Updated Aug 13, 2026 Β· current deploy 7d68484e Β· voice-first surgical case logger
Everything found this cycle and closed. Each verified with offline tests β no live Claude calls.
| Item | Area | What was wrong β fix | Status |
|---|---|---|---|
| C2 | Accounts | A 2nd account on a shared browser inherited the 1st's cases / app-lock / API key β local data now wiped unless it provably belongs to the same account. Cross-account data isolation. | Fixed |
| E2 | CPT | Catalog was 18 codes β scraped ACGME's full table: 4,491 codes with authoritative category mappings. | Fixed |
| E3 | CPT | Datalist β live typeahead: type a code, procedure, or eponym (whipple) β tap to add. | Fixed |
| P7 | Parse | "Colorectal surgery" didn't resolve β rotation list merged with ACGME's full list (36). | Fixed |
| β | Categories | Adding a CPT now suggests its ACGME categories to tap (non-destructive). | Fixed |
| P8 | Parse | Trauma cases didn't check "Involved trauma" β now auto-checks from the trauma category / CPT. | Fixed |
| P9 | Parse | "Unable to determine procedure" dumped a sentence + junk CPT β now blanks + flags cleanly. | Fixed |
| E1 | Form | Approach silently defaulted to "Lap" β now blank (β). | Fixed |
| D1 | Dictation | Screen lock killed the mic mid-dictation β screen wake-lock while recording. | Fixed |
| D2 | Dictation | Mic dropped silently β visible "mic stopped β tap to resume" warning. | Fixed |
| M1 | Multi-case | Couldn't see what was said per case β each review shows its own "What you said". | Fixed |
| T1 | Techniques | "lap" vs "laparoscopic appendectomy" made 2 techniques β normalized dedup (open/lap stay separate). | Fixed |
| T2 | Techniques | Seeded "Sleeve gastrectomy" looked like real data β now badged "Sample". | Fixed |
| B1 | Bug report | One screenshot only β up to 4 screenshots. | Fixed |
| P1 | Parse | ~60s cold first-parse β free warm-up on capture-open (pre-connects to Anthropic; no cost). | Fixed |
| P11 | Offline | "Load failed" toast β "π΄ You're offline β parsed on-device instead." | Fixed |
Confirmed working in your self-test (clean input).
Not oversights. Each with the reason and my recommendation.
| Item | Why it's open / recommendation | Status |
|---|---|---|
| P5 MRN "one 2345" | Recommend leaving. Auto-converting spoken number-words into an exact patient ID risks logging a wrong MRN. Blank + flag (you type it) is the safe default. | Held |
| P3 Garbled fields | Can't reproduce from code β anomalous model output, not a code bug. If it recurs, screenshot the transcript + bad fields and we'll catch the cause. | Held |
| P6 Parse accuracy | Needs live-parse iteration, which the "no live Claude calls on your credit while testing" rule blocks me from doing solo. Revisit deliberately. | Held |
| O1 Default site | Hard-defaults to "LA General" β fine for this cohort; make blank / configurable for a general or native-app release. | Decision |
| O2 Roster onboarding | Rotations now seeded; attending + site onboarding (CSV import or quick paste) is the remaining piece β a feature to design. | Feature |
| A1 Admin access | Admin page is a little hard to reach. Minor UX; low priority (only you use admin). | Minor |
The next testing pass β tap to check off. Your progress is saved on this device.
Raw chronological notes live in OpTally-Testing-Log.md. Security items tracked in OpTally-Security-Audit.md. This page's checklist state is stored only in this browser.