Travel from counting-rod tables in Luoyang, determinants in Edo, a Hanover letter, a Geneva rule, Göttingen observation, and London’s new algebra into quantum states, compression, optimization, graphics, search, and AI. Watch a number table change from a place to arrange problems into an operation that transforms worlds.
QUESTION FOR THE ROUTE
Why did equations, spaces, observations, and data—often without a direct line of transmission—repeatedly need a language that tables relations and transforms the table itself?
WHAT THIS RIVER DOES NOT CLAIM
The river is not geography. Distance downstream stands for time passing, and the light turns from dawn to dusk as the centuries go by. The objects by each stele are symbols of the kind of event and of how each century band wrote and calculated; they do not reconstruct any real artefact. The land around each stop sketches the natural geography of the scene’s real place, and an iconic building appears only if it already stood in that year. Each scene keeps its real place and evidence basis; open it on the map to read where it happened. This route does not call the Nine Chapters’ counting-rod array a modern abstract matrix or crown Seki, Leibniz, Cramer, Gauss, Sylvester, or Cayley as a sole inventor. Determinants, elimination, vector spaces, matrix algebra, and numerical linear algebra grew from different problems at different times. Modern applications do not mean matrices explain all meaning or social judgment in quantum theory, search, or AI.
WHAT YOU SEE ON THIS RIVER
- Diagram in the sky
- A lattice sheared by a transformation
- Emblem at the source
- A table of counting rods
- The real place around each stop
- Around each stele the land takes on the natural geography of that scene’s real place — sea or lake, plain, hills or mountains, the colour of the ground and its common trees — and, where one defines the place, its landform: a volcano, snow peaks, granite domes, a mesa, dunes, a fjord, islands, a rock hill, a gorge or loess terraces. The water near the stop takes the colour of the real river or sea, and the haze the place’s climate. A small globe on the stele marks where it is, with the route from the previous place. Where a city has an iconic building that already stood in the scene’s year, its schematic silhouette rises behind the stop and is named on the card. The land follows today’s terrain and climate as a sketch and the silhouettes are not measured reconstructions. Between stops the river itself stays symbolic.
- A figure board at every stop
- Each board draws the mathematics of that scene. When the boat arrives, the construction is drawn in and the key result rises in red. The drawings are schematic reconstructions, not historical manuscripts.
- Century bands along the banks
- to 499 · Sandstone stele · braziers · earthen villages and beacons · rafts · flocks of birds
- 1450–1749 · Marble stele · iron lanterns · three-arch bridge · windmills and clock-tower towns · sailing ships
- 1750–1899 · Cast-iron plaque · gas lamps · iron truss bridge · factory chimneys, railway and steam train · steamboats
- 1900–1969 · Concrete marker · electric streetlights · concrete bridge · pylons, apartment blocks and radio masts · barges · aircraft
- 1970 onward · Glass marker · LED lights · cable-stayed bridge · glass towers, wind turbines and data centres · ferries · satellites
Where the century band changes, the boat passes under a bridge of the new band. Villages, mills, factories, pylons and towers stand for the technology of each century, not for any real place or architectural style.
01·263 CE·Luoyang(basis: Reception)
Reducing Rows in a Counting-Rod Table — The Nine Chapters and Liu Hui
The Fangcheng chapter of the Nine Chapters arranged counting rods in rectangular cells and manipulated columns to solve several unknowns. Liu Hui’s 263 commentary explained why procedures worked. This resembles later elimination, but it was not yet an abstract matrix with modern row-operation notation. Luoyang is an editorial anchor for the Wei intellectual setting; the exact place of composition is unknown.
- Pause and ask
- Why can several unknowns be reduced together after a word problem is moved into counting-rod cells?
- How thinking changed
- Instead of solving each equation alone, align coefficients and values and apply elimination to the whole table.
- What we cannot claim
- Luoyang is an editorial anchor, not proven discovery site. Counting-rod column operations did not already contain the modern matrix object, row notation, or Gauss’s name.
- This place
- The Wei capital’s editorial setting supplies context for Liu Hui’s 263 explanation of accumulated practical procedures through reasons and checks. The exact writing place is unknown. (Luo River · loess terraces · broadleaf trees · 34.6°N 112.5°E)
- Figure board
- A counting-rod table of three columns; repeated column subtraction empties the upper cells until one unknown stands alone.
