different claim, which is the argument for verifying
claims rather than asserting them. The setting does survive deletion — set to
Maximum privacy, deleted everything, still Maximum privacy.
A DESIGN THAT WAS WRONG BEFORE IT WAS WRITTEN
The first draft cancelled the reminder schedule on delete. Seems obviously
right; is not. ReminderWorker reads the forecast each run and NoData maps to no
decision, so scheduled work already does nothing while there is nothing to say —
and scheduling only happens from ReminderCoordinator and the settings screen, so
cancelling would have left reminders silently off until the user next toggled
something, long after logging a new period. Checked the call sites instead of
reasoning from the name.
Delete touches health data only. UserPreferences is a separate store precisely
so a privacy action cannot reset a choice somebody made, and there is no undo —
§45 says irreversible, and an undo snackbar keeps the data alive for its
timeout.
Round 4 recorded in docs/qa/. 194 tests pass; ./gradlew check, schema-guard and
doc-claims all pass. Driven on PeriodMinSdk26.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
At font_scale 2.0 the bottom navigation read "Calenda / r" and "Setting / s".
NavigationBarItem's label had no maxLines, so Compose wrapped it rather than
truncating, breaking a word across two lines in a 4-item tab bar.
maxLines = 1 with an ellipsis degrades to "Calen…" instead — still recognisable,
and the icon above it carries the meaning regardless.
FOUND BY DOING THE THING NOBODY HAD DONE
Font scaling was a standing QA gap: written down, never run. Driven at 1.3 and
2.0 on PeriodMinSdk26, through onboarding to Today.
Everything else held. The 72sp hero survives 2.0, Today scrolls so nothing below
the fold is lost, and every onboarding step keeps its primary button reachable —
including the two that carry three buttons or three option cards beneath an
illustration, which is why their art is 104dp where the rest take 120-128dp.
That sizing was guessed when the artwork landed and is now checked.
This defect is older than the artwork work and unrelated to it; the font-scale
pass is simply the first thing that looked.
Recorded in docs/qa/ClaudeQACoverage.md, where the gap it closes was listed.
Still unreached: font scaling in dark mode, and on anything other than a phone.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
A second AVD exists — PeriodMinSdk26, API 26, Pixel 6 — and the app was built,
installed and driven on it end to end. Everything before today ran on API 36,
which the standing gaps list has said since Batch 01.
The QA plan now names both emulators, because rounds cite PeriodQA by name and
nothing defined it. It also carries the recreate command and three mechanical
traps that each cost time today:
- a windowed emulator dies with its session's X server (XIO: fatal IO error on
X server ":0") and took one run down mid-flow; -no-window has no such
dependency and screencap still works
- adb regularly has three devices attached on this machine, since other projects
keep emulators running, so the serial must be resolved from the AVD name
rather than assumed to be emulator-5554
- the debug build's application id carries a .debug suffix, so
`monkey -p dev.privacyllc.period` aborts with "No activities found to run"
after a successful install
The coverage doc's "No device at minSdk" gap is marked partly closed, with what
was reached (onboarding to first forecast, the date picker, all four tabs, a
relaunch after a cold boot) and what was not (notifications, the instrumented
suites, anything needing more than one cycle of history).
Also fixes a guard failure I introduced in ad085fb: that commit's log entry
named docs/architecture/scripts/forgejo-issue.py in backticks, and doc-claims.sh
reads a backticked path as a claim the file exists. I had run the guard before
writing the entry and not after. Named bare now, per the convention
DOC_TRUST_MAP.md already documents; doc-claims reports 255 paths, all present.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Every document in the tree audited against the source, the tracker and git
history, each finding then given to a second reader tasked with refuting it.
74 raised, 12 refuted, 57 applied. No code changed.
THE README DESCRIBED A SKELETON
Its Status table — the one place a claim about what is built is allowed to live
— still read "there is no usable app yet", with Not built against Room, the four
core screens, fertility and notifications, and No round run against QA. Five
batches had shipped and three QA rounds had run.
TWO DOCUMENTS WERE SILENTLY NEVER FIRING
architecture/README.md and design/README.md wrote Governs: as prose ("the Gradle
module graph", "the design tokens in core/designsystem"). Neither contains a
path token, so doc-triggers.py reduced them to globs matching nothing, and one
real glob apiece made them look path-governing rather than subject-governing —
the state the script's own header calls invisible. Editing a Room entity never
fired the document owning the migration table. Both now fire, proved by running
the script.
SECURITY.md CLAIMED FOUR UNBUILT PROTECTIONS
App lock listed among what works offline; biometric/PIN gating described as
protecting app launch; the incognito launcher as existing; Play Billing in the
third parties table without the "not yet integrated" marker its neighbours
carry. All are Batch 06/07 work.
The advertising boundary was overstated in SECURITY.md and the README alike:
both said the ads module declares no dependency and a guard proves it. There is
no ads module. The pre-declared ":core:ads" to emptySet() rule is stricter than
the sentence it replaced and matches nothing until Batch 07, which is why the
guard is proved by injection rather than trusted.
