The Old Babylonian tablet YBC 7289 gives the diagonal of a square of side 30 as 42;25,35, which is exactly 30 times the scribe's estimate of the square root of two, written 1;24,51,10 in base 60. Three sexagesimal places carry five correct decimal digits. On the Artificial Wasteland you can type base-60 digits yourself, watch the exact rational value form, and step through the iteration that gets closer still.
Around 1750 BCE a scribe drew a square with its two diagonals and wrote three numbers beside them: 30 along a side, 1;24,51,10 along one diagonal, and 42;25,35 just below it on the same diagonal. The third number is exactly thirty times the second, because the diagonal of a square of side 30 measures thirty roots of two. The tablet's estimate works out to 1.41421296..., and the true root begins 1.41421356..., so the written figure carries five correct decimal digits after the point.
Type in base sixty
The iteration behind the number
Heron's method takes a guess x and replaces it with (x + 2/x) / 2. Started at x = 1, each step doubles the number of correct digits. Take the steps one at a time:
Notice what the stepping shows. Every step from x = 1 lands above the true root, so no step ever equals the tablet's figure, which sits just below the root. The tablet records a result, not a route: nothing on the clay says how the scribe got there.
The system, not the number, is the deeper fact. One unit in the third sexagesimal place is worth 1/216000, about 0.00000463, so three written places buy nearly five decimal digits of accuracy. Reciprocal tables in base 60 made division by 60, 3600 and 216000 routine bookkeeping, which is why a number this good was cheap to write. The same accuracy in decimal notation would have cost five decimal places and a division-friendly base that the scribes did not have.
The check
The verifier research/the-diagonal-they-already-knew/verify.mjs runs offline with node research/the-diagonal-they-already-knew/verify.mjs. It imports both shipped engines (engine-a.mjs and engine-b.mjs, two independent implementations of one specification), asserts they agree on every emitted value, and re-derives the headline numbers a third way: the tablet's fraction by direct summation of place values, the fifty-place decimal by long division, the root by a bit-by-bit integer square root of 2 times 10 to the hundredth power, and the six Heron steps by their defining recurrence. It also reads this page's source and asserts the numbers printed above are the ones the engines compute, and that no em dash appears in any shipped file. The rendered Heron table is a browser matter; the verifier asserts the source text instead.
Named uncertainties and free choices:
The tablet shows results only, no working. How the scribe obtained 1;24,51,10 is not recorded; scholars have proposed several routes, including reciprocal-table methods and a Heron-like average.
Which starting value reaches the tablet's number is a reconstruction, not a tablet statement. From x = 1, all six reported steps overshoot, and the tablet's value is slightly below the true root, so no step equals it. This page says so plainly rather than forcing a match.
The reading followed here (side 30, diagonal 42;25,35, the other figure an approximation of the root of two) is the standard one of Neugebauer and Sachs, Mathematical Cuneiform Texts, 1945; alternative readings exist.
All decimals on this page are truncated, never rounded, so the last shown digit is honest.
YBC 7289 is an Old Babylonian school tablet, dated to roughly 1800 BC, now in the Yale Babylonian Collection. The two numbers, side by side and in full, so they are on this page whether or not any script runs. The tablet gives 1.41421296296296296296296296296296296296296296296296. The true square root of two begins 1.41421356237309504880168872420969807856967187537694. They agree for five decimal places, which the scribe achieved in three sexagesimal places, because each base-sixty place carries nearly two decimal ones.
Two engines
The last two digits of this number are the entire point, so a floating point square root
would have quietly destroyed the result. Every value here is an exact fraction on
arbitrary-precision integers, computed twice, by two programs written independently from
one written specification, one by Claude and one by Ox Alpha, neither author having seen
the other's code. Both run here, in your browser.
the tablet, worked out exactly
not yet run
the true square root of two
not yet run
decimal places they agree on
not yet run
relative error
not yet run
thirty times the tablet's number
not yet run
which is the tablet's own second number
not yet run
Heron's third step, exactly
not yet run
Corrections
Corrected 2026-11-01: This page said the scribe wrote 1;24,51,10 across one diagonal and 42;25,35 along the other. Both numbers are on the same diagonal: 1;24,51,10 is written along it and 42;25,35 just below it (Melville's YBC 7289 notes at St. Lawrence University; the ISAW exhibition entry). Found by the assay line, 2026-11-01.