- On the river
- A gateway another tradition passes through · to 499 (Sandstone stele · braziers · earthen villages and beacons · rafts · flocks of birds)
02·1683 CE·Tokyo(basis: Composition)
Holding Entangled Equations in One Quantity — Seki’s Determinants
Seki Takakazu organized coefficients of equations and presented determinant-like rules in 1683 within Japan’s wasan tradition. His path was independent of Leibniz in Europe. It did not complete modern linear algebra or instantly create today’s determinant method for linear systems.
- Pause and ask
- What can one learn by compressing the entanglement of coefficients into one rule?
- How thinking changed
- Move from calculating individual unknowns to tracking permutations and signs in an array of coefficients.
- What we cannot claim
- Seki’s path was independent of Leibniz but not identical to modern determinant notation or all linear algebra, and no direct route to Cayley is claimed.
- This place
- Edo’s wasan schools, problem tablets, manuscripts, and print networks supported reconstruction of Chinese mathematics and competitive development of independent symbolic procedures. (Tokyo Bay · distant Mt Fuji · broadleaf and pine · 35.7°N 139.7°E)
- Figure board
- A three-by-three array of coefficients, extended by two columns, with every slanted product line carrying a + or − sign.
- On the river
- A desk holding a written record · 1450–1749 (Marble stele · iron lanterns · three-arch bridge · windmills and clock-tower towns · sailing ships)
03·1693 CE·Hannover(basis: Letter sent)
Asking about Solutions through Arrays of Coefficients — Leibniz’s Letter
In a 1693 letter to l’Hôpital, Leibniz gave coefficients two indices and discussed solution conditions through determinant-like expressions. The work was not a widely published standard method in his lifetime, so the letter is not treated as the single beginning of European linear algebra.
- Pause and ask
- How can row-and-column addresses on coefficients shorten a condition for solutions?
- How thinking changed
- Rewrite long equations as doubly indexed coefficients and combinations, exposing the array’s structure.
- What we cannot claim
- A private letter was not a widely read textbook, does not monopolize priority over Seki’s independent work, and is not evidence of a direct line to Cayley’s algebra.
- This place
- A Hanover court and library post plus wide correspondence let Leibniz exchange pre-publication ideas with colleagues across Paris, Basel, and London. (Leine riverside · broadleaf trees · flat land · 52.4°N 9.7°E)
- Figure board
- Three equations with two-digit indexed coefficients become an index table; products taking one cell per row and column state the solution condition.
- On the river
- A post holding tied letters · 1450–1749 (Marble stele · iron lanterns · three-arch bridge · windmills and clock-tower towns · sailing ships)
04·1750 CE·Geneva(basis: Publication)
Making One Ratio for Each Unknown — Cramer’s Rule
In an appendix to his 1750 book on algebraic curves, Gabriel Cramer expressed solutions of a square linear system as ratios of determinants. The rule is elegant for small symbolic examples, but it is not an efficient practical algorithm for large systems.
- Pause and ask
- Can each solution be written as a ratio of determinants when equations and unknowns match?
- How thinking changed
- Replace a full elimination sequence with a general symbolic formula read from coefficient arrays.
- What we cannot claim
- It is an expression for nonsingular square systems, not a recommended large-scale numerical algorithm. Cramer did not invent determinants or linear equations themselves.
- This place
- Geneva’s academy and European print and correspondence networks supported publication and circulation of Cramer’s general rule in a book appendix. (Lake Geneva · Mont Blanc · broadleaf trees · 46.2°N 6.1°E)
- Figure board
- Each unknown is written as a ratio of two coefficient arrays; in the numerator its own column is replaced by the column of values.
- On the river
- A stack of books · 1750–1899 (Cast-iron plaque · gas lamps · iron truss bridge · factory chimneys, railway and steam train · steamboats)
05·1809 CE·Göttingen(basis: Publication)
Eliminating Orbits and Measurement Error — Gauss
Gauss reduced systems of linear equations step by step in astronomical and geodetic computation and connected them with least squares. We now say Gaussian elimination, although related elimination procedures are much older. His achievement was not exclusive invention but integration of calculation with observation and error adjustment.
- Pause and ask
- How can one calculate one persuasive orbit from many imperfect observations?
