The Riemann hypothesis from Euler to November 2026, read against the sidereal sky, with an honest test of what the sky can say.
Under the fig tree
Selasca, Lago Maggiore · 20 July 1866
Bernhard Riemann died on a Friday in July 1866, at the Villa Pisoni in Selasca, on the western shore of Lake Maggiore. He was thirty-nine, and tuberculosis had been taking him for four years. His friend Richard Dedekind, who wrote the short biography printed with Riemann's collected works, recorded that on the day before he died Riemann sat in the shade of a fig tree, glad of the view, and was still working, on the mechanics of the ear.
After his death a housekeeper threw away part of the papers in his study. What survived went to Göttingen. Sixty-six years later Carl Ludwig Siegel worked through those notes and found something Riemann had never published: a practical method for computing the zeros of the function that bears his name. Riemann had not only guessed where those zeros lie. He had calculated some of them.
The guess is the Riemann hypothesis, and it is still, at the time of writing, unproven in any form the mathematical world has accepted. Its history is unusually well dated. It was first stated on a Thursday in November 1859. It has milestones fixed to the day by academy sessions, arXiv timestamps and press releases. That makes it an unusual subject for Jyotiṣa, which holds that the sky at a moment bears on what happens at that moment.
Anton Coops proposed charting that history. This essay does so from a single point of reference, the small asteroid that carries Riemann's name, and reads the sky at each turning point from there. Then it does one thing more: it writes down what a correspondence should look like, and tests it.
IThe music of the primes
St Petersburg 1737 · Braunschweig 1792 · Berlin 1837
A prime is a whole number with no divisors but itself and one: 2, 3, 5, 7, 11, 13. Every whole number is a product of primes in exactly one way, which makes them the atoms of arithmetic. They also thin out as numbers grow, without any simple rule for where the next one falls. Between 1 and 100 there are 25 primes. Between 1,000,000 and 1,000,100 there are six.
On 25 April 1737 Leonhard Euler presented a paper to the Academy in St Petersburg that tied the primes to something smooth. Take the sum 1 + 1/2s + 1/3s + 1/4s + … over every whole number. Euler showed it equals a product with one factor for each prime, 1/(1 − p−s). The identity is a translation of unique factorisation into analysis, and Euler used it to show that the reciprocals of the primes, 1/2 + 1/3 + 1/5 + 1/7 + …, add up to infinity.
Half a century later a fifteen-year-old in Braunschweig was counting. Carl Friedrich Gauss had been given a book of logarithm tables that also listed the primes, and in 1792 or 1793 he began tallying them in blocks of a thousand. Near a number x, he saw, about one number in log x is prime. He kept counting for decades and told almost no one. When the astronomer Johann Franz Encke wrote to him about the same question, Gauss replied on Christmas Eve 1849 with an account of his tables and of work begun half a century earlier. Adrien-Marie Legendre had meanwhile published a formula of his own, in 1798.
Gauss's law says the number of primes up to x, written π(x), is close to the logarithmic integral li(x), the area under 1/log t. The plate below lets you see how close.
The third person in this prelude is Peter Gustav Lejeune Dirichlet. On 27 July 1837 he read a paper to the Berlin Academy proving that any arithmetic progression such as 3, 7, 11, 15, … contains infinitely many primes, provided its first term and step share no factor. His tools were Euler's products, adapted, and calculus. Number theory had acquired analysis. A decade later a shy young man from Hanover, who had begun university as a student of theology, came to Berlin to hear Dirichlet lecture, and it changed his life.
IIEight pages in November
Breselenz 1826 · Göttingen and Berlin 1859
Georg Friedrich Bernhard Riemann was born on 17 September 1826 in Breselenz, a village near Dannenberg in the Kingdom of Hanover, the second of six children of a Lutheran pastor. He studied in Göttingen and Berlin, became Dirichlet's successor in Gauss's old chair at Göttingen in 1859, and in the same year, on 11 August, was elected a corresponding member of the Berlin Academy.