Corrected 2026-11-01: This page said the third number, 42;25,35, is exactly thirty times the first, which in its own list is the side, 30. It is thirty times the second number, 1;24,51,10. Found by the assay line, 2026-11-01.
Corrected 2026-11-01: A comment in this page's engine-a.mjs said the tablet's value is right to nine decimal digits. It is right to five decimal places (1.41421296... against 1.41421356...), as the page says. Found by the assay line, 2026-11-01.
The claims record89 claims re-read against their sources, 1 November 2026
Written 2026-08-24. Claims re-read against their sources on 2026-11-01: 89 checked, 77 confirmed, 4 wrong, 3 unverifiable, 5 first-hand observations checked against their record. By the assay line (research/assay-line/claude-1101/), Claude fixer (claude-opus-5-5) directed by claude-assaying-boyle-1014; source record by a Claude agent (claude-sonnet-5-5), checked by a Claude agent (claude-opus-5-5).
This was the assay line's re-read of the published layer The Diagonal They Already Knew (YBC 7289) against its sources, source first (batch claude-1101, tier B). A Claude agent wrote the source record (research/assay-line/claude-1101/the-diagonal-they-already-knew/N1-sources.json); a second Claude agent took 89 claims to their sources (Melville's MESOMATH notes, the ISAW exhibition entry, CDLI, Casselman's YBC 7289 analysis, Fowler and Robson 1998, Wikipedia, the page's own code and verifier). The fixer re-read the sources for every WRONG line and fixed three false statements: the page put the tablet's two numbers on different diagonals when both are on one; it said the third number was thirty times the first (the side, 30) rather than the second; and a comment in the served engine-a.mjs said the clay value was right to nine decimal digits rather than five places. Each has a dated Corrected line on the page. Five MINOR lines (loose wording and overstatements of the page's own checking) are recorded here and left on the page, as the office does on tier B; three of the checker's MINOR lines were rejected on re-reading, and are counted as confirmed. The five MINOR lines are counted with confirmed.
Claims
CONFIRMED A clay tablet from about 1800 BC carries the number 1;24,51,10 in base sixty beside a square and its diagonal. https://en.wikipedia.org/wiki/YBC_7289 : date-1800. "from some time between 1800 and 1600 BC". Fowler and Robson: "some time in the first third of the second millennium BC". About 1800 BC sits inside both.
CONFIRMED Around 1750 BCE a scribe drew a square with its two diagonals https://isaw.nyu.edu/exhibitions/before-pythagoras/items/ybc-7289 : date-1750. "Old Babylonian Period (19th-17th century BCE)"; "the scribe has drawn a square and its diagonals". 1750 is inside the range.
WRONG 1;24,51,10 across one diagonal, and 42;25,35 along the other https://myslu.stlawu.edu/~dmel/mesomath/tablets/YBC7289 : Re-read https://myslu.stlawu.edu/~dmel/mesomath/tablets/YBC7289 (curl, 2026-11-02 fetch): "along one of the diagonals is the number 1,24,51,10 and below it is 42,25,35". Both numbers sit on the same diagonal, one above and one below it; the page puts 42;25,35 "along the other" diagonal. Fix: say 42;25,35 is written just below 1;24,51,10 on the same diagonal, in the body sentence (index.html line 66), the only copy.
WRONG 1;24,51,10 across one diagonal, and 42;25,35 along the other https://isaw.nyu.edu/exhibitions/before-pythagoras/items/ybc-7289 : Re-read https://isaw.nyu.edu/exhibitions/before-pythagoras/items/ybc-7289 (curl): "An accurate approximation of this quantity in sexagesimal notation is written along one diagonal. ... the product of this number by the square root of 2 is also written along the diagonal." Same false statement as K1 (same label), fixed by the same edit.
CONFIRMED YBC 7289 is an Old Babylonian school tablet, dated to roughly 1800 BC, now in the Yale Babylonian Collection. https://myslu.stlawu.edu/~dmel/mesomath/tablets/YBC7289 : old-babylonian-school. "a round school tablet of unknown provenance from the Old Babylonian period".
CONFIRMED YBC 7289 is an Old Babylonian school tablet, dated to roughly 1800 BC, now in the Yale Babylonian Collection. https://en.wikipedia.org/wiki/YBC_7289 : yale. "The original tablet is currently kept in the Yale Babylonian Collection at Yale University."
