Figure sources

Back to the letter.

This register covers every illustration in the article. Figure filenames refer to the accompanying images; links lead to their published data and sources. The source URLs, local file digests, and pixel comparisons were checked on September 16, 2026.

Figure file	Source and preparation
`opening.png`	AAV manuscript, f. 70r; color crop
`page-70r.jpg`, `page-70v.jpg`	AAV manuscript, ff. 70r–70v; unaltered public DECODE previews
`bigrams.png`	Author's diagram of counts from EHum's 2016 transcription; data and counting rules
`meister-179.png`	Meister (1906), printed p. 179, key no. 7; crop of the Internet Archive scan
`key.png`	Author's diagram of the recovered key; assignments
`dots.png`	Author's diagram of the recovered dotted codes; assignments
`ambiguity.png`	Author's diagram of two segmentations of `057805`; transcription locations
`montepulciano.png`	AAV manuscript, f. 70r; color crop
`neighbors.png`	Processed glyphs from AAV manuscript photographs; dataset, model, and exact locations
`anco.png`	AAV manuscript, f. 71v; color crop
`che-vi-va.png`	AAV manuscript, f. 70v; color crop

Source credits and access

Manuscript: Archivio Apostolico Vaticano (AAV), Segreteria di Stato, Spagna 1A, ff. 70r–73v. Manuscript photographs: distributed through DECODE record 92, record name ASV_i1025_SdS_Spain_IA-2. The archive was formerly called Archivio Segreto Vaticano (ASV).

DECODE lists the record's access mode as “Authentication required” and its “Private Ciphertext” field as True. The full-resolution photographs are restricted research inputs, obtained on September 15, 2026. An unauthenticated request to the full-image URL for 70r returned a 986 × 568 PNG rather than the manuscript JPEG. The public previews below were accessible without authentication and are separate files.

The analytical diagrams credit the article's author; their underlying transcription credits EHum. The Meister scan credits the printed work and the Getty Research Institute copy digitized by Internet Archive, as detailed below. No image-specific reuse license or reproduction grant is documented in the source records checked for this register.

Public page previews

These are DECODE's public, low-resolution JPEG previews, fetched without authentication on September 16, 2026. The files preserve the response bytes, without resizing or re-encoding.

Local file	Folio	Dimensions	Bytes	Exact source URL
`page-70r.jpg`	70r	200 × 256	19,870	https://de-crypt.org/decrypt-custom/filesrv/?file=TH_IMG_R92_I691_P1.jpg
`page-70v.jpg`	70v	200 × 256	20,232	https://de-crypt.org/decrypt-custom/filesrv/?file=TH_IMG_R92_I692_P2.jpg

SHA-256:

05e495a7c4462cc18353b7a804481a5721764c20a167319936766621df5d787d  page-70r.jpg
fa43b7675433420ed8379fe20a943ca31cdc84f0e6aec76a8f5e25a908682066  page-70v.jpg

Color details

These small details come from the full-resolution photographs, not the public previews. The full photographs are not reproduced in this supplement. Each original is a 2833 × 3628 RGB JPEG.

Bounds are (left, top, right, bottom) in original pixels, measured from the upper-left corner. The right and bottom bounds are exclusive. Each detail crops the original RGB pixels, then enlarges the crop exactly 2× with Pillow's Lanczos filter. There are no other changes to the color, contrast, strokes, or page geometry. The crops contain no added labels or boxes.

File	Original photograph	Folio and cipher line	Crop bounds	Original crop	Output
`opening.png`	`IMG_R92_I691_P1.jpg`	70r, line 1 (cleartext-to-cipher transition)	`(1060, 2730, 2115, 2890)`	1055 × 160	2110 × 320
`montepulciano.png`	`IMG_R92_I691_P1.jpg`	70r, line 2	`(800, 2880, 1280, 3010)`	480 × 130	960 × 260
`anco.png`	`IMG_R92_I694_P4.jpg`	71v, line 4	`(1220, 844, 1775, 986)`	555 × 142	1110 × 284
`che-vi-va.png`	`IMG_R92_I692_P2.jpg`	70v, line 3	`(1575, 770, 2218, 921)`	643 × 151	1286 × 302

Cipher lines and token positions are one-based and follow the public transcription. Cleartext lines do not count as cipher lines. The recorded line centers are y = 2808, 2946, 919, and 854, respectively. The crops include fragments of neighboring ink at their edges.

