From c819e99189e652dc137ffe42822a071b82e1119c Mon Sep 17 00:00:00 2001 From: uvtqt Date: Fri, 13 Nov 2026 20:50:43 +0100 Subject: [PATCH 1/2] AGENTS.md: the thin-data stress item is fixed, not open MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit §4.1 listed "stress shows a confident score on ~20 min of data" as still open. That was #148, the 100 - readiness fallback removed in adb51173 (closed 2026-08-02). Stress now comes only from the Baevsky SI over a scored night's beats, and the stager scores a night only from 60 min of sleep (90 min in the daytime), so the reported case cannot recur. Co-Authored-By: Claude Opus 5.5 --- AGENTS.md | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/AGENTS.md b/AGENTS.md index 4e2962023..e8645d7de 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -157,8 +157,10 @@ two copies of a `100 - readiness` stress fallback; RHR falling back to daytime H oxygen dips; a skin-temp section gated on `spo2` presence; `StageBars` drawing an *invisible gap* for an absent sleep stage; pace showing absurd numbers instead of "—"; a false empty state instead of a retryable error; Bluetooth-off reported as -"no strap found". One is still open: stress shows a confident score on ~20 min of -data. +"no strap found". The "confident stress score on ~20 min of data" report (#148) +was the `100 - readiness` fallback, removed in `adb51173`: stress now comes only +from the Baevsky SI over a scored night's beats, and a night is scored only from +60 min of sleep (90 min in the daytime). **Ask on any metric diff:** what does this return when the input is missing or thin? Anything other than null/"—"/an honest low-confidence envelope is a bug. From 827074c4d159902c69102e4c18402e888e7ace08 Mon Sep 17 00:00:00 2001 From: uvtqt Date: Fri, 13 Nov 2026 22:11:59 +0100 Subject: [PATCH 2/2] AGENTS.md: the 60/90 min gate applies to auto-detected nights only A user-entered sleep window is staged with no gates, so it can still yield a stress score from a short window. Co-Authored-By: Claude Opus 5.5 --- AGENTS.md | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/AGENTS.md b/AGENTS.md index e8645d7de..41bc63636 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -159,8 +159,9 @@ oxygen dips; a skin-temp section gated on `spo2` presence; `StageBars` drawing a "—"; a false empty state instead of a retryable error; Bluetooth-off reported as "no strap found". The "confident stress score on ~20 min of data" report (#148) was the `100 - readiness` fallback, removed in `adb51173`: stress now comes only -from the Baevsky SI over a scored night's beats, and a night is scored only from -60 min of sleep (90 min in the daytime). +from the Baevsky SI over a scored night's beats. An auto-detected night needs +60 min of sleep (90 min in the daytime); a user-entered sleep window skips those +gates, so a short logged window can still produce a stress score. **Ask on any metric diff:** what does this return when the input is missing or thin? Anything other than null/"—"/an honest low-confidence envelope is a bug.