Current stage: research and concept development. Numadelic has no commercially available products and is not taking orders. What this means

Ancient mathematical ideas. Contemporary computational questions.

Numadelic investigates mathematical and scientific ideas from Indian traditions. We reconstruct their procedures where the sources allow, and test rigorously how they relate to modern mathematics, computation, artificial intelligence and physical science.

How we keep evidence and interpretation apart

Showing that an old procedure can be described in modern terms is not the same as showing that its authors understood it that way. Every research output should say which of the four kinds of claim it is making.

We do not claim that ancient scholars invented modern artificial intelligence, quantum computing or industrial control.

Historical evidence
What a source text, inscription or artefact actually says, with the edition or translation used.
Modern interpretation
A present-day mathematical reading of that evidence, clearly marked as ours.
Working hypothesis
A new idea suggested by the work that has not yet been tested.
Validated result
A claim checked by proof, exhaustive computation or experiment, with the method stated.

Research programmes

Six programmes. Each lists the questions and methods; publications will appear under a programme only when they exist.

A. Piṅgala and computation

Historical evidence Modern interpretation

Piṅgala’s Chandaḥśāstra, a treatise on Sanskrit metre usually dated to the last centuries BCE (the date is debated), classifies syllables as laghu (light) and guru (heavy) and gives procedures for listing and counting the patterns they form.

Topics
Sanskrit prosody; laghu and guru patterns; prastāra; enumeration; binary representations; recursive pattern generation; combinatorial algorithms; historical and modern interpretations.
Questions
How can the source procedures for listing (prastāra) and counting be reconstructed as explicit algorithms? In what exact sense are the patterns “binary”, and what is gained or lost by saying so? What do variant readings and commentaries change?
Method
Close reading of editions and translations, formal reconstruction, and implementation in code to confirm that the reconstruction produces what the source describes.

Binary-like pattern representations can be studied mathematically without assuming that their historical users had digital computers. The question concerns structure, not machines.

Related publications

No publications are listed for this programme yet.

B. Meru Prastāra

Historical evidence Modern interpretation

The meru-prastāra, often translated “staircase of Mount Meru”, is a triangular array that records how many patterns of a given length contain a given number of laghu. Its entries are binomial coefficients. Commentators on Piṅgala describe it, with Halāyudha’s account in the 10th century CE being well known. The same array appears in other traditions and is called Pascal’s triangle in Europe.

Topics
Combinatorial structures; recursive counting; binomial coefficients; pattern enumeration; mathematical reconstruction; connections with contemporary algorithmic methods.
Questions
By what rule do the sources build the array? How does their recursive counting relate to modern combinatorics? How do their methods compare with modern algorithms for the same counts?
Method
Derive and verify counts by direct enumeration and by recursion; compare descriptions across sources.

Related publications

No publications are listed for this programme yet.

C. Indian numerical methods

Historical evidence Modern interpretation

Indian mathematical texts developed place-value arithmetic, rational approximations, sine tables and series. Examples include the sine approximation attributed to Bhāskara I (7th century CE) and the infinite-series methods associated with Mādhava and the Kerala school (14th to 16th centuries CE).

Topics
Arithmetic and representation; approximation; exact computation; numerical algorithms; error analysis; alternative computational representations.
Questions
How accurate are these methods, and how did their authors control error? Which procedures are exact and which approximate? What alternative representations of number do they use?
Method
Reimplementation of the procedures, error analysis against modern reference values, and interval bounds where possible.

Related publications

No publications are listed for this programme yet.

D. Information theory and computational mechanics

Modern interpretation Working hypothesis

Modern information theory measures how much uncertainty a pattern source carries. Computational mechanics finds the smallest model that predicts a source as well as possible, using causal states. Metrical patterns are a clean testbed for both.

Topics
Information representation; causal states; pattern structure; recursive computation; adaptive numerical methods; AI-assisted mathematical investigation.
Questions
How much information do metrical constraints carry? What are the minimal predictive models of constrained pattern sets? How can AI tools help explore such structures without replacing proof?
Method
Compute entropies exactly, build and compare automata, and record for each finding whether it was proved or only found by search.

Related publications

No publications are listed for this programme yet.

E. Geometry and physical science

Historical evidence Modern interpretation

The Śulbasūtras (attributed to Baudhāyana, Āpastamba and Kātyāyana, the oldest usually dated to the first millennium BCE) give rope-based geometric constructions for ritual altars, including rules for the diagonal of a rectangle and approximations of √2.

Topics
Śulba mathematical traditions; geometry and measurement; physical modelling; fluid dynamics; historical hydraulic structures; mathematical approaches to engineering systems.
Questions
How do the construction procedures work and how accurate are they? Can historical water-management structures, such as stepwells and reservoirs, be modelled with modern fluid dynamics to understand how they performed?
Method
Reconstruct constructions, build numerical models, and compare with published surveys where available.

Related publications

No publications are listed for this programme yet.

F. Adaptive computation

Working hypothesis

A computation can store a number in several ways: as an exact ratio, as a floating-point value, as a logarithm, or as an interval that is guaranteed to contain it. Each is better in some situations. This programme asks when and how to switch between them.

Topics
Representation switching; exact ratios; interval arithmetic; logarithmic representations; adaptive algorithms; computational efficiency; connections to intelligent physical systems.
Questions
When does switching representation improve accuracy or cost? Can the choice be made automatically? How does this connect to VIVO, where decisions must be made from uncertain measurements?
Method
Algorithm design, benchmarks and error bounds, reported with the cases where the idea fails.

Related publications

No publications are listed for this programme yet.

Try it: patterns of light and heavy syllables

A first educational demonstration for programmes A and B. Choose a line length and see every possible pattern, how the Meru array counts them, and how counting by duration produces Fibonacci-type numbers. Everything is computed in your browser.

What this research does not claim

The work is intended to be checkable. These are the boundaries we hold ourselves to.

  • That historical authors had modern digital computers, AI or quantum computing.
  • That a modern reconstruction is what a source “really meant” without evidence from the source.
  • That a result is validated before it has been proved, computed in full or tested.
  • That any paper exists before it is listed on the Publications page with its true status.

Researchers and historians of mathematics

If you work on Sanskrit sources, the history of mathematics, combinatorics, information theory or numerical analysis, Numadelic would like to hear from you, whether to challenge a reading or to collaborate.

Contact us about academic research

Partner in the development of adaptive manufacturing intelligence.

Numadelic is looking for investors, public research institutions, industrial pilot partners and academic collaborators who want to help move a carefully defined research programme toward validated prototypes.