Gitti’s 1878 edition and our transcriptions

Page dated 12 October 2026.

In one paragraph

The 1494 print is the only source for the reading. Five machines read it without any modern edition. Gitti's 1878 edition was then used as a check: his matching words were shown beside most of the places where the first machines disagreed, and he is cited in 29 of the 77 changes made at the second stage. It did not supply the reading, and the translations (modern Italian and English) were made from our own text without it. Two other models (GPT-6-Sol and Gemini) re-read the doubtful places without sight of the first decisions.

Where Gitti was and was not used

What his work did, from least to most influential

1. Confirmed a reading the models already shared (12 of the 29 DRAFT 2 changes; for example denançe). 2. Decided a mark-only difference by the Gitti-fit rule: the abbreviation sign was chosen to match his expanded word. This affects spelling, not meaning. 3. Supplied the word where the type is damaged or unclear and the image could not settle it (for example monete, coperta). 4. Broke a tie where all machines were wrong or all differed (for example nominarle). These are the places where he did most. A further 7 small points moved the text off a reading both models shared (see the sorting below).

The 29 DRAFT 2 changes that cite Gitti, sorted

Method: for each change, compare the DRAFT 2 reading with what Codex and Gemini each read (blind, without Gitti) and with the first pass. The recorded basis in every case is the image and Gitti, so this shows where his word agreed with ours, not that he alone decided.

A (12): oꝑatione (p434 l30), Maxĩe (p436 l36), ſeꝑata (p440 l30), ſopᵃ (p447 l10), qualch̃ ãico (p455 l13), oꝑa (p434 l13), poter (p434 l54), ℥ (p435 l20), uie (p437 l42), denançe (p442 l55), lĩa (p443 l04), ĩpoꝛtano (p455 l48).

B (10): ſeꝑata dala (p447 l09), noĩarle (p435 l52, all three differed), haᵉ ala (p435 l58), d̃ ĩp̃ſti (p435 l58), p̃mamẽte (p436 l55), copta (p437 l37, all differed), ĩla ĩmediate (p441 l40), facilmẽte (p448 l13), ꝑ te o ꝑ notaro (p450 l14, all differed), Bruça (p455 l12).

C (7), all small points: vᵒ for vᵃ (p435 l39), m{o}nete (p437 l30, a letter not inked in the print), uerebono for uerẽbono (p450 l52), ꝑtitᵉ for ptitᵉ (p451 l40), lihaueſſe for libaueſſe (p452 l04, Gothic h like b), suꝑflua for sup̃flua (p453 l03), eldi for el di (p455 l03).

So the picture is: in 12 of 29 he agreed with both models; in 10 he supported one reading among differing ones; in 7 he moved the text off a reading both models shared, always on a mark, letter or spacing, never on a whole word. Those 7 are listed for your judgement. I sorted them from the logged readings, not by looking at the images again.

What the whole-text comparison is for

The word-by-word comparison with Gitti was run as a no-fault check, not as a source of the reading.

The result: in the first draft, 2,601 of 3,034 aligned abbreviated words (85.7%) were consistent with his expanded forms. That is a check passing on that set of words. It is not a measure of the whole text, and it does not mean we borrowed from him.

The remaining 14% are reported, not acted on. They are on the unexplained list, and no text was changed because of them. They have four possible causes that I cannot yet tell apart: alignment noise, his modernising, a change he made silently (he did this in Ch. 9 and Ch. 17), or a misreading by us.

Agreement shows there is no gross error. It does not prove a reading right, because two sources can share a mistake, and he did not mark his own expansions.

The 433 unexplained words, checked against the page images (11 Oct 2026)

Method: line crops rebuilt with each line matched to its OCR line by text (the earlier crops used line numbers and were wrong on many pages). Six further Claude instances each looked at the crop for every word, with Gitti's text to search. Two lines were spot-checked by the coordinator; the rest are the machines' readings.

Separate check, 19 "partita without a bar" spellings: all 19 were misread by us. The print has a p with a crossed descender, so the word is ꝑtita/ꝑtite (partita/partite). The meaning was already right; only the sign was missing.

So of the 433 words, 4 were our errors (about 1%), plus the 19 partita signs. The third draft of the reading corrected 22 of these 23 (the 23rd was already right).

Gitti and the 1523 reprint

Gitti's own preface says he took the 1494 as his base and revised it "word by word" against the 1523 reprint, using the copies in the Biblioteca Marciana, and that the two editions are "in everything and for everything" identical. He also thanks two students of the Venice commercial school who copied the treatise and compared the proofs against both editions. So he was not working from the 1494 alone, and he is not independent of the 1523. His claim of identity is too strong: our comparison found many differences between the two printings (see the page on the 1523 reprint).

A place where Gitti and our first translation were both wrong

At p435 L58 the print has ala cameᵃ d̃ ĩp̃ſti. Gitti reads "alla Camera denari imprestati" (money lent), and our first translation followed him, taking d̃ as denari. Anthony Mellor, reading as an accountant, saw that "I find I have lent" made no sense. d̃ is de: it is the Camera d'imprestiti, the Venetian public loan office, which Pacioli names again later ("la Camera d'imprestito o altro monte"). So the entry records a holding in the public loan fund, an asset. This is corrected in the Italian and English drafts. A later check of the other lines where d̃ stands alone found one more (p457 L46, "de ducati 240"), also corrected.

Limits

• The 85.7% figure measures agreement on one set of aligned abbreviated words in the first draft, not dependence and not the whole text.
• Gitti's text is a human edition with its own doubts and silent changes; it is not treated as an authority.
• Machine output; no person has proofread it against the scan.

Appendix A: second-draft changes that cite Gitti (29, as recorded at the time; some readings were changed later)

Appendix B: first-draft decision notes that mention Gitti (40, as recorded at the time; some readings were changed later)