Recommended Free Tools
Meta says mathematicians used Muse Spark in the regular meta.ai chat interface to work on six research papers, five addressing questions it describes as previously open. The work was a collaboration: researchers chose and guided the problems, developed and checked arguments, and reviewed the resulting papers. It is not evidence that the AI independently solved six problems.
What Meta reported—and what “regular chat window” means
In an announcement published October 2, 2026, Meta AI Research said mathematicians worked with Muse Spark 1.1 and 1.2 in Thinking Mode through the regular meta.ai chat interface. Meta says they did not use a custom research scaffold. The collaboration produced six papers spanning probability, differential equations, group theory, optimization, arithmetic physics and non-associative algebra; Meta characterizes five as answers to previously open questions.
As an Amazon Associate I earn from qualifying purchases.
That describes the interface, not an absence of human expertise or a special autonomous theorem-proving system. Meta says researchers guided exploration and argument development, and a second group reviewed the work. The papers also mark passages primarily drafted by researchers or AI and credit prior research. Meta describes contributions ranging from candidate proofs and technical drafting to calculations and revisions; for the group-theory work, it specifically says Muse Spark generated a GAP search program, which the mathematicians checked and used in completing the argument.
Meta’s institutional announcement says: “Our goal here wasn’t to mass-produce papers, but to empower researchers and help them develop mathematical insights that others can understand and build on.” That is Meta’s stated aim, not an independent assessment of the papers’ correctness or of the model’s general mathematical ability.
#1 Best Overall
What the six papers address
| Area | Paper’s question | Kind of result |
|---|---|---|
| Probability | When can Gaussian points lie on a centered ellipsoid? | Asymptotic threshold |
| Differential equations | Must certain radial negative-energy solutions blow up in finite time? | Blow-up theorem |
| Group theory | Must every finite semiabelian group be monomial? | Counterexample |
| Optimization | When is a cycle-based relaxation exact? | If-and-only-if characterization |
| Arithmetic physics | How are a string calculation and a curve’s height function related? | Connection between two calculations |
| Non-associative algebra | Does a proposed test identify the intended class of evolution algebras? | Counterexample and alternative rule |
Probability: fitting a Gaussian ellipsoid
Aykut Arslan’s paper studies independent standard Gaussian vectors and asks whether one centered ellipsoid, represented by a positive semidefinite matrix, can fit all of them. It reports a sharp asymptotic threshold near n = d2/4: below that ratio, a fitting positive definite matrix exists with probability tending to one; above it, no fitting matrix exists with probability tending to one. The paper makes no claim for the boundary case where the ratio tends to the threshold.
This result was not exclusive to Meta’s collaboration. Meta acknowledges three independent papers posted in August 2026: Misiakiewicz and Wen independently proved the Gaussian threshold; De la Cerda, Potechin, Tulsiani and Xu established it up to a vanishing multiplicative factor; and Koehler and Sohn obtained a broader universality result that includes the Gaussian threshold as a special case. Meta says these groups worked independently using different approaches.
Differential equations: finite-time blow-up
Leonard Dinh’s paper concerns the focusing mass-critical biharmonic nonlinear Schrödinger equation. Its theorem is specifically about radial solutions with negative energy and initial data in H2(RN), for dimensions N ≥ 2: each such solution blows up in finite time both forward and backward. In this setting, the result rules out blow-up occurring only at infinite time. Meta describes the question as long-standing since 2015 and says simulations had predicted the result in 2002; those historical descriptions are Meta’s account.
Group theory: a finite counterexample
Joseph Phillip Brennan and Milana Golich’s paper disproves M. Kida’s conjecture that every finite semiabelian group is monomial. It gives a semiabelian, non-monomial group of order 384, identified in GAP’s SmallGroups library as SmallGroup(384, 20127). Meta says Muse Spark generated the GAP search program; the mathematicians verified the counterexample and completed the argument. Meta also notes that the AI agent Nilradical reported a different counterexample independently on September 16, 2026.
Rank #3
Optimization: when a relaxation is exact
Arslan’s optimization paper studies a cycle-based relaxation for binary polynomial optimization. For the completed support of one length-three alpha-cycle, it gives an exact condition: the relaxation equals the multilinear polytope if and only if each of the three pairwise-only intersections has size one. This is a characterization for that specified cycle structure, not a general claim that the relaxation is exact for every optimization problem.
Arithmetic physics: a string function and a curve
Anindya Dey, Gabriel Herczeg, An Huang, Nicolas Jaramillo Torres and Jacob H. Swenberg connect a string two-point function with a height function on a curve. Meta says the work extends a known connection for the Tate curve to a broader class of curves, linking number theory and p-adic string theory. In simpler cases, the calculation can be understood through how many initial base-p digits two point coordinates share.
Rank #4
- Exercise your mind with this collection of brainteasers, logic puzzles, and more! 359 puzzles
Non-associative algebra: a proposed test fails
Andres Barei’s paper gives a three-dimensional evolution algebra that passes a proposed solvability test but does not belong to the class the test is meant to identify. It also proposes an alternative rule based on whole subspaces. Meta acknowledges independent counterexamples by Hu and Wen, so the result should not be presented as an uncontested first.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →What the report does—and does not—show
The examples show several distinct forms of mathematical assistance: a threshold analysis, a theorem in a narrowly specified equation and solution class, counterexamples to conjectures or criteria, an exactness characterization, and a connection between calculations in two fields. They do not establish that Muse Spark can autonomously choose important questions, produce reliable proofs without expert review, or solve open mathematics generally. The papers concern bounded technical claims, and the ellipsoid paper expressly leaves its threshold boundary undecided.
The most supportable reading is that a conversational AI interface served as one tool in researcher-led mathematical work. The reported interface was ordinary meta.ai chat, but the contributions and safeguards were substantive: mathematicians supplied direction, checked arguments, revised the work and reviewed the papers. The concurrent results in probability, group theory and algebra also show why “helped tackle” is more accurate than claiming that Meta’s model alone discovered every result.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