New members were expected to report on their research. Riemann chose the distribution of primes, a subject, he wrote, made worthy by the long interest of Gauss and Dirichlet. He sent the paper from Göttingen on 19 October. On Thursday, 3 November 1859, Ernst Kummer read it to the Academy, and it appeared in the November Monatsberichte. It runs to eight printed pages and is the only paper Riemann ever wrote on number theory.
In those pages Riemann let the variable s in Euler's sum be a complex number, extended the function to the whole complex plane, and named it ζ(s). He found its central symmetry, a functional equation relating ζ(s) to ζ(1 − s), which mirrors the plane across the vertical line where the real part of s is ½. And he wrote down an exact formula for counting primes in which each correction to Gauss's curve comes from a zero of ζ.
Then comes the sentence that has occupied mathematicians ever since. Riemann had rewritten the function so that zeros on the critical line appear as its real roots, and he observed:
„Es ist sehr wahrscheinlich, dass alle Wurzeln reell sind.“It is very likely that all the roots are real.
A proof would be welcome, he added, but after a few brief attempts he had put the search aside, since it was not needed for his purpose. That is the whole of the hypothesis: every non-trivial zero of ζ has real part exactly ½.
Why should anyone care where the zeros are? Because in Riemann's formula each zero acts like a tone. The zero ½ + iγ contributes an oscillation of frequency γ (on a logarithmic scale) to the count of primes, and its real part sets how loud it can grow. If every real part is ½, every tone swells no faster than √x, and the primes keep as close to Gauss's curve as they possibly can. The next plate rebuilds the prime count from its zeros, one tone at a time.
IIIListening to the zeros
Copenhagen 1903 · Göttingen 1932 · Manchester 1950 · 2020
The first person known to have located zeros of ζ was Riemann, though no one knew it until Siegel opened the Nachlass in 1932. In print, the Danish actuary Jørgen Pedersen Gram was first: in 1903 he confirmed by hand that the first ten zeros lie on the line. The working tool is a function now called Hardy's Z(t): it is real along the critical line and changes sign exactly where a zero sits there, so finding zeros becomes a matter of watching a real graph cross its axis.
Siegel's find gave the fast way to compute Z, the Riemann–Siegel formula, and the record climbed: 1,041 zeros by Titchmarsh's team in the mid-1930s. In 1950 Alan Turing ran the computation on the Manchester Mark 1, one of the first stored-program computers, extending the check to 1,104 zeros. He hoped, by his own account, to find a zero off the line. He found none, and the method he devised to certify that no zero had been missed is still in use.
Andrew Odlyzko later computed hundreds of millions of zeros near the 1020-th. In 2020 David Platt and Tim Trudgian verified, with rigorous interval arithmetic, every zero up to height three trillion. Not one has strayed from the line. That is strong evidence, and it is not proof: the arithmetic of very large numbers is full of patterns that hold for trillions of cases and then fail, as Part IV will show.
IVA century of siege
1896 · Paris 1900 · 1914 · Rouen 1940 · 1974 · 1989 · 2026
Riemann's paper did not prove Gauss's law; it showed how a proof might go. In 1896 Jacques Hadamard and Charles de la Vallée Poussin, working separately, completed it by showing that no zero of ζ lies on the line Re(s) = 1. That is the prime number theorem. The hypothesis asks for much more: not just that the zeros keep away from 1, but that they all sit exactly on ½.
On 8 August 1900 David Hilbert addressed the International Congress of Mathematicians at the Sorbonne and set out the problems he thought the new century should solve. The Riemann hypothesis sits inside the eighth. Hilbert is reported to have said that if he woke after sleeping a thousand years, his first question would be whether it had been proved. A year later Helge von Koch showed the hypothesis is equivalent to the sharpest possible error in Gauss's law: the primes stay within about √x log x of li(x).
Spring 1914 brought two notes in the Comptes Rendus. On 6 April Hardy proved that infinitely many zeros lie on the line. On 22 June his collaborator J. E. Littlewood proved something unsettling: π(x) − li(x), negative for every x anyone had ever checked, changes sign infinitely often. Plate 1 showed li(x) always above the staircase. Somewhere far beyond any table, it is not. Evidence from small numbers can mislead.
