research: add GitHub platform usage metrics as paper dimension — commits, Copilot usage, AI cost vs. throughput

[ccisnedev]

Context

The research paper (bootstrap-validation.md) currently tracks per-cycle metrics via metrics.yaml (#72). This captures the micro view: individual APE cycles, delta failures, test deltas, etc.

What's missing is the macro view — platform-level usage data from the author's GitHub account that demonstrates the aggregate impact of using Inquiry as a methodology.

Proposed dimension

Add a new research dimension that tracks GitHub platform usage reports across all repositories and organizations:

1. Commit frequency over time

• Total commits per month across all repos and orgs (before vs. after Inquiry adoption)
• Commit distribution by repo/org
• Active days per month

2. GitHub Copilot usage metrics

• Copilot completions accepted/rejected over time
• Copilot Chat usage frequency
• Lines of code suggested vs. accepted

3. AI cost vs. project throughput

• Monthly AI cost (Copilot subscription, API usage if any)
• Number of projects advanced per month
• Issues closed per month across all repos
• Cost-per-issue or cost-per-commit as efficiency metrics

Data sources

• GitHub contribution graph API — commit counts, active days
• GitHub Copilot usage API / billing reports — usage metrics, costs
• GitHub billing page — monthly AI spend
• gh CLI — cross-repo issue/PR counts

Relationship to existing research

• Extends bootstrap-validation.md §4 (Metrics to Graph) with platform-level dimensions
• Complements per-cycle metrics.yaml (feat: implement metrics.yaml collection for research paper data #72) with aggregate longitudinal data
• Strengthens the paper's empirical claims beyond N=1-per-cycle to N=1-over-time trends
• Addresses the "no control group" threat (§6.3 in bootstrap-validation) by showing before/after adoption curves

Acceptance criteria

• Define schema for platform-level metrics (monthly granularity)
• Document data extraction process (API calls, billing exports)
• Collect baseline data (pre-Inquiry period) for comparison
• Collect post-Inquiry adoption data
• Add section to bootstrap-validation.md §4 for platform metrics
• Create at least one visualization comparing before/after trends
• Document AI cost-efficiency analysis (cost vs. throughput)

Labels

research, paper, metrics