$ paw-test compare alpha.paw beta.paw suite.yaml
exit code: 0
--- stdout ---
Differences (2/3):

  Input: today
    alpha -> 2026-09-11
    beta -> "2026-09-11"
    pass differs: alpha=True beta=False
    expected differs: want 2026-09-11 -- alpha=True beta=False
  Input: new year
    alpha -> 2026-01-01
    beta -> "2026-01-01"
    pass differs: alpha=True beta=False
    expected differs: want 2026-01-01 -- alpha=True beta=False

Whitespace- or quoting-only differences (1/3): same value once parsed as JSON (or, if either side isn't JSON, once
Unicode-normalized and whitespace-collapsed), or one side is the other as a JSON string scalar -- not byte-identical,
but not a real disagreement either. Collapsed to input only:

  February 30th (pass differs: alpha=True beta=False -- same value, but only one side satisfies the suite's assertions)

Correct against expected: alpha 2/2 (100.0%) -- all 2 scored; beta 0/2 (0.0%) -- all 2 scored
  2 case(s) where exactly one adapter matched the answer key -- listed under Differences above.

Summary: 3 cases, 0 identical output (byte-for-byte), 0 equivalent output (byte-for-byte + JSON/whitespace-normalized),
3 equivalent output once unwrapped (also counting a JSON-string-quoted/bare pair), alpha pass 3/3, beta pass 0/3,
only-alpha-pass 3, only-beta-pass 0, errored 0/3