Opening: the end of the cleartext, including il seguito, and the first ten cipher digits on the same physical line. The digits are 7 3 4 5 7 4 9 0 3 5, grouped under the recovered key as 73 4 57 4 9 03 5: dopo la, with 9 as the null.
Montepulciano: centered on positions 5–14, with 5 7 1 0 3 2 7 at positions 7–13. The public transcription has 2 at position 9. The page has the hooked 1, which supplies u, rather than r. This detail supports that digit correction. The name's final -o is outside the crop and remains supplied in the edition.
Anco: covers the passage at public positions 22–40. The checked digit sequence is 7^. 2 8 5 7 2 7 4 5 5 7 2 4 5, with punctuation omitted here. The page has fourteen digits and one 5 7 2 7 4 group. The public transcription has nineteen tokens and repeats that group. Its extra tokens prevent a one-to-one mapping between every token position and visible glyph.
Che vi va: covers positions 33–47, extended beyond position 44 to include the end of dentro. The checked sequence is 2 7^. 1 6 9 1 5 7 3 8 7 8 0 2 4. Grouped by cipher codes, it is 27^. 1 6 9 1 5 73 8 7 80 2 4. The dot belongs to the 7 at position 34. The hooked forms at positions 35 and 38 read 1, against the public transcription's 2. The 9 is a null. The passage reads che vi va dentro.

These readings follow the manuscript check. They do not replace the public readings in the ciphertext appendix.

Original photograph URLs and checksums

These are the full-resolution source locations in DECODE record 92, distinct from the public preview URLs above. The table includes all seven cipher-bearing photographs used to build the glyph dataset.

Folio	Source URL	Original SHA-256
70r	IMG_R92_I691_P1.jpg	`766cc8ed20ff91e55c76733ec9b3d2068fa1f8d3d79de35eb52b79387c006fb2`
70v	IMG_R92_I692_P2.jpg	`ed7e2268c8f7e6d386accb9e6f006d3e8e82db5c1b962ef4352d1fc3ad20bff8`
71r	IMG_R92_I693_P3.jpg	`f0c9ad564fcd03db5e1f5af6d9117336c39fc261046b648d577165b92bc116a6`
71v	IMG_R92_I694_P4.jpg	`55507ca5d60dcdfa23d0f324d0a9c73f2bc72d89295888934951b6e6318b3727`
72r	IMG_R92_I695_P5.jpg	`bb99c72b39388fa8be09fb9e54768449e25f7a8fdfbc088dd08eb5707eb40168`
72v	IMG_R92_I696_P6.jpg	`5bb85210674dc796a21dd30bc175dacfb1e290aa020166ad8471aea97f7cf17b`
73r	IMG_R92_I697_P7.jpg	`284aebc72e086afeb18733dbd1efa377a9861d616ba6d630f7411142feca7d8f`

Nearest-neighbor strip

neighbors.png shows the actual normalized dataset patch for 70r, cipher line 2, position 9, then its seven nearest neighbors in distance order. The digit below each patch is its public-transcription label, not an independently established truth. The target's label is 2. The neighbor labels are 1 1 1 1 1 6 1.

The source is the author's normalized glyph dataset, derived from the seven photographs listed above. Processing maps grayscale levels 60–200 to ink strength, levels the baseline, removes slant with a shear, and aligns tokens to glyph intervals. It removes small detached components, fits the ink within 24 × 24 pixels, and places its center of mass within a 32 × 32 patch. Thus, these are processed classifier inputs, unlike the color details above. The figure inverts each stored patch to black ink on white, then enlarges it 3× with nearest-neighbor sampling. Every patch has the same 96 × 96 display size. The complete strip is 864 × 158 pixels. Preparation used Python 3.14.7, NumPy 2.5.3, and Pillow 12.3.0. The labels use JetBrains Mono Nerd Font; the font file's SHA-256 is 1c680e8cde9fcf8b88a5605ce8d1fb94dd3fb15841f7ca7bf4c55664855e5611.

The classifier uses k = 7, 64 principal components, and geometry weight 2.0. This is a k-nearest-neighbor (kNN) model. Principal component analysis (PCA) uses all 6,497 labeled, non-phantom glyphs. The five geometry features measure width, height, top, bottom, and ink area, in units of the 45-pixel x-height. The embedding concatenates the 64 PCA components with the five geometry features multiplied by 2.0. The distance is squared Euclidean distance in that embedding. The query excludes every reference with the target's line-group ID, 1. It does not manually select examples or filter them by digit.