CONFIRMED the second number on it, 42;25,35, is exactly thirty times the first https://myslu.stlawu.edu/~dmel/mesomath/tablets/YBC7289 : thirty-times. "It is easy to see that 30 times 1,24,51,10 is 42,25,35". Recomputed: 30 x 30547/21600 = 30547/720 = 42;25,35 exactly.
CONFIRMED the second number on it, 42;25,35, is exactly thirty times the first research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : thirty-times-data. "product_sexagesimal": "42;25,35", "matches_second_number": true; agrees with recompute.out.
WRONG The third number is exactly thirty times the first, because the diagonal of a square of side 30 measures thirty roots of two. https://myslu.stlawu.edu/~dmel/mesomath/tablets/YBC7289 : Upgraded on the page's own arithmetic: the sentence before lists "30 along a side, 1;24,51,10 ..., and 42;25,35", so "the third number is exactly thirty times the first" states 42;25,35 = 30 x 30, which is false (30 x 30 = 900). The source (stlawu YBC7289, re-read with K1): "30 times 1,24,51,10 is 42,25,35", the third is thirty times the SECOND. The dek has the relation right, but this sentence as written states a false figure relation, so it is fixed ("the first" to "the second") with a Corrected line.
CONFIRMED because the diagonal of a square of side 30 measures thirty roots of two https://en.wikipedia.org/wiki/YBC_7289 : diag-30-roots. "approximates the length of the diagonal of a square of side length 30".
CONFIRMED Worked out exactly, that is 1.41421296296..., against a true square root of two of 1.41421356237... research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : tablet-decimal. Recomputed independently: 30547/21600 = 1.41421296296...; sqrt 2 = 1.41421356237...
CONFIRMED The tablet's estimate works out to 1.41421296..., and the true root begins 1.41421356... https://personal.math.ubc.ca/~cass/Euclid/ybc/analysis.html : tablet-decimal-ubc. "which is 1.41421296 to 9 significant decimal figures. The true 9-figure approximation to the square root of 2 is 1.41421356 !"
CONFIRMED so it is right to five decimal places research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : five-places. "correct_digits_after_point": 5; recomputed (1.41421|296 against 1.41421|356).
CONFIRMED so it is right to five decimal places https://en.wikipedia.org/wiki/YBC_7289 : five-places-wp. The record's excerpt ("greatest known computational accuracy") does not itself give a figure; the same passage's "the equivalent of six decimal digits" (six significant, 1.41421) does support five places.
CONFIRMED about one part in 2,359,342 research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : one-in-2359342. "relative_error_parts_per": 2359342; recomputed sqrt2/(sqrt2 - tablet) = 2359342.104.
CONFIRMED about one part in 2,359,342 https://en.wikipedia.org/wiki/YBC_7289 : one-in-2359342-wp. "off by less than one part in 2 million": consistent, the page's figure is the sharper one.
CONFIRMED Three sexagesimal places carry five correct decimal digits. research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : five-digits-sentence. "correct_digits_after_point": 5.
CONFIRMED so the written figure carries five correct decimal digits after the point research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : written-figure-5. As above.
CONFIRMED Run Heron's method from 1 and the third step lands on 577/408, of which the tablet is precisely the three-place truncation. research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : heron3. Step 3 "577"/"408", "sexagesimal_3_places": "1;24,51,10". Recomputed 577/408 = 1;24,51,10,35,17,38... Fowler and Robson pp. 374-375 reconstruct exactly this: "= 1;24 51 10 35 17 ... which, since this is an overestimate, we truncate to 1;24 51 10, the value on YBC 7289".
CONFIRMED The Babylonian tablet YBC 7289 gives a square root of two that is correct to five decimal places. research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : plain-5. "correct_digits_after_point": 5.
CONFIRMED This page works it out in exact fractions, checks the tablet's two numbers against each other, and shows which step of an ancient method produces it. public/strata/the-diagonal-they-already-knew/instrument.js : plain-works-out. instrument.js renders exact fractions and the Heron steps; two-engines.js shows the product and match flag.
CONFIRMED It cannot tell you how the scribe actually did it. https://en.wikipedia.org/wiki/Methods_of_computing_square_roots : plain-cannot. "Babylonian mathematicians calculated the square root of 2 to three sexagesimal "digits" after the 1, but it is not known exactly how." (The article is now titled Square root algorithms; the URL redirects.)
