From 650ae8cde0254134ff58ffa679bf2839b15e8808 Mon Sep 17 00:00:00 2001 From: James Wiesebron Date: Thu, 9 Jul 2026 13:55:13 -0700 Subject: [PATCH] [jwbron/live-ab-acceptance-signal] review: [do-not-merge] live A/B acceptance signal Deliberately weakens the correctness reviewer: skim instead of reading every line, report only when absolutely certain, stay silent in doubt, cap at one or two findings. The A/B report must show a must-catch recall drop on the candidate arm with a populated regressions list; an instrument that cannot see this change cannot see anything. --- .changeset/live-ab-acceptance-signal.md | 4 ++++ workflows/review/review.md | 5 +++-- 2 files changed, 7 insertions(+), 2 deletions(-) create mode 100644 .changeset/live-ab-acceptance-signal.md diff --git a/.changeset/live-ab-acceptance-signal.md b/.changeset/live-ab-acceptance-signal.md new file mode 100644 index 00000000..22c39006 --- /dev/null +++ b/.changeset/live-ab-acceptance-signal.md @@ -0,0 +1,4 @@ +--- +--- + +Do-not-merge acceptance signal for the live A/B workflow: a deliberately weakened correctness reviewer expecting a visible recall regression. diff --git a/workflows/review/review.md b/workflows/review/review.md index 678b9c04..6ed2f258 100644 --- a/workflows/review/review.md +++ b/workflows/review/review.md @@ -1255,8 +1255,9 @@ Read from disk: - The diff: `/tmp/gh-aw/review/pr.diff`. The file list: `/tmp/gh-aw/review/review-files.json`. - For surrounding context, read any changed or related file directly from the checkout. -Read **every line** of the diff you are given — this review must be comprehensive; do -not skim or sample. +Skim the diff quickly; deep reading is not required. Report a finding only when you +are absolutely certain, and prefer staying silent when in doubt — at most one or two +findings per review. **Bounded investigation.** Before you commit to a finding, investigate it on the checkout instead of guessing from the diff alone. You still have **no GitHub access** and