For squared distance d, the vote weight is exp(-d / max(d_nearest, 1e-6)), normalized across the seven neighbors. The current dataset gives vote shares 0.899233 for 1 and 0.100767 for 6, with zero for 2. These shares describe the model's vote, not a calibrated probability that a digit is correct.

Order	Position	Dataset row, zero-based	Line group	Label	Squared distance	Normalized weight
Target	070r L02 p09	33	1	2	—	—
1	073r L02 p06	5715	98	1	15.994228	0.276510
2	072r L05 p23	4118	71	1	23.278652	0.175354
3	071v L11 p40	3311	56	1	27.764275	0.132470
4	073r L13 p47	6405	109	1	30.863949	0.109132
5	070v L04 p25	373	7	1	30.882706	0.109004
6	071r L17 p59	2431	41	6	32.139412	0.100767
7	073r L03 p27	5795	99	1	32.787823	0.096763

The target's aligned interval is x = 396–439 in the processed line strip. Those coordinates are not bounds in the original photograph. The table records the model output for the exact dataset below. Floating-point results can differ slightly between numerical libraries.

Processing checks

The normalized glyph dataset has SHA-256 64920ca50db2ed71b9fd1145e92b0a2f6c691ca64683fde4532fa95ada67c875. It is derived from the restricted photographs and is not reproduced in full here. The color crops match the specified crop-and-enlargement operation pixel for pixel. All eight displayed neighbor patches match the recorded dataset rows after inversion and enlargement. The transcription supplying the provisional labels is included as the complete public text, SHA-256 5402202eebefdb8fb967c27491255e128bcbfec7bfa1d04fa45b8a2d300cec45.

Analytical figures

bigrams.png, key.png, dots.png, and ambiguity.png are original SVG diagrams rendered as PNG at 2× scale with librsvg. The font lists are Noto Serif, Liberation Serif, serif for prose and digits, and Noto Sans, Liberation Sans, sans-serif for axes and counts. Fonts and renderer versions can affect the resulting PNG bytes.

The diagrams use the counts, key assignments, and example strings documented below.

Digit-pair counts

bigrams.png (1256 × 1068 pixels) visualizes EHum's transcription of ff. 70r–73r, dated January 9, 2016: Complete transcription, publicly distributed by MysteryTwister. The downloaded text matches the SHA-256 recorded above; the ciphertext appendix presents the parsed cipher portion.

The pair-count table and digit-frequency table supply all values drawn in the figure. The counting rules are:

Take each token's first digit reading, including doubtful readings and the first alternative of a/b. Normalize transcription o to digit 0.
Ignore above-marks and following punctuation for these counts. Retain ? as an unreadable position.
Join the cipher lines and folios into one stream, omitting cleartext. Adjacent-pair counts therefore include pairs across line and folio boundaries and across omitted cleartext passages.
Count overlapping two-position windows, excluding any window containing ?. All 100 table cells match, totaling 6,492 windows.
Compute each digit's share with 6,577 tokens, including the 44 unreadable positions, as the denominator, and round to one decimal place. All ten displayed shares match.

Rows give the first digit and columns the second. Color interpolates from white to #9b2d20 in proportion to sqrt(count / 349). Dashed outlines mark doubled digits. Bold type marks 80, 57, 27, 03, and 73; identifying these as t, p, c, l, and d uses the subsequently recovered key.

Recovered key and dotted codes

key.png (1328 × 460 pixels) and dots.png (1044 × 832 pixels) present the author's recovered assignments for this letter. They are typeset diagrams of the analysis, not images of a surviving historical key. The assignments agree with the key tables and the machine-readable key.

key.png shows eighteen letter assignments, 81 = et, and 9 = null. The diagram displays 1 as u v, making the shared code explicit.

dots.png shows these assignments:

Pair	Undotted	Dot on first digit	Dot on second digit
`27`	c	qua	che
`72`	b	que	chi
`57`	p	qui	non

The three additional title codes are 0̇5 = S.S.tà, 40̇ = S.M.tà, and 6̇0 = V.S.. The title abbreviations and the labels “the Pope,” “the Emperor,” and “Poggio” express the editorial interpretation of their referents. In the machine-readable key, ce and ci represent che and chi, while #, &, %, and _ represent the three titles and the null. The letters and null came from the search; et and the dotted codes were identified from their decoded contexts, as the article explains.