The next stage was to show that a positive share of the zeros lies on the line. Atle Selberg did it in 1942. Norman Levinson pushed the share past a third in 1974, Brian Conrey past two fifths in 1989, and successive refinements reached about 41.7% by 2020. Then, in August 2026, the number jumped.
Meanwhile the strongest evidence came from neighbouring worlds. Analogues of ζ can be built for curves and varieties over finite fields, and they have their own Riemann hypotheses. André Weil proved the one for curves in 1940, in circumstances no one would choose. He had failed to report for military duty and was held in the Bonne-Nouvelle military prison in Rouen. He worked intensely there, wrote to his sister Simone and to Henri Cartan, and sent a three-page note to the Paris Academy that Élie Cartan presented at the session of Monday, 22 April 1940. Pierre Deligne proved the general case for varieties in 1974. In those worlds the hypothesis is true, and the reason is understood. In the original world it is neither.
VTea at the Institute
Princeton · Thursday, 6 April 1972
Hugh Montgomery had come back to the Institute for Advanced Study in Princeton to show Atle Selberg a new result about how the zeros are spaced. At afternoon tea in Fuld Hall the Indian number theorist Sarvadaman Chowla insisted on introducing him to the physicist Freeman Dyson. Montgomery was reluctant. Dyson asked what he was working on, and Montgomery wrote down his formula for how often two zeros sit a given distance apart: 1 − (sin πu / πu)².
Dyson recognised it at once. It was the pair-correlation function for the eigenvalues of large random Hermitian matrices, the Gaussian Unitary Ensemble, which physicists had been using to model the energy levels of heavy atomic nuclei. The zeros of ζ, it seemed, are spaced like the energy levels of a quantum system. They repel one another; two zeros almost never sit very close together.
The match has since been confirmed to many decimal places by Odlyzko's computations high up the line. You can test it yourself on the first two thousand zeros.
VIPrizes, claims, and 2026
Paris 2000 · 2004 · Heidelberg 2018 · 2024 · August–November 2026
On 24 May 2000, at the Collège de France, the Clay Mathematics Institute announced seven Millennium Prize Problems with a million dollars attached to each. The Riemann hypothesis was the oldest of them. The rules require a solution to be published and to survive two years of scrutiny before the Institute will consider it.
Claims followed, as they always had; Stieltjes had announced one in 1885. On 8 June 2004 Purdue University issued a press release for Louis de Branges's posted proof, an approach that had already met counterexamples. On Monday, 24 September 2018, Sir Michael Atiyah, then 89, presented a short argument at the Heidelberg Laureate Forum. Neither was accepted.
Real progress came in smaller, harder pieces. On 18 January 2018, at 02:47 UTC, Brad Rodgers and Terence Tao posted a proof of Newman's conjecture: a constant Λ attached to ζ is at least zero. The hypothesis is equivalent to Λ ≤ 0, so if it is true it is only just true, with no room to spare. On 31 May 2024 Larry Guth and James Maynard posted the first improvement in more than eighty years to Ingham's bound on how many zeros can lie far from the line.
Then came 2026. On 10 August Anthropic reported that an internal Claude model had raised the proven share of zeros that are simple and on the critical line from about 41% to more than two thirds (Plate 5); the paper by Levent Alpöge and Ralph Furman appeared on arXiv on 13 August. On 6 October OpenAI released several hundred machine-written manuscripts, one claiming that no zero has real part above 7/8, a statement called the quasi-Riemann hypothesis; it is unrefereed. On Friday, 20 November, Alpöge posted a four-page note and a very large Lean formalisation claiming, with Claude, a proof of the Riemann hypothesis and of its generalised form. As this essay is written, a day later, verification has begun and nothing is settled.
VIIThe sky, seen from ζ
Thirty-eight charts, 1737–2026
The study casts no nativity, neither Riemann's nor one for the hypothesis. It takes a single point of reference instead: the main-belt asteroid (4167) Riemann, marked here with the letter ζ. Every chart below is read from ζ outward, and every measurement in Parts VIII and IX is a measurement from ζ to something else.
