Skip to content

Find code by asking what it does. A CLI for coding agents that uses Jev to discover relevant files and source context.

docs: retain failed Django confirmation and reopen quality diagnosis

David Zhang9f4bae00d37b
diff --git a/specs/jevgrep/README.md b/specs/jevgrep/README.md
index e0e7814..709f32f 100644
--- a/specs/jevgrep/README.md
+++ b/specs/jevgrep/README.md
@@ -21,7 +21,9 @@ all ten frozen no-call validations. [Freshness evidence](assets/source-freshness
owns the archive identity, failure reproductions and verification results. Do not
transfer the previous cohort's measured results to this corrected archive.
[Current confirmation](assets/freshness-confirmation.md) records completed cells;
-scikit-learn solved officially but exceeded baseline cost. The choices ledger has
+scikit-learn solved officially but exceeded baseline cost, and Django failed its
+official grade, contradicting acceptance for this cohort. Finish the registered
+attempts and diagnose that failure; do not replace or pool cells. The choices ledger has
been consolidated against current code; final reconciliation and archive remain.
Preserve healthy retrieval semantics and the canonical skill while fixing
diff --git a/specs/jevgrep/assets/freshness-confirmation.md b/specs/jevgrep/assets/freshness-confirmation.md
index 982f1af..ec9c4fd 100644
--- a/specs/jevgrep/assets/freshness-confirmation.md
+++ b/specs/jevgrep/assets/freshness-confirmation.md
@@ -1,6 +1,8 @@
# Corrected-package confirmation
-Status: running. This is a separate prospective ten-task cohort for the
+Status: running; acceptance already contradicted by a lost Django baseline solve.
+Finish and retain the registered cohort rather than replacing that attempt.
+This is a separate prospective ten-task cohort for the
[source-freshness correction](source-freshness.md), not a replacement of individual
cells from the preceding accepted study. The frozen plan, archived runner and
attempt evidence live at `evals/runs/swebench/installed-jg-freshness-v1/`.
@@ -13,6 +15,7 @@ separately and excluded; incomplete observations are subtotals, not totals.
| Task | Official result | Sol | Baseline Sol | Jev observed API cost |
| --- | --- | ---: | ---: | --- |
| scikit-learn__scikit-learn-13124 | solved; not a cost win | $0.3434830 | $0.2944360 | known $0.046611348; total unknown |
+| django__django-15629 | unresolved | $1.0695890 | $1.5055472 | known $0.158282796; total unknown |
Scikit-learn returned five relevant files in 5,038 bytes, implementation reading
leads and two test source blocks. The agent then read the splitter and tests, searched documentation,