Impact Efficiency Report - 2026-06-16 #39551

Executive Summary

In the 180-day window (2025-12-18–2026-06-16), the highest-impact work was two bug fixes for core sample-replay failures (PR #37863, PR #37565), delivering 170/205 total objective value (83%) in the bug/compiler category. A third merged outcome—PR #38185 (linter-miner)—added 35 units under improvement. AIC data is unavailable (workflow-logs.json: 0 runs recorded); Impact Efficiency cannot be computed. Four rejected documentation outcomes from Daily Documentation Healer/Updater contributed zero accepted value, while 993 of 1,000 analyzed PRs (99.3%) lacked a traceable Closes #N link.

Summary

Agentic Work by Objective

Bug (value: 85) — 2 accepted outcomes, total value: 170, AIC: 0, IE: n/a

• PR #37863 → root #37835 (bug, compiler)
• PR #37565 → root #37545 (bug, compiler)

Improvement (value: 35) — 1 accepted outcome, total value: 35, AIC: 0, IE: n/a

• PR #38185 → root #38143 (improvement, quick-win)

Highest AIC-consuming category: cannot determine (no AIC data).

Which Workflows Drove That Impact

Unattributed (not ranked): 2 accepted outcomes, value 170 (82.9%) — PR #37863, PR #37565. Only PR Description Updater footer found in bodies; creating workflow unknown.

• Most total value (named workflow): Linter Miner — PR #38185 (improvement, 35 pts)
• Largest share of delivered value: Unattributed bucket (170/205 = 82.9%)
• Most AIC: cannot determine

Top Outcomes by Outcome Value

All 7 in-scope outcomes listed.

Unmapped Outcomes

None. All 7 analyzed outcomes had root issue labels matching the objective mapping.

Interpretation

Accepted outcome count vs Impact Efficiency: 3 of 7 in-scope outcomes were accepted (43%). Bug fixes drove 83% of value (170/205). Impact Efficiency cannot be computed (AIC = 0 from deterministic source).

Key findings:

• Bug category delivered highest per-outcome value (85 pts each): Two compiler/sample-replay fixes dominate the window.
• 82.9% of delivered value is unattributed: The most impactful outcomes have no confirmed creating workflow.
• Documentation workflows delivered 0 accepted value: 4 PRs created, all rejected. These consume AIC (unquantified) with no measurable return.
• Low linked-issue rate (0.7%): The primary data-quality gap. 993 PRs excluded for lack of Closes #N.

Data Quality

PR root tracing: 993/1,000 PRs (99.3%) excluded for lacking Closes #N. Highest-priority fix.

Safe-output coverage: 0 safe-output outcomes recorded (safe-output-issue-summary.json: all zeros).

Workflow attribution: 2/3 accepted outcomes are unattributed (66.7%; 170 value pts). PR #38185 uses title-inferred attribution only. workflow-logs.json had 0 runs — attribution relies entirely on PR body metadata.

AI Credits: Deterministic precomputed source (workflow-logs.json: 0 runs). Total AIC = 0. Impact Efficiency undefined. No live fallback invoked. AIC values in PR bodies (e.g., "389.9 AIC" from PR Description Updater) belong to secondary workflows and were excluded.

Human Work

Because workflow-logs.json has 0 runs, all 500 merged PRs technically qualify (no matching run). Count is inflated by empty log data — ~131 workflow-attribution comments exist in merged PR bodies, indicating many are workflow-generated.

By objective (linked-issue merged PRs):

Bug (value 85): PR #37863 (→ #37835), PR #37565 (→ #37545) — total value: 170

Improvement (value 35): PR #38185 (→ #38143) — likely workflow-generated ([linter-miner] prefix); appears here due to empty logs — total value: 35

Generated by 📊 Impact Efficiency Report · 852.9 AIC · ⌖ 14.5 AIC · ⊞ 22.5K · ◷