Jyotiṣa has a familiar way to do this. When a birth time is unknown, a jyotiṣī can count the houses from the Moon's sign instead of the rising sign, as Chandra lagna. Here ζ's sign plays that part: it is the first house, the next sign is the second, and so on around the zodiac. A graha then relates to ζ in two ways. One is by degree, the angular distance between them. The other is by sign: whether it shares ζ's sign or casts its glance (dṛṣṭi) on it. Every graha aspects the seventh sign from itself; Mars also aspects the fourth and eighth, Jupiter the fifth and ninth, Saturn the third and tenth.
The thirty-eight dated moments are cast in the sidereal zodiac with the Lahiri ayanāṃśa, using the mean node. Riemann's birth is one of them, charted and read from ζ like any other milestone; it is not used as a nativity. Where the clock time is unknown, noon is assumed and the Moon is marked as uncertain by about ±6½°; for events known only to the year, only the slow bodies are shown.
Seen across all the dated moments at once, ζ wanders the whole zodiac. It goes around in a little over four years, so the milestones of a century and a half scatter it through every sign. The next plate shows where it stood each time.
VIIIA model of correspondence
Symbol · measure · test
A list of charts is not yet a claim. To say the sky corresponds to this history is to say something definite: that on the days that mattered, particular bodies stood in particular relations to ζ more often than on ordinary days. Stated that way, the claim can fail. The model has three layers.
The symbols. Each body is paired with something in the story, using classical Jyotiṣa where it speaks and modern practice where it does not. The reference point itself, ζ, stands for the name.
| In the story | Measured from ζ to | Reasoning | Test |
|---|---|---|---|
| Calculation, proof | Mercury, Budha बुध | Kāraka of गणित gaṇita, writing and reasoning. | Z1 |
| Visibility, recognition | The Sun | ζ conjunct the Sun is hidden in its glare; opposite, it is at its brightest. Either is a moment of emphasis. | Z2 |
| The zeros; the critical line | Rāhu–Ketu axis | शून्य śūnya, the void, is Ketu's; and as the zeros mirror across Re(s) = ½, the nodes stand exactly opposite one another. | Z3 |
| Academies, prizes | Jupiter, Guru गुरु | Learning, institutions, recognition. | Z4 |
| 167 years unresolved | Saturn, Śani शनि | Time, delay, endurance, hard labour. | Z5 |
| The ground of number | Galactic Center | At 3° Dhanus in मूल Mūla, "the root". | Z6 |
| Elegance of a proof | Venus, Śukra शुक्र | Beauty, refinement, harmony. | Z7* |
| Attack on the problem | Mars, Maṅgala मङ्गल | Drive, contest, courage, technical force; interpretively, mathematics itself. | Z8* |
| Machines; illusion; depth | Uranus; Neptune; Pluto | Modern usage only. | Z9–Z11* |
| All of it together | Every body above | A composite of all eleven contacts, and a count of grahas whose sign-based dṛṣṭi falls on ζ's sign. | Z12*, Z13* |
* Added when the analysis was rebuilt around ζ, after the first round of results had been seen. Z1–Z6 were stated before any statistic was computed.
Two significators for mathematics. In Jyotiṣa, mathematics is interpretively associated with Mars, Maṅgala मङ्गल. Mercury’s connection to intelligence makes it a naturally sought significator as well, and it is the one this study named first, as Z1. Mars entered the tests later, as Z8; Part IX returns to it.
The measures. Each pairing becomes an angle measured on any date: the separation Δλ from ζ to the other body. Over a set of n days the score is the average of cos(kΔλ). With k = 1 the score approaches 1 when the body is conjunct ζ; with k = 2 it approaches 1 for conjunction and opposition alike, which suits the Sun, the nodal axis and the Galactic Center. The composite adds all eleven contacts with a 3° Gaussian orb. The dṛṣṭi count asks how many of the six visible grahas, Sun to Saturn, occupy or aspect ζ's sign.
The tests. Thirteen in all, on the 21 events whose day is known. The asteroid's own discovery is left out, because asteroids are found near opposition to the Sun. Each score is compared with 20,000 sets of random dates under two baselines: one redraws each date independently within ten years of the true one; the other shifts the entire history by a single random offset of up to twenty years, which keeps its real clustering. With thirteen tests, a false-discovery correction and the stricter Bonferroni bound are both applied.