- How thinking changed
- Combine triangular reduction of linear systems with least squares and error adjustment to handle inconsistent observation.
- What we cannot claim
- Gaussian elimination is a later name. It does not erase earlier elimination including Liu Hui’s tradition, and Gauss’s normal equations are not always the most numerically stable route.
- This place
- Göttingen’s university, observatory, and geodetic work joined repeated celestial and terrestrial observations, calculators, and state surveying needs. (Wooded hills · broadleaf trees · small river · 51.5°N 9.9°E)
- Figure board
- Scattered observations are fitted by one orbital arc, and the linear system behind the fit is reduced step by step to triangular form.
- On the river
- A stack of books · 1750–1899 (Cast-iron plaque · gas lamps · iron truss bridge · factory chimneys, railway and steam train · steamboats)
06·1844 CE·Leipzig(basis: Publication)
Writing Spaces That Could Extend beyond Three Dimensions — Grassmann
Grassmann’s 1844 Theory of Linear Extension treated directed quantities through laws of addition and multiplication in spaces of arbitrary dimension. Its unfamiliar notation and an outsider’s position delayed reception, yet it supplied foundations for vector spaces and exterior algebra. Leipzig marks publication; Grassmann worked in Stettin.
- Pause and ask
- Can addition and multiplication remain lawful when a space has more than three directions?
- How thinking changed
- Treat coordinate collections not as mere lists but as extensible quantities governed by laws independent of dimension.
- What we cannot claim
- Modern vector-space axioms and column-vector notation are not projected unchanged into the 1844 book, and Leipzig publication is not turned into Grassmann’s workplace.
- This place
- Leipzig print made the Stettin teacher’s radical manuscript a book, but could not guarantee rapid reception for work outside established posts and notation. (Riverside woods · broadleaf trees · flat plain · 51.3°N 12.4°E)
- Figure board
- A directed segment sweeps out a parallelogram and then a parallelepiped; the same laws extend on to n directions.
- On the river
- A stack of books · 1750–1899 (Cast-iron plaque · gas lamps · iron truss bridge · factory chimneys, railway and steam train · steamboats)
07·1858 CE·London(basis: Publication)
Making the Number Table Itself an Object of Calculation — Sylvester and Cayley
Sylvester coined the term “matrix” in 1850, and Cayley’s 1858 memoir systematically treated matrix addition, multiplication, and inverses. The fact that reversing order can change a product signaled a new algebra. Their work belongs after long histories of tabular calculation and determinants, not as an isolated invention by two men.
- Pause and ask
- What changes when the number table itself can be added and multiplied as one object?
- How thinking changed
- Move from an auxiliary array for determinants to an independent algebraic object with ordered products and inverses.
- What we cannot claim
- Cayley’s 1858 scene belongs to London activity and publication, not Cambridge. The two men did not replace from nothing the earlier histories of tables, Seki, Leibniz, Cramer, and Gauss.
- This place
- Within London’s legal, Royal Society, and mathematical networks, practicing lawyers Sylvester and Cayley sustained exchanges on invariants and matrices alongside legal work. (Thames banks · broadleaf trees · flat basin · 51.5°N 0.1°W · landmark: St Paul’s Cathedral (1710), Tower of London (1100))
- Figure board
- Two number tables A and B multiplied as whole objects: the products AB and BA come out as different tables.
- On the river
- A stack of books · 1750–1899 (Cast-iron plaque · gas lamps · iron truss bridge · factory chimneys, railway and steam train · steamboats)
08·1901 CE·London(basis: Publication)
Finding the Axes That Best Explain a Cloud of Points — Pearson
Pearson posed the problem of fitting the closest line or plane to observations in many dimensions. Together with Hotelling’s 1930s formulation, it became the language of principal component analysis: summarizing variation through a few directions. Pearson’s biometric work must be read alongside his eugenics, and the 1901 paper was not yet all of modern PCA.
- Pause and ask
- Can a high-dimensional cloud be laid onto one or two directions with the least loss?
- How thinking changed
- Replace one-variable-at-a-time inspection with the whole covariance structure and its closest-fitting axes.
- What we cannot claim
- The 1901 paper is not all of modern PCA and does not erase Hotelling’s later formulation. Statistical axes do not discover neutral natural types of people.
