Daily Spending Forecast — github/gh-aw
Forecast date: 2026-09-29
Source run: §36551435406
History window: 30 days (workflow-native windows reported in output)
Executive Summary
- Total observed spending from sampled runs (
run_samples[].aic): 418,016.17 AIC across 8,498 runs.
- Weekly P10 (10th percentile — optimistic scenario: 9 out of 10 months will cost at least this much): 129,824.95 AIC.
- Weekly P50 (50th percentile — median/expected scenario: equal probability of spending more or less than this): 172,658.53 AIC.
- Weekly P90 (90th percentile — conservative scenario: only 1 out of 10 months is expected to exceed this amount): 215,412.36 AIC.
- Monthly P10/P50/P90 totals: 556,757.20 / 739,965.29 / 923,196.31 AIC.
Charts
Key Metrics and Issues
- 1761 run samples have zero AIC (20.7% of samples).
- 69 workflows have sparse samples (<5 runs).
- 137 workflows have broad weekly confidence bands ((P90-P10)/P50 > 2.0).
- Top workflow table is sorted by projected weekly P50 AIC and limited to highest-impact entries.
Workflow Forecast Table
| Workflow |
Samples |
Observed AIC |
P50/Run |
P90/Run |
Weekly P50 |
Weekly P90 |
Monthly P50 |
Success Rate |
Confidence Range |
| agentic-token-audit |
120 |
39,534.16 |
284.67 |
578.16 |
12,131.26 |
15,921.02 |
51,990.02 |
99.2% |
8,318.95–15,921.02 |
| Go Linting |
31 |
24,697.66 |
807.43 |
961.11 |
8,377.13 |
9,783.14 |
35,902.00 |
74.2% |
6,946.86–9,783.14 |
| markdownlint |
31 |
8,086.66 |
261.50 |
341.06 |
2,548.84 |
3,117.44 |
10,923.60 |
74.2% |
1,979.86–3,117.44 |
| Add to Projects |
167 |
18,114.72 |
98.55 |
160.66 |
2,448.32 |
3,402.06 |
10,492.81 |
100.0% |
1,479.46–3,402.06 |
| Update Triaged At and Project Date |
168 |
17,214.47 |
96.76 |
128.56 |
2,393.23 |
2,902.78 |
10,256.69 |
100.0% |
1,859.42–2,902.78 |
| Setup Rust and Cargo Caching |
31 |
7,568.27 |
242.97 |
326.90 |
2,366.62 |
2,889.57 |
10,142.67 |
74.2% |
1,837.43–2,889.57 |
| Docs formatting |
93 |
13,268.16 |
120.87 |
206.16 |
2,245.13 |
3,282.17 |
9,621.99 |
100.0% |
1,213.74–3,282.17 |
| Update Issue Stats |
177 |
14,643.66 |
81.90 |
121.64 |
2,048.27 |
2,626.67 |
8,778.29 |
100.0% |
1,477.66–2,626.67 |
| stale |
120 |
13,254.69 |
93.49 |
131.44 |
2,012.00 |
2,687.80 |
8,622.84 |
100.0% |
1,344.61–2,687.80 |
| Push Notifications |
57 |
16,987.10 |
269.56 |
457.74 |
1,887.15 |
3,303.00 |
8,087.80 |
98.2% |
408.00–3,303.00 |
| Auto Label and Categorize New Issues and Pull Requests |
26 |
7,869.98 |
306.97 |
363.75 |
1,865.68 |
2,101.09 |
7,995.77 |
100.0% |
1,628.54–2,101.09 |
| Discussion Labeling |
119 |
8,688.81 |
73.02 |
87.89 |
1,755.28 |
2,040.09 |
7,522.61 |
99.2% |
1,469.28–2,040.09 |
Data Quality and Accuracy Assessment
- Reconciliation: all observed totals are computed directly from nested
workflows[].run_samples[].aic; the top-level run_samples field is absent in this output format, so reconciliation uses per-workflow samples only.
- Zero AIC samples likely represent no-op/early-fail/empty-usage runs; impact: can bias per-run percentiles downward for affected workflows.