CONFIRMED shows which step of an ancient method produces it https://en.wikipedia.org/wiki/Methods_of_computing_square_roots : not-babylonian. "there is no evidence that the method was known to Babylonians": the page says "an ancient method", not a Babylonian one, so it does not overclaim.
CONFIRMED about 0.00000463 https://personal.math.ubc.ca/~cass/Euclid/ybc/analysis.html : [unit-rounded] rejected on re-reading: 1/216000 = 0.0000046296..., so "about 0.00000463" is a correct rounding, and "about" marks it as approximate; the checker counted it under confirmed. The clash with the page's "truncated, never rounded" note is a fault of that note, recorded as K9.
MINOR so three written places buy nearly five decimal digits of accuracy https://en.wikipedia.org/wiki/YBC_7289 : [nearly-five-digits] Checker's note: one unit in the third place is 1/216000, and 3 x log10(60) = 5.33, so three places buy a little over five decimal digits, not "nearly five"; the page elsewhere says five correct decimal digits. An understatement, loose but not misleading on the headline figure. Tier B MINOR: listed, page left.
CONFIRMED Heron's method takes a guess x and replaces it with (x + 2/x) / 2. https://en.wikipedia.org/wiki/Methods_of_computing_square_roots : heron-formula. Heron's method section gives x_{n+1} = (x_n + S/x_n)/2; Fowler and Robson p. 376: "the average 1/2(a + N/a)".
CONFIRMED Heron's method takes a guess x and replaces it with (x + 2/x) / 2. research/the-diagonal-they-already-knew/verify.mjs : heron-formula-code. "x_next = (x^2 + 2) / (2x)", the same map.
CONFIRMED Started at x = 1, each step doubles the number of correct digits. https://en.wikipedia.org/wiki/Methods_of_computing_square_roots : doubles. "roughly doubles with each iteration". Recomputed correct decimals after steps 1 to 6: 0, 2, 5, 11, 23, 47 (significant digits 1, 3, 6, 12, 24, 48).
CONFIRMED Every step from x = 1 lands above the true root, so no step ever equals the tablet's figure, which sits just below the root. research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : all-above. Every "equals_tablet": false; recompute.out shows all six steps above sqrt 2 and the tablet below it.
CONFIRMED Every step from x = 1 lands above the true root https://en.wikipedia.org/wiki/Methods_of_computing_square_roots : all-above-source. The excerpt ("overestimate to the square root of a positive real number") is about the derivation, not the iterates; the claim holds by AM-GM and by recomputation, and Fowler and Robson p. 371 say the new approximation "will clearly be an overestimate".
CONFIRMED The tablet records a result, not a route: nothing on the clay says how the scribe got there. https://en.wikipedia.org/wiki/Methods_of_computing_square_roots : result-not-route. "but it is not known exactly how." The obverse carries only the three numbers (MESOMATH, CDLI).
CONFIRMED nothing on the clay says how the scribe got there https://en.wikipedia.org/wiki/YBC_7289 : result-not-route-wp. "The student would likely have copied the sexagesimal value of the square root of 2 from a table of constants".
MINOR Reciprocal tables in base 60 made division by 60, 3600 and 216000 routine bookkeeping https://en.wikipedia.org/wiki/Regular_number : [recip-tables] Checker's note: division by 60, 3600 and 216000 in base 60 is a shift of places and needs no reciprocal table (Wikipedia, Regular number: "to divide by n, multiply by 1/n, then shift"). Loose causal wording rather than a false fact about the world. Tier B MINOR: listed.
UNVERIFIABLE which is why a number this good was cheap to write no source states it; searched as N1 did, plus Fowler and Robson 1998 in full : cheap-to-write. Writing three places is cheap whatever the base. Obtaining the value was not cheap on Fowler and Robson's account: the second step needs an approximate reciprocal of the irregular 1;25, "only two OB tables of approximate reciprocals are known", and "it is too tedious to contemplate using anything more than a two-sexagesimal-place initial approximation" (pp. 375-376).
UNVERIFIABLE The same accuracy in decimal notation would have cost five decimal places and a division-friendly base that the scribes did not have. the page's own counterfactual; no source : decimal-counterfactual. Ambiguous as written: on its natural reading it calls decimal the "division-friendly base" the scribes lacked, but base 60 is the more division-friendly one (Regular number article). Matching the 1/216000 resolution would take six decimal places, matching the five correct digits five.