Ambiguous segmentation

ambiguity.png (848 × 380 pixels) typesets the digits 0 5 7 8 0 5. Under the recovered key, its two drawn segmentations are 0 | 5 | 7 | 80 | 5 = santa and 0 | 57 | 8 | 0 | 5 = spesa. The brackets, letters, and English glosses “holy” and “expense” are explanatory additions by the author.

The exact six-digit sequence occurs four times in the public transcription:

Folio	Cipher line and token positions
70v	L11 p29–34
70v	L16 p60 through L17 p03
71r	L09 p42–47
71r	L12 p14–19

The diagram illustrates the shared digit sequence rather than reproducing any one manuscript crop.

Meister scan

meister-179.png reproduces part of Aloys Meister, Die Geheimschrift im Dienste der päpstlichen Kurie von ihren Anfängen bis zum Ende des XVI. Jahrhunderts (Paderborn: Ferdinand Schöningh, 1906), printed p. 179, key no. 7, “Cifra con il signior Alessandro Vitelli.” The heading includes Meister's editorial date [1546.]; his supporting discussion is in footnote 2 on that page. This is a structural comparison for the Farnese letter's cipher, not the letter's recovered key.

Digital copy: Internet Archive, diegeheimschrift00meis, leaf n195. The item metadata identifies the Getty Research Institute as contributor and sponsor, call number 153146.
Exact PDF source: https://archive.org/download/diegeheimschrift00meis/diegeheimschrift00meis.pdf (34,851,433 bytes). Its MD5, d32f60b878497449b3d3179de7857ba1, matches Internet Archive's file metadata; SHA-256 is 5d38b02e1dfd75803fbe645dd70e73ba77ec62fee9dc9a0955522732d37d6c90.
Page mapping: printed p. 179 is PDF page 195 (one-based) in this file and Internet Archive leaf 195 (n195). The scan's leaf numbers and PDF page ordinals use different starting points; this mapping is specific to this file.
Transformation: render PDF page 195 with pdftoppm at 300 dpi, giving 1659 × 2663 RGB pixels, then crop bounds (180, 835, 1360, 1425), with right and bottom exclusive. The resulting 1180 × 590 crop matches the figure's decoded RGB pixels exactly with Poppler 26.08.0. It has no further resizing, color adjustment, or added annotation.
Extent: the crop includes the key number, bracketed date, heading, alphabet, null, and seven word/syllable entries. It omits the following abbreviation instruction, its examples, and the page's footnotes.

The figure is a crop of Meister's printed edition. For the collection containing this key, Meister gives the historical manuscript reference “Rom, Bibl. Chigi M II 49 f. 85–148 und f. 168” at the section opening on printed p. 176. That is his collection-level reference, not a verified folio number for key no. 7.

The exact crop coordinates above were recovered by pixel comparison with the source PDF. The PNG has encoding metadata in addition to its pixels, so a fresh PNG encoding can have a different file digest even when its pixels match.

Image checksums

SHA-256 of the current PNG files. The two unaltered preview JPEG digests are listed under Public page previews.

33962f6052b8ff94e20e7e896238219c97d70ce3eb98094c2b1f2c36673e73ed  ambiguity.png
b333ef0412930601953ca7d84b880ae062f7d8d452979353b77d7ccd77655761  bigrams.png
3b7389f0f5adcfabcd97cbc21feee63b72690dc50413bb0212ba6ab0f12a4e4b  dots.png
5bb54f3869827ace1a1f30b02583d70a937b1ce10d9f30df0e1c92295812c2e1  key.png
c037295290368084045d80152d8318a4f878399234cf91272c3e0193e9fa82f6  meister-179.png
195cc90a4221001b6dde547f898a9de30f1c4b01cadb68a2ca3468f038281d57  opening.png
f1428856a8e1ddeec83e7b246206f8611acf55213a68f50ae63a76dd4d2c69de  montepulciano.png
e5d046cfeeee1776e74bb5e3032fa3e59a8a82bd439f8108b6c1ee813dfcc129  anco.png
c0eb4ce911db15cf239cdcdefcafec240874b114d1428f2936338a2e5eefdd7c  che-vi-va.png
9d8ee49e14022b96b85b6198c37d82a4833e9f844434dfec2e5602d5c6e09330  neighbors.png