Build 843 Working Hardware Evidence Review¶
| Field | Value |
|---|---|
| Classification | Controlled review of working/support evidence; not formal verification evidence |
| Subject | JRC-FORMAL-VV |
| App | BionicLoop 1.0 (843) / CC-DEV |
| Source | walkabout-candidate-2026-08-21-build-843 / 1455cda7ad5d3e164b4a955ea0e3bc725a9996e9 |
| Exercise window | 2026-08-24 14:30 EDT through 2026-08-25 11:10:29 EDT |
| Review date | 2026-08-27 |
| Updated | 2026-09-08 |
| Sponsor status | Supporting evidence accepted under D08; no separate formal hardware-validation claim |
Purpose And Baseline Boundary¶
This review correlates the completed Build 843 formative real-device evaluation with the device-local recovery export, operator observations, and BionicScout records. It supports an impact and evidence-reuse decision for the intended IDE study build.
Build 843 is not the same configuration identity as engineering freeze
ide-software-freeze-2026-08-21 at
91c0e98a9bc9429a0486bebdebffc7d8dbbe300e. Exact-freeze formal evidence remains
governed by the 2026-08-21 Freeze Execution Report. This review did not itself
designate Build 843 and does not promote working evidence to formal evidence;
the separate designation was recorded on 2026-09-03 in Appendices A10C and
A18.
Evidence Hierarchy¶
For software reconstruction of the tested algorithm session, the preferred working-evidence order is:
- the clinical-gated, device-local recovery ZIP containing the protected step CSV and byte-exact files for the selected Algo2015 session
- controlled automated evidence bundles for requirements exercised by those suites
- the operator record for visible UI, notification, and physical handling observations
- BionicScout/Dynamo and app-log records as corroborating backup, sequence correlation, and gap investigation
Scout is not required for local algorithm execution or dosing. Official study sources are Dexcom Clarity for CGM outcomes, the BionicLoop exported CSV for insulin-delivery outcomes, and staff-entered electronic CRFs for adverse events and other safety information.
The available ZIP was exported at 09:38:30 EDT on 2026-08-25. It is therefore primary local evidence only through its 09:33:48 EDT final step. Temporary Target and suspension exercises after that cutoff currently rely on the operator record plus corroborating telemetry. A post-run ZIP is required if a single local export is expected to cover those later steps.
Retained Device-Local Export¶
The controlled source repository retains the supporting ZIP and extracted-file hashes under the Build 843 engineering evidence record.
| Artifact | Size | SHA-256 |
|---|---|---|
JRC-FORMAL-VV-algo-files-20260825-093830.zip |
297,010 bytes | 446914e79d535f32c646e003a8ddc639b6dde661838cfe1eced74a44f8cec409 |
BionicLoop_StepTelemetry.csv |
99,327 bytes | 77f7d00a1dd65705faf49f0bdb920a676fa499b9e756a3f10838b9aebf8a1067 |
BP_1787597333.txt |
69,425 bytes | 3fdad25b0bda38eaf993c1aaf3210497ef330e32f791c1f717a617b02fce234e |
BP_LOG_1787597333.txt |
118,181 bytes | 52b0c2fa9952848b9816692c051a642e8cbe131cb822ec27f9a5fdab27dc5ba4 |
BP_ReadMatrix.m |
9,589 bytes | e84eec770f5d1fc2fcd4f39559b8076dcad46b3bc81c45f2e954799eb3bb4d5a |
The ZIP opens, contains four nonempty files, and contains one algorithm epoch.
Epoch 1787597333 is 2026-08-24 14:48:53 EDT and matches the session start. The
CSV contains 222 data rows plus its header, from step 0 at 14:48:52 EDT through
step 225 at 09:33:48 EDT. No unrelated algorithm session was included.
The CSV records manual BG consumption at step 7 (133 mg/dL) and step 164
(99 mg/dL). It also covers the ordinary-use, meal, relaunch, extended-CGM-
outage, same-Pod fallback-recovery, and canceled-partial-meal portions performed
before the export cutoff.
Deterministic Checker And Operator Correlation¶
The bounded checker run used signature set 0.13 and checker commit 564f912f
over 2026-08-24 18:30:00Z through 2026-08-25 15:11:00Z. It reviewed 7,510
timeline events and 6,105 app-log lines, all identified as Build 843.
| Exercise ID | Checker result | Correlated disposition |
|---|---|---|
| INT-01 | Partial | App logs are present, but the expected remote_logging_config_applied marker is absent; tool/config evidence disposition required |
| INT-02 | Pass | Readings present; no discrepancy reported |
| INT-03 | Pass | Algorithm session continued through relaunches |
| INT-04 | Pass | Meal delivery/progress cleared; commands and feedback present |
| INT-05 | Missing by frozen signature | Manual-BG submission used bg.modal.submit_button; signature 0.13 incorrectly searches for home.bg_button on the submit event. Operator and ZIP prove entry and step consumption. Preserve as an execution-tool deviation rather than changing the signature after execution. |
| INT-06 | Pass | CGM-outage guidance, reminder, backup-basal bridge, BG recovery, and resumed stepping correlated |
| INT-07 | Not executed | Conditional Pod-change exercise; no Pod change occurred in the bounded run |
| INT-08 | Pass | Four launches, same algorithm session, no step 0 or stuck meal state |
| INT-09 | Pass | Operator confirmed expected notification presentation; scheduled-notification evidence present |
| INT-10 | Pass | Short Pod signal-loss alert progression and recovery were expected |
| INT-11 | Pass | 120 mg/dL target activated, first applied, and ended early for one target identity |
| INT-12 | Pass | 130 mg/dL target applied, persisted across cold relaunch, expired, notified, and restored permanent target |
| INT-13 | Pass | Suspend/resume produced paused Home state, blocked meal, reminder, escalation, Pod-confirmed resume, and successful live step |
Additional high-risk working observations:
- a canceled 4.25 U meal delivered 0.35 U, survived force-quit/relaunch, and was consumed exactly once at the next applicable step without full-dose reconstruction or a stuck meal state
- a long same-Pod outage reconciled 0.15 U of confirmed backup-basal delivery over bounded applicable steps and resumed live execution
- during suspend, no pump command was sent; resume recovery recorded a verified zero delivery delta for the bounded fallback-active interval before the next live command
Limitations And Promotion Requirements¶
- The working operator sheet lacks phone model/iOS and operator signature/date; this does not block its use as supporting formative evidence because no separate formal hardware-validation claim is made.
- Pod lot/expiry and CGM identity were not captured in this record.
- INT-07 and a same-run G7 replacement were not exercised.
- The recovery ZIP predates the later Temporary Target and suspend/resume exercises.
- A checker pass or operator report remains supporting evidence unless it is promoted through the study's controlled evidence process.
- Build 843 was designated for IDE submission on 2026-09-03. This supporting
evidence was accepted under
D08; no separate formal hardware-validation claim is made.
Conclusion¶
The evidence supports use of the recovery ZIP as the primary device-local
technical record for its covered session interval, with Scout telemetry as
corroborating backup. The evidence closes the engineering observation that
Build 843 was installed and operated with a real Pod. Sponsor evidence
acceptance is recorded under D08; Build 843 was approved for IDE submission
on 2026-09-03. The no-separate-formal-
hardware-validation claim boundary was clinically approved on 2026-08-27.