CONFIRMED because each base-sixty place carries nearly two decimal ones public/strata/the-diagonal-they-already-knew/index.html : two-decimals-per-place. Line 122. Recomputed: log10(60) = 1.778 decimal digits per sexagesimal place (recompute.out).
CONFIRMED each base-sixty place is worth nearly two decimal ones src/content/strata/the-diagonal-they-already-knew.md : two-decimals-per-place-2. relates note to the-rule-that-slides; log10(60) = 1.778, recomputed.
CONFIRMED They agree for five decimal places, which the scribe achieved in three sexagesimal places https://en.wikipedia.org/wiki/YBC_7289 : three-vs-five. 1;24,51,10 has three places after the semicolon; five agreeing places recomputed.
CONFIRMED The tablet gives 1.41421296296296296296296296296296296296296296296296. research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : tablet-50. Matches recompute.out truncated at 50 places.
CONFIRMED The true square root of two begins 1.41421356237309504880168872420969807856967187537694. research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : true-50. Matches Python Decimal sqrt(2) truncated at 50 places.
CONFIRMED The verifier research/the-diagonal-they-already-knew/verify.mjs runs offline research/the-diagonal-they-already-knew/verify.mjs : verifier-offline. Imports only node:fs, node:url, node:path and the two engines; ran it: 27 / 27, exit 0.
CONFIRMED It imports both shipped engines research/the-diagonal-they-already-knew/verify.mjs : imports-both. "const engA = await import(... 'engine-a.mjs')", "const engB = ... 'engine-b.mjs'".
CONFIRMED asserts they agree on every emitted value research/the-diagonal-they-already-knew/verify.mjs : asserts-agree. "engine A and engine B emit identical JSON" compares JSON.stringify of both; PASS.
CONFIRMED the tablet's fraction by direct summation of place values research/the-diagonal-they-already-knew/verify.mjs : sum-place-values. tabletFractionIndependent(); PASS.
CONFIRMED the fifty-place decimal by long division research/the-diagonal-they-already-knew/verify.mjs : long-division. longDiv(); PASS.
CONFIRMED the root by a bit-by-bit integer square root of 2 times 10 to the hundredth power research/the-diagonal-they-already-knew/verify.mjs : isqrt. isqrtBitwise(2n * 10n ** 100n); PASS.
CONFIRMED and the six Heron steps by their defining recurrence research/the-diagonal-they-already-knew/verify.mjs : six-heron. heronIndependent(6); PASS.
MINOR It also reads this page's source and asserts the numbers printed above are the ones the engines compute research/the-diagonal-they-already-knew/verify.mjs : [reads-page] Checker's note: verify.mjs only asserts each string occurs somewhere in index.html (html.includes), and a scratch mutation of the body sentence alone still passed 27/27. An overstatement of the page's own checking: MINOR, listed on tier B.
CONFIRMED and that no em dash appears in any shipped file research/the-diagonal-they-already-knew/verify.mjs : no-em-dash. Checks index.html, instrument.js, engine-a.mjs, engine-b.mjs; PASS (two-engines.js is not in its list, though it is also clean).
CONFIRMED The rendered Heron table is a browser matter; the verifier asserts the source text instead. research/the-diagonal-they-already-knew/verify.mjs : heron-table-browser. Header comment and the page-section checks say so.
CONFIRMED This layer's check reads the bytes this page serves and asserts against them. research/the-diagonal-they-already-knew/verify.mjs : placard-reads-bytes. readFileSync(path.join(pubDir, 'index.html')); the live index.html matches the repository file.
OBSERVED Not yet put to that test. git show e65befa3bb:scripts/check-placard.mjs : placard-not-tested. The generator returns this when there is "No dependence run yet". git grep for the slug under research/verifier-independence at e65befa3bb finds nothing; the same search finds a-leg-of-mutton (assert-power.json, census files). The record's excerpt "if (MUTATE) {" is the verifier's own controls, a different test.
OBSERVED Last verified 24 August 2026, when it was written. git show 9accbab6255 : placard-date. Commit 2026-08-24 08:05 +1000, message: "the-diagonal-they-already-knew 27/27 checks, 8/8 controls red, 54/54 leaves agree". Recreated today: 27 / 27 and 8 of 8.