IXWhat the sky said
Thirteen tests, two baselines, 20,000 redraws each
Measured from ζ, most of the sky is indifferent. The nodes, Jupiter, Saturn, the Galactic Center, Uranus, Neptune and Pluto show no preference for ζ on the milestone days. The Sun is no more often on ζ's axis than chance would put it, and Venus leans only slightly toward ζ, for a reason given at the end of this part. The count of grahas aspecting ζ's sign runs a little high, , which is not enough once thirteen tests are taken into account.
Two contacts stand apart, and they arrive with different credentials.
Mercury conjunct ζ was stated in advance. On the milestone days its score is , where chance gives about (p = ). It passes a false-discovery threshold under the baseline that keeps the history's clustering (q = ) but not under independent redraws (q = ), and fails the Bonferroni bound under both. A single day carries much of it: at Riemann's birth ζ sat 1.4° from Mercury. Part of it may also be borrowed: ζ tends to stand on the Sun's side of the sky on these days, and Mercury never strays far from the Sun.
Mars conjunct ζ was not stated in advance. It appeared when the study was rebuilt around ζ and every body was measured from it. It is now the strongest result: a score of where chance gives (p = ). It passes the false-discovery threshold under both baselines and the Bonferroni bound for thirteen tests under the rigid shift; with independent redraws it falls just short of Bonferroni. It is also unusually sturdy. Dropping any one event leaves p at or below , and it holds without the 2026 events and with the 1859 cluster counted once. On of the 21 days Mars is within 25° of ζ.
A jyotiṣī can tell a story for Mars, the warrior among the grahas, standing beside the name on the days the problem was attacked; and Mars is also, interpretively, a significator of mathematics itself. But this study did not name that contact before the numbers came in, and that the story arrives after the number is exactly why it cannot count yet. Thirteen tests were run, and the family of seven added tests was defined after the first results were known. Mars ☌ ζ is a candidate, the strongest this study has produced, and the right next step is the one taken in Part X: write the rule down and score the future against it.
The plate below lets you look at the raw material directly: the angle from ζ to Mars or to Mercury on every day from 1826 to the end of 2028.
Noticed along the way
ζ and the Sun.
The claimed proofs.
Autumn 1859.
XA test still to come
August – November 2026, and after
A model earns trust by predicting, so here are two predictions that can be scored. If the contacts are real, the decisive moments still ahead should fall on days when Mercury, or Mars, is within 15° of ζ more often than chance allows. Moments to score include an expert statement that the November 2026 proof is sound or flawed, its acceptance by a refereed journal, a published refutation, and any decision by the Clay Institute. One hit would mean little. A tally over the next decade, scored against these base rates, would be a real test. The model takes no position on whether the proof is correct.
Resonance, not prediction
Coda
The symbols fit beautifully. The zeros are a kind of śūnya; the critical line is an axis of symmetry, as the nodal axis is; a small body carrying Riemann's name circles the Sun every four years and can be read, as the Moon is read, as a lagna of its own. A reader who knows Jyotiṣa will feel those resonances, and they are real as resonances.
What the sky does not do, on this evidence, is keep time with the problem in the ways the symbols first suggested. Of the six contacts stated in advance, five are indistinguishable from chance and the sixth, Mercury beside ζ, is suggestive but fragile. The strongest signal came from a contact no one predicted, Mars beside ζ, and an unpredicted signal is a question, not an answer.
That stance belongs to both traditions this essay draws on. Riemann called his hypothesis very likely and set it aside, unwilling to claim more than he could show. A careful jyotiṣī reads a chart as a field of tendencies, not a timetable. The zeros will be shown to lie on the line, or not, by proof. Whether ζ keeps company with Mars when the next chapter is written is now a matter of record waiting to happen.
