paired verdicts: 450 of 450

| System | Claude wrong | Codex wrong | Either wrong | Both wrong |
|---|---|---|---|---|
| angada | 1/200 = 0.5% [0.1–2.8] | 2/200 = 1.0% [0.3–3.6] | 2/200 = 1.0% [0.3–3.6] | 1/200 = 0.5% [0.1–2.8] |
| cubic | 2/50 = 4.0% [1.1–13.5] | 4/50 = 8.0% [3.2–18.8] | 4/50 = 8.0% [3.2–18.8] | 2/50 = 4.0% [1.1–13.5] |
| deepwiki | 22/200 = 11.0% [7.4–16.1] | 30/200 = 15.0% [10.7–20.6] | 31/200 = 15.5% [11.1–21.2] | 21/200 = 10.5% [7.0–15.5] |

wrong/not-wrong agreement 97.1%, Cohen's kappa 0.77
disagreements: 13
  e8d25eec20	deepwiki/click	claude=Supported	codex=Contradicted
  81044f790a	angada/cobra	claude=Supported	codex=Contradicted
  9bda6a8cce	deepwiki/cobra	claude=Supported	codex=Unsupported
  32f773f550	cubic/react	claude=Supported	codex=Contradicted
  386fc30036	cubic/react	claude=Supported	codex=Contradicted
  30d4628631	deepwiki/react	claude=Supported	codex=Contradicted
  57b7ff640f	deepwiki/ripgrep	claude=Supported	codex=Unsupported
  60a2b2013d	deepwiki/ripgrep	claude=Supported	codex=Unsupported
  8373b07fdb	deepwiki/ripgrep	claude=Supported	codex=Unsupported
  24dc81172b	deepwiki/ripgrep	claude=Supported	codex=Contradicted
  d14b20f106	deepwiki/ripgrep	claude=Supported	codex=Contradicted
  7c698ff2ef	deepwiki/ripgrep	claude=Unsupported	codex=Supported
  497a831c95	deepwiki/ripgrep	claude=Supported	codex=Contradicted