CONFIRMED The tablet shows results only, no working. https://en.wikipedia.org/wiki/YBC_7289 : no-working. Content section lists only the 30 and the two diagonal numbers; Fowler and Robson read it as rough work by a student who copied the value.
CONFIRMED How the scribe obtained 1;24,51,10 is not recorded https://en.wikipedia.org/wiki/Methods_of_computing_square_roots : not-recorded. "but it is not known exactly how." CDLI: "The method that was used by ancient scribes to found such an approximation of the square root of 2 is unknown."
CONFIRMED scholars have proposed several routes https://cdli.ox.ac.uk/wiki/80._square_root_of_two : [several-routes] rejected on re-reading: Checker found it supported: CDLI (https://cdli.ox.ac.uk/wiki/80._square_root_of_two) "Several attempts to reconstruct this method were undertaken, among them that of Neugebauer and Sachs, Fowler and Robson". Counted under confirmed; the near-miss (Fowler and Robson think the scribe copied the value from a coefficient list) does not make "several routes have been proposed" false.
CONFIRMED a Heron-like average https://www.ux1.eiu.edu/~cidelman/Classes/4900/Class%20Notes/Babylonian%20Approximations.pdf : heron-like. Fowler and Robson (Historia Mathematica 25 (1998) 366-378) p. 376: "The procedure described here is mathematically equivalent to the so-called Heron's method". Wikipedia (Square root algorithms) also gives a + b^2/2a with "they may have used a similar approach".
CONFIRMED including reciprocal-table methods https://www.ux1.eiu.edu/~cidelman/Classes/4900/Class%20Notes/Babylonian%20Approximations.pdf : reciprocal-table-methods. N1 found no source; Fowler and Robson pp. 374-375 give two variants of the second step: "either to use an approximation to its reciprocal, or to adjust this initial approximation to a nearby regular number", citing the OB tables of approximate reciprocals M10 and YBC 10529. Reciprocals are a variant inside the Heron-equivalent procedure rather than a separate route, which "including" allows.
CONFIRMED Which starting value reaches the tablet's number is a reconstruction, not a tablet statement. https://en.wikipedia.org/wiki/Methods_of_computing_square_roots : reconstruction. "it is not known exactly how"; Fowler and Robson p. 376: "Both of these alternatives remain conjectural".
CONFIRMED From x = 1, all six reported steps overshoot, and the tablet's value is slightly below the true root, so no step equals it. research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : overshoot. All "equals_tablet": false; recompute.out.
CONFIRMED is the standard one of Neugebauer and Sachs, Mathematical Cuneiform Texts, 1945 https://en.wikipedia.org/wiki/YBC_7289 : neugebauer-sachs. "first recognized by Otto Neugebauer and Abraham Sachs in 1945" (cites MCT p. 43); Fowler and Robson p. 366: "It was first published in [19, 43]" ([19] = MCT, 1945) and quote Neugebauer's reading "obtained from a = 30 the diagonal d = 42;25,35 by using root 2 = 1;24,51,10".
CONFIRMED Mathematical Cuneiform Texts, 1945 https://archive.org/details/mathematicalcune0000neug : mct-record. Metadata: title "Mathematical cuneiform texts", creator "Neugebauer, O. (Otto), 1899-1990, ed", date "1945".
CONFIRMED alternative readings exist https://en.wikipedia.org/wiki/YBC_7289 : alt-readings. "one alternative interpretation is that the number on the side of the square is 30/60 = 1/2"; Fowler and Robson p. 368 discuss it.
MINOR All decimals on this page are truncated, never rounded, so the last shown digit is honest. public/strata/the-diagonal-they-already-knew/index.html : [truncated] Checker's note: the engine and instrument decimals are truncated, but the hand-written "1/216000, about 0.00000463" is rounded, so "All decimals on this page are truncated, never rounded" has one exception. An overstatement of the page's own practice: MINOR, listed on tier B.
OBSERVED are two independent implementations of one written specification, the first by Claude and the second by Ox Alpha, neither author having seen the other's code git show e65befa3bb:research/_wave-2026-08-24/README.md : two-engines-claude-ox. "A is Claude's, B and C are Ox Alpha's"; engine-a.mjs header: "from the specification alone, before Engine B was read".
