Recursive Asterisms

AI-assisted mathematical discovery

Turn questions into
mathematical discoveries.

Recursive Asterisms is an AI-assisted mathematical discovery platform in development. It is designed to help people investigate mathematical questions, uncover relationships, develop conjectures, and connect results to the reasoning and evidence behind them. Its purpose is to make mathematical discovery easier to explore, understand, verify and build upon.

The ambition is a shared mathematical world where an idea can grow from a conjecture into a result, with its evidence and history intact. Researchers, students and curious explorers could follow the same inquiry at different depths.

The project creator

Ash Shaver.

Creator of Recursive Asterisms and author of its opportunity thesis. The proposed product grew from firsthand experience carrying mathematical inquiry across AI conversations, notes and checking tools.

Recursive Asterisms is presented as an Anavi app. The initial work brings the product idea and an interactive mathematical exhibit together; the next stage needs expertise in research software, formal methods, product engineering and independent assessment.

Investment entity, asset ownership and team commitments are part of the diligence discussion. No outside appointments, credentials or advisory relationships are implied here.

Read the investor briefing →
[…] Who has put me here? By whose order and direction have this place and time been allotted to me? […]
Blaise Pascal · Pensées, fragment 205
Read the context & source

Blaise Pascal, *Pensées*, fragment 205 in this edition; English translation by W. F. Trotter. E. P. Dutton, 1958; Project Gutenberg ebook 18269.

Pascal considers a brief human life against immense time and space, within a Christian apologetic work. The questions retain anxiety and theological context, not just wonder. Fragment numbers vary by edition. The well-known infinite-silence sentence belongs to 206 here, not 205.

Gutenberg text · 1958 edition’s library record

Author’s viewpoint, not proof evidence or an endorsement of RA.

Different contexts. Explicit connections.

“Multiverse” can name the branching worlds of a mathematical inquiry: alternative axioms, spaces or representations. A theorem in one axiom system does not silently become a theorem in another.

RA’s proposed meta-level asks what a representation preserves and what evidence would change a claim. A resemblance between mathematical fields starts a question; it is not proof of transfer.

From a picture to its mathematical structure

An ellipse.
A hidden pair of directions.

2x² + 2xy + 2y² = 3u² + v²

Turn the coordinates through the eigenvector directions: u = (x+y)/√2 and v = (x−y)/√2. A coupled expression separates. Geometry and linear algebra describe the same object in different ways.

This established connection illustrates the intended experience: begin with a shape, ask what lies beneath it, develop a candidate explanation, inspect the mathematics and pursue the next question.

Explore the spectral connection ↗

Contribution concept · no submission service

Build on a result.
Keep its history.

01 Draft

Author credit, source provenance, a stable version ID and a naming proposal.

02 Formal check

Exact environment, declared assumptions and recorded outcome.

03 Review

Intended meaning, prior art, compatibility and explanatory accuracy.

04 Accepted

Admission to a declared collection; linked corrections and affected dependents.

Passing Lean alone does not establish novelty or collection acceptance. Names do not establish priority, and popularity does not establish truth. This proposed process draws on Mathlib review practice.

The first discovery pilot.

We are seeking research partners to test whether RA helps turn a mathematical question into a useful, supported insight. A pilot would follow candidate claims, counterexamples, reasoning and reuse, comparing the whole investigation with the group’s existing workflow.

The initial customer hypothesis is a small mathematics research group already using AI and formal tools. Proposed paid workspaces would bring inquiry, evidence and collaboration together. The pilot would compare expert-reviewed results, total effort and willingness to pay; no customer outcomes are claimed yet.

Copy a pilot discussion brief
Development status and privacy

This interactive preview demonstrates the experience through established mathematics. Integrated discovery, formal proof search and shared contribution services remain proposed capabilities. AI assisted the content, design and website implementation.

The preview accepts no research submissions or payments. Its controls work in your browser; only your sky-motion preference is stored. A copied companion prompt runs in the AI service you choose, under that service’s policies.