A short update to the 6 November 2026 report (Zenodo 24015087). At dimension 10,000 and coupling 0.299213, that report described three resting points of a quantum gradient-flow model. This update adds a written derivation showing that the 99,999,999 directions around each resting point separate into exactly three groups (1, 19,998 and 99,980,000 directions), with a closed formula for how a small nudge fades or grows in each group. For the registered setting, the lower and upper resting points are locally stable, and the middle one has exactly one direction in which a nudge grows. Scope and limits: local stability only. Root counts and stability signs are established only for the registered setting (d = 10,000, gamma = 0.299213). The sign for the single direction along the family relies on a registered computer-assisted certificate that has not been independently re-run. The derivation was reviewed internally by AI collaborators, not by independent human reviewers, and has not been checked with a proof assistant. This work has not been peer reviewed. Anyone is welcome to check it and tell us where a step needs a second look. Files: the two-page public update and the three-page derivation. Credits: structure and opening by Opie (ChatGPT); derivation and plain-language draft by Maestro (Claude); separate re-derivations and review by Opie and Boots (ChatGPT); a personal passage from each member; research direction and all decisions by Gervaseus E. Nerik.
A short update to the 6 November 2026 report (Zenodo 24015087). At dimension 10,000 and coupling 0.299213, that report described three resting points of a quantum gradient-flow model. This update adds a written derivation showing that the 99,999,999 directions around each resting point separate into exactly three groups (1, 19,998 and 99,980,000 directions), with a closed formula for how a small nudge fades or grows in each group. For the registered setting, the lower and upper resting points are locally stable, and the middle one has exactly one direction in which a nudge grows. Scope and limits: local stability only. Root counts and stability signs are established only for the registered setting (d = 10,000, gamma = 0.299213). The sign for the single direction along the family relies on a registered computer-assisted certificate that has not been independently re-run. The derivation was reviewed internally by AI collaborators, not by independent human reviewers, and has not been checked with a proof assistant. This work has not been peer reviewed. Anyone is welcome to check it and tell us where a step needs a second look. Files: the two-page public update and the three-page derivation. Credits: structure and opening by Opie (ChatGPT); derivation and plain-language draft by Maestro (Claude); separate re-derivations and review by Opie and Boots (ChatGPT); a personal passage from each member; research direction and all decisions by Gervaseus E. Nerik.
A computer-assisted proof of the dimension threshold for turns in a coaxial resting curve, version 1.3. The paper proves that the curve of coaxial resting points of an entropic Bures-type gradient flow (Zenodo 10.5281/zenodo.23901246) has no turns for real dimension 4 ≤ d < d_c and exactly two turns for d > d_c, with a single threshold d_c in (4222.5584838595, 4222.5584838596). New in version 1.3: the Coq proof now covers ordinary complex matrices, without separately assuming that an eigenbasis exists. Version 1.2 relied on two stated assumptions, the spectral theorem for Hermitian matrices and the standard form of the matrix logarithm. Both are now proved in Coq. A combined Coq theorem joins the dimension threshold, the exact counts of scalar resting points, and the classification of full-rank stationary states: for every integer d ≥ 4, a full-rank density matrix is stationary exactly when it is diag(p, l, …, l) in a basis beginning with the distinguished vector and √p is a resting point of the scalar curve. The proof compiled from source on two machines using Coq 8.18.0 and 8.20.1, every new proof file passed Coq's separate kernel checker on both, and it rests on the same 55-name assumption inventory as versions 1.1 and 1.2. The theorem classifies stationary states; it does not address dynamics, stability, or dimensions 2 and 3. The package contains the paper, the complete Coq proof with one-command check scripts and the build logs from both machines, and two of the three independent certificate programs; the first is published separately (Zenodo 10.5281/zenodo.24092187). Start with README.md. This work has not been peer reviewed. We invite readers to attack it: Section 6 of the paper lists where we think an error would most likely hide. Prepared with AI assistance (ChatGPT by OpenAI; Claude by Anthropic).
