Client-only photography feasibility tests (T01–T07, T11, T12). Nothing is uploaded anywhere. All processing happens on this device. No accounts, no storage services. Spike v3 — disposable. (v2: filenames inserted as text only; runs report attempted/succeeded/failed and incomplete runs are labeled. v3: photos saved to the on-device queue are restored into the session on load — with their slot/QC metadata when present, safe defaults for older records — so recovered photos can be quality-checked and exported; failure messages are always recorded as real text.)
Standalone display: open this page from a Home Screen icon and check the "Display mode" row — it should read standalone.
For each photo, the journal records the file name, the format the page actually receives, and its size — that delivery format is core T02 evidence.
Pick a processed photo, then re-encode it at different quality levels and compare byte size vs legibility at 100% zoom. Quality always wins over the size target — record the lowest quality where labels/flaws stay readable.
Thresholds are PROPOSED starting points to calibrate on real photos: mean brightness < 60 → DARK, > 235 → BRIGHT, sharpness (Laplacian variance) < 60 → BLUR. Flags are advisory only.
Then follow the test plan's interruption steps (switch apps, lock the phone, Airplane Mode, swipe-close) and reopen this page. The journal records hidden/visible events, and this section shows what was recovered. If the page reloads mid-run, that is recorded too — it is T03/T06 evidence, not an error to hide.
Listing photos are included by default. Measurement photos are excluded by default — check one only when you deliberately want it in this package; the manifest flags it as owner-included. Files are numbered in package order (01, 02, …).
After testing, send this journal to Bookie (copy/paste or share). It is the evidence record for the matrix — no test is passed without it.