SMALLER, EACH A REAL TRAP
WORK_CYCLE.md pointed at docs/architecture/scripts/forgejo-issue.py, a template
path absent here — missed by doc-claims.sh, which reads backticked prose and not
fenced blocks. ClaudeReport.md's Round notes said "No rounds yet" after three
rounds because ClaudeQAPlan.md's after-a-round list never named that section;
the playbook is fixed first. The instrumented-test count was eight in three
places and is four. HISTORY.md said the repository had no code and that nothing
had been tried and dropped, when three approaches had.
DELIBERATELY UNCHANGED
ClaudeReport.md's last verified build SHA stays at 0451fbe — no round has run
since, and moving it would claim a verification nobody performed. Every
DEVELOPMENT_LOG entry stays as written.
Guards: ./gradlew check, schema-guard.sh, doc-claims.sh (235 claimed paths, all
present), doc-triggers.py, and a link sweep over 21 markdown files.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
"Period" was always a working name — PRODUCT_PLAN.md §55 said so. The real
identity arrived with the brand guide and artwork, and this makes the project
call itself by it.
The repository is renamed in place on Forgejo (null/Period ->
null/Privacy-Period-Tracker), which keeps all 30 commits, all 27 issues, all 8
milestones and the four severity labels — verified by counting them on both
sides rather than assuming a rename is lossless. The git remote follows in the
same breath, so this commit's automatic push is what proves the new URL works.
WHAT DELIBERATELY DID NOT MOVE
"Period" is the product name AND the central domain word, and a mass rename
would have turned the data model into nonsense. PeriodRecord, PeriodWriteResult,
confirmPeriodStart, the period_records table and the button that says "Started
period" all describe a menstrual period rather than a product, and all 239
references to them are untouched.
So are the Kotlin package and applicationId. dev.privacyllc.period already reads
correctly under the new name — privacyllc is the company, period is the app —
and changing it would rewrite 68 files, rename the Room schema directory and
break the hardcoded path in schema-guard.sh for no gain. period.db stays for a
sharper reason: renaming a database file orphans the data on every device that
already has it.
Sixteen files changed, and the diff is small on purpose.
THE NAME HAS A SPACE AND THE ARTWORK DOES NOT
Canonical is "Privacy: Period Tracker". The supplied wordmark sets it without
one, so logo.webp and banner.webp now disagree with every document and with the
app itself. BRAND_GUIDE.md §10 is updated to the space form and says plainly
that the drawn mark has not caught up. Filed separately rather than papered
over, because closing that gap needs an artist and not a rename.
Round 3 adds pass F, which became runnable for the first time. Eight
instrumented tests assert what a lock screen would render; the row says plainly
that nobody has yet looked at an actual locked screen, because the tests check
the notification object and the last mile is what the system chooses to draw.
Three guards have now failed their first proof — schema, boundaries,
permissions. Recorded as a rule rather than a run of bad luck: assume a new
guard is broken until it has been watched failing.
Round 2 covers every screen Batch 03 built, pass by pass, with each partial and
not-run row carrying why. Three new standing gaps, all of them things nobody has
done rather than things that failed: TalkBack has never been run, text has never
been scaled, and nothing has run at minSdk.
The rule this session keeps demonstrating is now stated plainly in both files:
the defects in this project are found by running it, not by reading it.
Round 1 was partial and says so, pass by pass, with each not-run row carrying
why. It found two defects that 70 green unit tests could not: the primary
button crashing the app on a second tap, and two API 34 date calls on the
recalculation path that would crash every device below Android 14.
Neither is filed in the tracker. Both were found and fixed inside the batch that
introduced them, before any build left this repository, and an issue closed by
the commit that created the code would be bookkeeping rather than a record.
New standing gap: nothing has been run on a device at this project's own minSdk.
Lint is a good guard for that and is not a substitute for one run on API 26.
Period was a bare directory holding one 2,527-line specification, with no git
repository, no tracker and no documentation convention. This is the adoption
from Projects/Template/START-HERE-New-Project.md, plus a project that compiles
so the hooks and future guards have something real to run against.
Documents. scaffold.sh created 19 paths, 0 skipped. The specification moved to
docs/planning/PRODUCT_PLAN.md unchanged in substance, with a status header; the
capitalised Docs/ is gone. Every scaffolded document was filled in for Period.
docs/OPERATIONS.md deleted — an offline app is not a deployed service.
DOC_TRUST_MAP.md written last, describing what is actually here, including what
this project deliberately does not have.
Code. Four Gradle modules. domain/cycle and domain/prediction are kotlin("jvm")
and cannot see the Android SDK, so the engine is testable without an emulator —
17 tests pass, 12 of them the acceptance cases from PRODUCT_PLAN.md §51.
BaselinePredictionEngine is a robust-median prototype and explicitly not the
product; it exists so Batch 02's replacement can be shown to be better rather
than merely different.
Versions verified against their official sources today rather than inherited
from the specification's own numbers, which that document asks for: Kotlin
2.4.10, AGP 9.3.1, Gradle 9.7.0, Compose BOM 2026.08.00, Room 2.8.4, Hilt
2.60.1. AGP 9 ships Kotlin built in, so org.jetbrains.kotlin.android is no
longer applied. compileSdk is 37 because current AndroidX requires it; targetSdk
stays 36, Play's floor from 2026-08-31, and the difference is deliberate.
Six scripts taken into scripts/; the rest declined and named in docs/TOOLS.md.
Three hooks in .githooks/, with pre-commit adapted to Gradle.
closes#1closes#2
null