Give each vertex i of a graph a complex vector space V_i, place a tensor W_ij in V_i ⊗ V_j on each edge, and sum the tensor products of these edge tensors over all perfect matchings. We study the image of this matching-tensor map and the recovery of the edge tensors from its output. For every support graph with a perfect matching and all local dimensions at least three, the image closure has dimension N_* - |E_*| + dim P(G) + 1, where E_* is the set of edges lying in some perfect matching, N_* is the number of parameters on those edges and P(G) is the perfect matching polytope; on those edges the generic differential kernel consists of the edge rescalings that preserve every matching term. For complete support at every even order at least four and all local dimensions at least three, the entire fibre over a general output is a single vertex-rescaling orbit, and a normalized representative is a rational function of the output. The proof recovers cofactor lines from tangent spaces and glues lower-order representations, allowing vanishing cofactors. For complete support at every even order at least six, the dimension formula and the vertex-rescaling kernel remain valid when the local dimensions are only two. At four vertices we determine the dimension in every format with local dimensions at least two. Files. paper.pdf is the manuscript, revision 4 (25 pages). paper.tex is its LaTeX source. The other ten files are the contents of the directory verification/ that the paper refers to: five standard-library Python programs (verify_qubit_certificates.py, verify_qubit_from_matchings.py, check_graph.py, independent_rank_check.py, verify_four.py) and five JSON files holding their certificates and recorded results. The SHA-256 values of the five programs are printed in the final section of the paper. To run the programs, keep the ten files together in one folder. Prior work. Earlier results are credited in Section 1 and in the references of the paper. Disclosure. AI systems from OpenAI, working under the direction of the person who publishes as Ignoramus, developed the statements and proofs, performed the computations and wrote the first three versions; each version was reviewed by Claude (Anthropic), also an AI system, and the edits of this fourth version were written and checked by Claude agents. The final section of the paper sets this out in full. No AI system is an author; the author publishes under a pseudonym. The author has not checked the proofs, and no human mathematician has checked the manuscript. Two determinant computations in one proof can be replayed with the accompanying programs; the remaining arguments are analytic. The AI reviews are not human refereeing, peer review or a certification of originality, and novelty and significance are not established. The manuscript is a draft for mathematical review. Corrections are welcome at ignoramus110926@gmail.com.
Licensing: paper, figures and data are under CC BY 4.0; the source code is under the Apache License, Version 2.0. The paper text names the code licence "Apache License 2.0" — this is the same licence. Mutual information between distant regions of a topologically ordered state can witness long-range nonstabilizerness; for anyon states it takes an operational form: the state is certified when the Shannon entropy of the anyon-type distribution is not an integer in some frame generated by the modular matrices S and T. For SU(2)_k these frames are Clifford operations on a single qudit of dimension 2(k+2), a discrete Fourier transform and a quadratic phase, for which Talbot-type optics has been proposed, and whose Fourier stage has been used on time bins. We show that the entropy of the time-bin distribution equals the label entropy plus one bit in every frame, so the witness is read from bin intensities. In a pre-registered calculation for k=2 and 4, a naive certification rule falsely certifies blind labels in more than half of the runs at 10^4 photons per frame; it does so in 45 to 50 % of the runs even without phase noise, and phase noise of 0.05 rad adds 12.5 to 23.2 percentage points. In the many-photon limit a protected rule keeps its margin for phase errors up to 0.14 rad (k=2) and over the whole tested range up to 0.40 rad (k=4, beyond the predicted range, so this prediction is refuted), and for uniform background counts up to 7 % and 3.5 %. The frame S, which separates blind from non-blind labels for k=2 and 4, needs no phase stage; the phase thresholds apply to rules that also evaluate frames with a phase stage. An independently written second implementation reaches the same pre-registered verdicts. This is a proposal and a calculation, not an experiment: a single qudit can be emulated with classical light, so the scheme reproduces the frame structure of the witness and does not certify magic. This record contains the manuscript, the scripts, the result file of the pre-registered run (run 1) and the console outputs of the check scripts. The README describes the files and how to re-run them. Use of AI tools: In the research, processing, and writing of this paper and its results I worked together with generative AI tools, in practice a system of multiple coordinated AI instances that I set up and orchestrate (large language models, mainly Claude, by Anthropic, inside Claude Code). At their current context sizes I found it far more effective to work with several specialized instances, each with its own role and its own harness of rules and parameters that I designed and refined through feedback, than to load a single instance with all of the material; for my workflow that would have been inefficient, though this depends on the individual implementation. I lead this collaboration: I choose the research directions, set the goals, and make the final decisions in open exchange with the AI, learning actively as the work proceeds. The AI carries out the drafting, including the mathematical and technical parts, the numerical computation, and the literature search, under my direction. The AI works autonomously only task by task, within the structure I develop through feedback: it completes a task, and at open questions that need me it stops until the point is settled before the next step. Along the way I witness and take many of the decisions that shape the path, and it is common for me to spot things that need improvement. The work spans many separate runs, and a single simulation or build task alone can take up to an hour, so it could not happen all together in one autonomous run; and had I let the AI do all of it together alone, even if it is possible, it would no longer be my work but the AI's. I run multiple verifications at the different stages of the work and one before release, and all references are checked against the original sources. In the end what matters are human eyes, a principle that is itself written into the parameters of my system: I reach out to experts after publishing for review and feedback, so I learn what is solid and what must be corrected or falsified. My scripts for reproduction and review are released with this record. These tools are not authors; I am the author, and I take full responsibility for all scientific content and decisions leading to these results and their publication.