- This place
- University College London’s biometric laboratory combined biological data, computing labor, and a statistics journal while also strengthening eugenic classification and institutional power. (Thames banks · broadleaf trees · flat basin · 51.5°N 0.1°W · landmark: Big Ben (Elizabeth Tower) (1859), St Paul’s Cathedral (1710))
- Figure board
- The line lying closest to a point cloud, judged by perpendicular distances, captures most of its variation.
- On the river
- A stack of books · 1900–1969 (Concrete marker · electric streetlights · concrete bridge · pylons, apartment blocks and radio masts · barges · aircraft)
09·1925 CE·Göttingen(basis: Publication)
Turning Observable Quantities into Multiplication Tables — Matrix Mechanics
Heisenberg computed atomic spectra using arrays of transition amplitudes; Born recognized matrix multiplication, and work by Born and Jordan and then all three developed matrix mechanics. Noncommutative multiplication, where order changes the result, became part of the structure of quantum observables. This was a collaborative development, not one isolated flash.
- Pause and ask
- What multiplication appears if a table keeps observable transitions instead of invisible electron orbits?
- How thinking changed
- Move from classical orbit pictures to arrays of transition amplitudes and accept a noncommutative structure where multiplication order matters physically.
- What we cannot claim
- Heisenberg is not made the sole inventor of matrix mechanics. The matrix formalism is one representation of quantum theory and does not settle every interpretive question about measurement.
- This place
- Göttingen’s theoretical-physics seminar and Born’s group supplied the density of collaboration needed to connect Heisenberg’s arrays with Jordan’s matrix knowledge and test them in papers. (Wooded hills · broadleaf trees · small river · 51.5°N 9.9°E · landmark: Göttingen Observatory (1816))
- Figure board
- Transitions between levels are entered cell by cell in a table that is multiplied as a whole — and the order of multiplication matters.
- On the river
- A stack of books · 1900–1969 (Concrete marker · electric streetlights · concrete bridge · pylons, apartment blocks and radio masts · barges · aircraft)
10·1936 CE·Chicago(basis: Publication)
Reducing a Large Table to Its Most Important Layers — Eckart and Young
Eckart and Young stated how to find the best lower-rank approximation to a matrix. Combined later with singular value decomposition, the principle became central to image compression, denoising, recommendation, and latent structure. Their 1936 paper did not invent SVD or every modern compression technique.
- Pause and ask
- Which information should remain first when a large matrix is replaced by fewer directions?
- How thinking changed
- Instead of preserving every entry equally, retain the largest singular directions to obtain the least-error low-rank approximation.
- What we cannot claim
- The paper did not first discover SVD and was not a direct blueprint for JPEG or recommenders. “Important information” depends on the chosen error metric and representation.
- This place
- The University of Chicago’s mathematical-physics setting supported turning orthogonal-transformation questions from vibration and quantum calculation into a general approximation theorem. (Lake Michigan · broadleaf trees · flat prairie · 41.9°N 87.6°W · landmark: Chicago Loop skyscrapers (1892))
- Figure board
- A large table splits into rank-one layers of shrinking weight; keeping the two largest gives the closest lower-rank approximation.
- On the river
- A stack of books · 1900–1969 (Concrete marker · electric streetlights · concrete bridge · pylons, apartment blocks and radio masts · barges · aircraft)
11·1947 CE·Washington DC(basis: Classified research)
Turning Resources and Goals into a Table of Constraints — Dantzig and the Simplex Method
For US Air Force planning, Dantzig and colleagues represented production and transport constraints and objectives as linear expressions and developed the simplex method in 1947. Matrices made complicated choices computable, but cannot decide values omitted from an objective function.
- Pause and ask
- How can feasible and better plans be separated amid thousands of resource constraints?
- How thinking changed
- Rewrite plans as matrix inequalities and an objective, then improve along vertices of the feasible polytope.
- What we cannot claim
- The scene does not erase Soviet linear-programming precedents or team implementation by making one isolated invention. An optimum is optimal only for the stated objective and constraints.
- This place
- Project SCOOP at the Pentagon concentrated postwar Air Force personnel, equipment, transport planning, and computation, making large linear programs an operational problem. (Potomac banks · broadleaf trees · low hills · 38.9°N 77.0°W · landmark: United States Capitol (1866))
- Figure board
- Each row of an inequality table becomes one edge of the feasible polygon; the plan improves by moving from vertex to vertex.