- Sparse-sample workflows (<5 runs) have unstable percentile projections; impact: wider uncertainty and less reliable budgeting for low-volume workflows.
- Broad confidence bands indicate materially uncertain forecast tails; impact: P90 budget guardrails should be used for conservative planning.
- Follow-up evidence: prepared output parsed successfully (
exit_code=0) and contained sufficient per-run evidence (8,498 samples), so no rerun was required.
Detailed Per-Workflow Evidence
Top workflows by observed AIC (sample evidence)
| Workflow |
Samples |
Observed AIC |
Weekly P50 |
Weekly P90 |
Monthly P50 |
Success Rate |
| agentic-token-audit |
120 |
39,534.16 |
12,131.26 |
15,921.02 |
51,990.02 |
99.2% |
| Go Linting |
31 |
24,697.66 |
8,377.13 |
9,783.14 |
35,902.00 |
74.2% |
| Add to Projects |
167 |
18,114.72 |
2,448.32 |
3,402.06 |
10,492.81 |
100.0% |
| Update Triaged At and Project Date |
168 |
17,214.47 |
2,393.23 |
2,902.78 |
10,256.69 |
100.0% |
| Push Notifications |
57 |
16,987.10 |
1,887.15 |
3,303.00 |
8,087.80 |
98.2% |
| Open and maintain issue for stale pull requests |
93 |
16,246.09 |
1,641.78 |
2,193.16 |
7,036.20 |
100.0% |
| Update Issue Stats |
177 |
14,643.66 |
2,048.27 |
2,626.67 |
8,778.29 |
100.0% |
| Docs formatting |
93 |
13,268.16 |
2,245.13 |
3,282.17 |
9,621.99 |
100.0% |
| stale |
120 |
13,254.69 |
2,012.00 |
2,687.80 |
8,622.84 |
100.0% |
| Add labels to pull requests from merge queue |
177 |
11,444.93 |
1,605.01 |
2,173.99 |
6,878.61 |
100.0% |
| Summary update |
30 |
11,059.72 |
0.00 |
5,381.94 |
12,103.84 |
46.7% |
| Add and remove labels based on branch naming conventions |
177 |
10,858.76 |
1,523.91 |
2,072.06 |
6,531.04 |
100.0% |
| project-item-copier (self) |
89 |
10,520.52 |
863.88 |
1,638.88 |
6,262.15 |
98.9% |
| markdownlint |
31 |
8,086.66 |
2,548.84 |
3,117.44 |
10,923.60 |
74.2% |
| Auto Label and Categorize New Issues and Pull Requests |
26 |
7,869.98 |
1,865.68 |
2,101.09 |
7,995.77 |
100.0% |
Assumptions and Next Actions
- Assumes prepared
gh aw forecast output captures representative 30-day execution behavior and that workflow mix remains broadly similar next period.
- Uses observed AIC from sampled historical runs as ground truth for historical spend; projected spend uses workflow-provided Monte Carlo percentiles.
- Budgeting recommendation: use weekly/monthly P50 for expected planning and P90 for conservative spend caps where SLA risk is high.
Generated by 📈 Daily Spending Forecast · codex · gpt53codex · 20.7 AIC · ⌖ 3.48 AIC · ⊞ 25.1K · ◷
Daily Spending Forecast — github/gh-aw
Forecast date: 2026-09-29
Source run: §36551435406
History window: 30 days (workflow-native windows reported in output)
Executive Summary
run_samples[].aic): 418,016.17 AIC across 8,498 runs.Charts
Key Metrics and Issues
Workflow Forecast Table
Data Quality and Accuracy Assessment
workflows[].run_samples[].aic; the top-levelrun_samplesfield is absent in this output format, so reconciliation uses per-workflow samples only.exit_code=0) and contained sufficient per-run evidence (8,498 samples), so no rerun was required.Detailed Per-Workflow Evidence
Top workflows by observed AIC (sample evidence)
Assumptions and Next Actions
gh aw forecastoutput captures representative 30-day execution behavior and that workflow mix remains broadly similar next period.