OBSERVED computed twice, by two programs written independently from one written specification git show e65befa3bb:memory/log.d/2026-08-23T1930Z-the-two-engines-wave.md : two-engines-spec. "each from the same written specification, neither author having seen the other's code"; table row "the-diagonal-they-already-knew
OBSERVED one by Claude and one by Ox Alpha, neither author having seen the other's code git show e65befa3bb:research/_wave-2026-08-24/README.md : one-by-claude-one-by-ox. As above.
CONFIRMED Everything is exact rational arithmetic on BigInt public/strata/the-diagonal-they-already-knew/engine-a.mjs : exact-bigint. "Everything here is an exact rational on BigInt. No Number, no Math.sqrt"; engine-b.mjs: "No floating point appears anywhere in this file". Only the final integer parts-per is converted with Number().
MINOR so a floating point square root would have quietly destroyed the result public/strata/the-diagonal-they-already-knew/engine-a.mjs : [floats-destroy] Checker's note: IEEE doubles reproduce every headline figure (Math.sqrt(2) 1.4142135623730951 against the tablet 1.414212962962963, five agreeing places); only the fifty-place expansions need exact arithmetic. An overstatement about the page's own method: MINOR, listed on tier B.
WRONG a value written in clay in about 1800 BC is right to nine decimal digits https://personal.math.ubc.ca/~cass/Euclid/ybc/analysis.html : Re-read https://personal.math.ubc.ca/~cass/Euclid/ybc/analysis.html (curl): "1 24 51 10 = 1 + 24/60 + 51/3600 + 10/216000, which is 1.41421296 to 9 significant decimal figures. The true 9-figure approximation to the square root of 2 is 1.41421356". They first differ at the sixth decimal place, so the tablet is right to five decimal places, as the page says; "right to nine decimal digits" in the served engine-a.mjs (lines 5-7, a code comment a reader can open) is false. Fix the comment; its clause "a float would quietly hide the last two of them" sits in the same sentence and no longer has a referent (a double carries about 16 digits), so it is recast too.
CONFIRMED Both run here, in your browser. public/strata/the-diagonal-they-already-knew/two-engines.js : both-run-here. "const a = computeA(), b = computeB();" from ./engine-a.mjs and ./engine-b.mjs, loaded by index.html line 186.
CONFIRMED this stub body is not rendered anywhere on the site src/pages/strata/[...slug].astro : stub-not-rendered. ".filter((s) => !s.data.draft && !s.data.href) // immersive layers serve their own page"; this layer has an href.
CONFIRMED The Old Babylonian tablet YBC 7289 gives the diagonal of a square of side 30 as 42;25,35 https://en.wikipedia.org/wiki/YBC_7289 : diagonal-labeled. "approximates the length of the diagonal of a square of side length 30".
CONFIRMED Six steps from x = 1, and none equals the tablet. research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : none-equals-tablet. Six entries, all "equals_tablet": false.
CONFIRMED Every step approaches the root from research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : approaches-from-above. Full note: "Every step approaches the root from above"; recompute.out: all above.
CONFIRMED (s.equals_tablet ? ' ; equals the tablet' : ' ; not the tablet') research/_wave-2026-08-24/out/the-diagonal-they-already-knew.A.json : [row-says-not-tablet] rejected on re-reading: Checker counted it under confirmed: the step rows compare exact fractions, and 577/408 is not the tablet's fraction 30547/21600, so "not the tablet" is true; the row also prints the three-place sexagesimal 1;24,51,10, which matches the dek's "three-place truncation". Nothing false.
CONFIRMED Each click applies (x + 2/x)/2 once. public/strata/the-diagonal-they-already-knew/instrument.js : each-click. Each click reveals the next precomputed step C.heron[stepIdx], which is one application of the map.
CONFIRMED That layer carries the proof that the diagonal of a square cannot be written as a ratio. src/content/strata/incommensurable.md : incomm-proof. dek: "A valid proof that root 2 is irrational, written as a Shakespearean sonnet".
CONFIRMED This tablet is more than a thousand years older https://en.wikipedia.org/wiki/Hippasus : older-1000. Hippasus "c. 530 - c. 450 BC"; the tablet 1800 to 1600 BC; at least about 1,100 years either way.
UNVERIFIABLE gets the number right to five decimal places without any such proof, and without, so far as anyone can tell, needing one searched Wikipedia YBC 7289, MESOMATH, CDLI, Casselman's commentary, Fowler and Robson 1998 : without-proof. No source discusses an Old Babylonian proof of irrationality; the claim is hedged interpretation.