- On the river
- A sealed box · 1900–1969 (Concrete marker · electric streetlights · concrete bridge · pylons, apartment blocks and radio masts · barges · aircraft)
12·1961 CE·Teddington(basis: Publication)
Driving a Matrix toward a Triangle to Find Eigenvalues — The QR Algorithm
At Britain’s National Physical Laboratory, Francis, and in Leningrad, Kublanovskaya, independently reached QR iteration around 1961. It became a leading stable method for finding eigenvalues of large matrices. Teddington marks only one of the two centers.
- Pause and ask
- What must change so a computer can iterate toward eigenvalues of a large matrix without runaway error?
- How thinking changed
- Factor a matrix into orthogonal Q and triangular R, reverse the product, and iterate toward a form that reveals eigenvalues.
- What we cannot claim
- Teddington is not the sole birthplace: Kublanovskaya published independently in Leningrad. QR factorization and the iterative QR algorithm are also not identical terms.
- This place
- NPL’s early computers and numerical-analysis community supplied machines and problems for testing a theoretical factorization under real precision and cost. (Thames banks · broadleaf trees · flat parkland · 51.4°N 0.3°W)
- Figure board
- Factor into Q and R and multiply back in reverse, again and again: the entries below the diagonal fade and the eigenvalues appear.
- On the river
- A stack of books · 1900–1969 (Concrete marker · electric streetlights · concrete bridge · pylons, apartment blocks and radio masts · barges · aircraft)
13·1963 CE·Cambridge, MA(basis: Composition)
Becoming the Transformation That Moves a Line on Screen — Sketchpad
Sutherland’s MIT doctoral project Sketchpad let a user draw with a light pen, impose constraints, and copy, rotate, or scale geometry. Homogeneous coordinates and transformation matrices became a consistent grammar for moving screen objects. One program is not claimed as the sole origin of every graphical interface.
- Pause and ask
- What should be stored to rotate or move a whole shape without relocating thousands of points one by one?
- How thinking changed
- Represent a shape not as a pixel list but as coordinate vectors joined to transformation matrices and constraints.
- What we cannot claim
- Sketchpad was pivotal but does not make one person the sole inventor of every GUI, CAD system, or computer graphic. Cambridge marks the MIT research context.
- This place
- The TX-2, light pen, interactive computing resources at MIT Lincoln Laboratory, and a doctoral setting enabled direct geometric dialogue with a screen. (Charles River · broadleaf trees · flat land · 42.4°N 71.1°W · landmark: MIT Great Dome (1916))
- Figure board
- The shape’s vertices are stored as (x, y, 1); one 3×3 matrix rotates and scales the whole shape at once.
- On the river
- A desk holding a written record · 1900–1969 (Concrete marker · electric streetlights · concrete bridge · pylons, apartment blocks and radio masts · barges · aircraft)
15·2017 CE·Mountain View(basis: Main activity)
Comparing Context through Matrices — The Transformer
The Transformer multiplies query, key, and value matrices derived from token representations to calculate what each token should attend to. Large parallel matrix operations helped scale translation and generative AI, but matrix multiplication alone does not explain language understanding. Mountain View is an institutional anchor for the Google research network, not a replacement for the authors’ multiple affiliations.
- Pause and ask
- Which three tables let every token in a sentence compare itself with every other token at once?
- How thinking changed
- Move from passing recurrent state step by step to attention that computes query–key score matrices and weighted values in parallel.
- What we cannot claim
- Matrices are an efficient computational representation, not a full explanation of understanding, reasoning, or creativity. Mountain View is a Google network anchor and does not erase Toronto affiliation or earlier attention work.
- This place
- Google’s translation data, TPU and GPU computation, multiple research teams, and conference publication supplied institutional conditions for testing matrix-shaped attention at scale. (Near the bay · oaks · flat valley floor · 37.4°N 122.1°W)
- Figure board
- Multiplying Q by K gives a score table of how much each token attends to every other; those weights mix the rows of V.
- On the river
- A place of ongoing work, marked only by the route’s emblem · 1970 onward (Glass marker · LED lights · cable-stayed bridge · glass towers, wind turbines and data centres · ferries · satellites)