Methods, data and sources
Apparatus
Reference point. No nativity is used. All measurements are taken from the asteroid (4167) Riemann, ζ; houses are counted from ζ's sign. Ephemerides. Planets, Moon and nodes: Swiss Ephemeris 2.10 (Moshier), geocentric and apparent, checked against JPL Horizons for Mars at four dates from 1737 to 2026 (agreement under 1″). Sidereal frame: Lahiri (Chitrapakṣa); Rāhu is the mean node. Galactic Center: Sgr A*, precessed to each date (IAU 2006); 3°00′ Dhanus throughout. ζ: JPL Horizons, solution JPL#78 (13,191 observations, 1968–2026), for every event; for the random baselines its orbit was integrated with all eight planets and Pluto as perturbers (REBOUND, IAS15) from 1700 to 2050, reproducing Horizons to within 23″ at every event date. Times. Local mean time before standard time, then the civil time then in force; noon where no time is recorded; year-only events at 12:00 UT. Zeros. The first 2,000 zeros were computed with mpmath to six decimals. Statistics. Thirteen one-sided tests; Z1–Z6 stated before any computation, Z7–Z13 added when the study was rebuilt around ζ; 20,000 draws per baseline with a fixed seed; Benjamini–Hochberg and Bonferroni corrections across all thirteen.
Show all 38 charts as a table (sidereal positions)
- Riemann, "Ueber die Anzahl der Primzahlen unter einer gegebenen Grösse", Monatsberichte der Berliner Akademie, Nov. 1859; German text and translation by D. R. Wilkins, Clay Mathematics Institute. The quotation in Part II is from the German original; the English is this essay's.
- Paper sent 19 Oct., read by Kummer 3 Nov. 1859: American Institute of Mathematics. Election 11 Aug. 1859 and letter to Weierstrass: J. Arias de Reyna.
- Riemann's last days and the fig tree, from Dedekind's memoir: J. Derbyshire, Prime Obsession, ch. 22; birth and death: MacTutor.
- Euler, E72, presented 25 Apr. 1737: J. Lagarias, "Euler's constant", ref. [99].
- Gauss to Encke, 24 Dec. 1849: Göttingen Academy, Gauss correspondence; Y. Tschinkel, Bull. AMS 43 (2006).
- Dirichlet, read 27 July 1837: translation by R. Stephan.
- Hilbert, 8 Aug. 1900: MacTutor; Clay's date of 9 Aug.: Millennium Problems.
- Hardy, C. R. 6 Apr. 1914; Littlewood, 22 June 1914: references in Hardy and Littlewood, Acta Math. 41.
- Weil, séance of 22 Apr. 1940: M. Hindry, Journées X-UPS 2012; J. Milne, "The Riemann hypothesis over finite fields".
- Turing and the zeta function: D. Hejhal and A. Odlyzko, "Alan Turing and the Riemann zeta function".
- Proportions on the line: Bui, Conrey and Young, arXiv:1002.4127; Goldston and Suriajaya, arXiv:2511.20059; Baluyot, arXiv:2501.14545.
- Montgomery and Dyson, 6 Apr. 1972: Institute for Advanced Study; Montgomery's account: M. Watkins.
- Asteroid 4167 Riemann: IMCCE discovery list; orbit and ephemeris: JPL Horizons.
- Clay Millennium Prize, 24 May 2000: Clay Mathematics Institute.
- de Branges, 8 June 2004: Purdue news release. Atiyah, 24 Sep. 2018: Heidelberg Laureate Forum blog.
- Rodgers and Tao: arXiv:1801.05914. Platt and Trudgian: arXiv:2004.09765. Guth and Maynard: arXiv:2405.20552.
- Alpöge and Furman: arXiv:2608.13637; Lamzouri: arXiv:2609.02882. OpenAI release of 6 Oct. 2026 and its status: arXiv:2610.12803, §1.1. The claim of 20 Nov. 2026: Al Jazeera, 21 Nov. 2026.
AyurAstro.com · Writings · 21 November 2026. The idea for this essay came from Anton Coops (DEEPsee · 360 Card View). Prepared with Claude (Anthropic). Positions, zeros and statistics were computed for this essay. All measurements are taken from the asteroid (4167) Riemann; six of the thirteen tests were stated before any computation, and the seven added later are marked.