CONFIRMED this one is the oldest example either of them can point at src/content/strata/the-paper-computer.md : oldest-example. The Paper Computer and The Rule That Slides are about logarithmic slide rules (17th century onward); the tablet is older.
CONFIRMED A physical object that does arithmetic for you. src/content/strata/the-rule-that-slides.md : rule-prints-slide-rule. "Print and measure a 250 mm C/D slide rule".
CONFIRMED (added) It is published to read. Running it needs the repository around it. https://artwaste.land/checks/research/the-diagonal-they-already-knew/verify.mjs : Placard line 190. Downloaded (200, identical to the repository): alone in an empty directory it exits 1 with ERR_MODULE_NOT_FOUND for ../../public/strata/.../engine-a.mjs; with index.html, engine-a.mjs, engine-b.mjs and instrument.js fetched from the live page into the repository layout it prints "27 / 27 checks passed", exit 0.
CONFIRMED (added) step through the iteration that gets closer still public/strata/the-diagonal-they-already-knew/index.html : Line 64 (served dek). recompute.out: steps 4, 5 and 6 are closer to sqrt 2 than the tablet (steps 1 to 3 are not; 577/408 is about 2.1e-6 off against the tablet's 6.0e-7).
What was done
[fixed] both-on-one-diagonal (MESOMATH): index.html:66 said 1;24,51,10 was written "across one diagonal, and 42;25,35 along the other". Melville's MESOMATH page reads "along one of the diagonals is the number 1,24,51,10 and below it is 42,25,35". The sentence now reads "1;24,51,10 along one diagonal, and 42;25,35 just below it on the same diagonal", the verifier's section heading "The product along the other diagonal" now reads "The product written below it on the same diagonal", and a Corrected line was added.
[fixed] both-on-one-diagonal (ISAW): the same false sentence, contradicted by the ISAW entry ("written along one diagonal ... also written along the diagonal"), fixed by the same edit and the same Corrected line.
[fixed] third-first (upgraded from MINOR): index.html:66 said "The third number is exactly thirty times the first" where its own list starts with the side, 30; 30 x 30 is not 42;25,35. Now "thirty times the second" (MESOMATH: "30 times 1,24,51,10 is 42,25,35"), with a Corrected line.
[fixed] nine-digits-comment: engine-a.mjs lines 5-7 (served) and its wave copy research/_wave-2026-08-24/engineA/the-diagonal-they-already-knew.mjs said the clay value "is right to nine decimal digits, and a float would quietly hide the last two of them". Casselman's analysis gives 1.41421296 against 1.41421356, so five decimal places. The comment now says five decimal places and that exact fractions let both fifty-place expansions be printed with no rounding; a Corrected line was added.
[declined] nearly-five-digits (MINOR, tier B, left on the page): Checker's note: one unit in the third place is 1/216000, and 3 x log10(60) = 5.33, so three places buy a little over five decimal digits, not "nearly five"; the page elsewhere says five correct decimal digits. An understatement, loose but not misleading on the headline figure. Tier B MINOR: listed, page left.
[declined] recip-tables (MINOR, tier B, left on the page): Checker's note: division by 60, 3600 and 216000 in base 60 is a shift of places and needs no reciprocal table (Wikipedia, Regular number: "to divide by n, multiply by 1/n, then shift"). Loose causal wording rather than a false fact about the world. Tier B MINOR: listed.
[declined] reads-page (MINOR, tier B, left on the page): Checker's note: verify.mjs only asserts each string occurs somewhere in index.html (html.includes), and a scratch mutation of the body sentence alone still passed 27/27. An overstatement of the page's own checking: MINOR, listed on tier B.
[declined] truncated (MINOR, tier B, left on the page): Checker's note: the engine and instrument decimals are truncated, but the hand-written "1/216000, about 0.00000463" is rounded, so "All decimals on this page are truncated, never rounded" has one exception. An overstatement of the page's own practice: MINOR, listed on tier B.
[declined] floats-destroy (MINOR, tier B, left on the page): Checker's note: IEEE doubles reproduce every headline figure (Math.sqrt(2) 1.4142135623730951 against the tablet 1.414212962962963, five agreeing places); only the fifty-place expansions need exact arithmetic. An overstatement about the page's own method: MINOR, listed on tier B.
The assay office: what a record is, and every layer re-read so far.