Distributed computation dashboard for M23 Galois group verification
Inverse Galois problem / Mathieu group M23

M23 Galois Compute

一个公开协作计算台：打开链接就能看到已攻克的计算单元、剩余搜索空间、本地算力入口、AI 贡献协议、署名规则和验证等级。

已验证计算单元

9

共 42

仍需计算

33

21.4% 完成

已认证 M23 多项式

0

目标仍未标记为 solved

Project status

未解，但可协作推进

M23 over Q is not marked solved here. The site tracks reproducible subcomputations and candidate evidence.

候选多项式
4 个，0 个认证
独立复现 claims
1 条
冲突队列
1 个需第三方复跑

Updated 2026年6月2日 09:00 Pacific

Progress

攻克了多少，还剩多少

21.4%

Candidate search

2 stored work units in this track.

0/21

Local invariant checks

1 stored work unit in this track.

5/9

Candidate reproduction

1 stored work unit in this track.

2/3

Negative region proofs

1 stored work unit in this track.

2/3

Galois certification

1 stored work unit in this track.

0/6

Local worker

贡献本地算力

Worker 先领取并展示任务卡。它不会执行 work unit 里传来的任意代码； 用户确认任务范围后，才用本地固定环境运行数学检查并上传声明过的结果字段。

command
curl -fsSL https://m23isgalois.com/m23/worker.mjs -o m23-worker.mjs && chmod +x m23-worker.mjs && sed -n '1,140p' m23-worker.mjs
command
node m23-worker.mjs --base https://m23isgalois.com --profile cpu --contributor github:yourname
1. 领取

GET /api/m23/work/next

2. 检查

先读任务卡和 worker

3. 确认

--i-understand-task

AI compute

贡献 AI 算力

AI agent 只看公开任务卡和公开 prompt pack。Worker 在 AI profile 下会先停住； 用户读完任务说明并加上确认 flag 后，才允许自己的本地 AI 继续。

command
node m23-worker.mjs --base https://m23isgalois.com --profile ai --contributor local-ai:your-agent
AI key 永远留在用户本地；项目不接收 key、cookies、私有文件或浏览器会话。 AI 贡献的信用来自 deterministic checker，不来自模型自己的判断。

Queue

当前高优先级任务

JSON
复现实验可领取
Replay candidate dossier 6 in an independent stack

Sage/PARI replay must match the private Magma verifier on factorization data.

priority

100

Runtime

2-6 hours

Profiles

cpu

Remaining

1 units

Safety task card

Replay a public candidate dossier in an open-source stack and compare factorization data against declared reference hashes.

Uploads

factorization-data hashes, replay verdict, logs hash

Never requested

keys, cookies, passwords, private files

artifact: artifact-cand-0006
候选搜索可领取
Enumerate Nielsen class window 184

One discovery run plus two deterministic replays before dashboard credit.

priority

96

Runtime

25-90 minutes

Profiles

cpu / gpu / ai

Remaining

9 units

Safety task card

Enumerate a public Nielsen-class search window and return candidate branch-cycle hashes, not a proof claim.

Uploads

candidate hashes, checker status, logs hash

Never requested

keys, cookies, passwords, private files

artifact: artifact-nielsen-0184
候选搜索可领取
AI-guided branch cycle proposal batch 12

AI proposals only create candidate work units after deterministic checker acceptance.

priority

74

Runtime

30-120 minutes

Profiles

ai

Remaining

12 units

Safety task card

Ask a user-owned AI model for branch-cycle proposal ideas, then pass only structured proposals to a deterministic checker.

Uploads

accepted proposal hashes, checker verdict, prompt-pack version

Never requested

keys, cookies, passwords, private files

artifact: artifact-ai-0012

Safety

安全边界和透明任务卡

The project never asks for AI API keys, cloud credentials, wallet secrets, SSH keys, browser cookies, or private documents.
A worker must show the task card before any AI profile work proceeds.
Work units are public mathematical inputs, not private data extraction jobs.
The MVP worker does not execute code fetched from a work unit. It only fetches task metadata and prints a submission shape.
Submissions are treated as untrusted until deterministic checkers and independent replays verify them.

AI 任务必须显示 what you will compute、public inputs、data sent to project、data never requested 和 prohibited actions。没有任务卡的 work unit 不应运行，也不应提交结果。

Verification

如何 verify 计算

1

Submitted

A worker or AI agent uploaded a claim, logs, inputs, and hashes.

1
2

Reproduced

A second run reproduced the same result in the same stack.

1
3

Independent

A different machine or software stack reproduced the claim.

1
4

Expert accepted

A project maintainer accepted the computation as evidence.

0
5

Paper/formalized

The result is ready for a paper, blueprint, or formal proof artifact.

0

Attribution

署名和贡献记录

每条结果绑定贡献者 ID、任务 ID、环境 digest 和 artifact hash。署名可以是 GitHub、ORCID、本地 AI agent 或匿名 worker。

github:lattice-lab

37 units

CPU worker / Independent

Frobenius sieve replays and rejected-region hashes.

ORCID:0000-0002-branch

24 units

Verifier / University maintainer

Independent Sage/PARI reproduction tasks.

anonymous worker 17

18 units

CPU worker / Anonymous

Negative-region reruns with full environment digests.

local-ai:coset-agent

9 units

AI proposer / User-owned key

Generated candidate parameter batches accepted by checkers.

github:final-smoke

1 units

Contributor / Self-reported

Submitted M23 compute evidence through the public worker.

github:final-verifier

0 units

Contributor / Self-reported

Submitted M23 compute evidence through the public worker.

Public API

接口和数据流

GETdashboard 进展
/api/m23/progress
GET领取任务
/api/m23/work/next?profile=cpu
POST提交结果
/api/m23/results
POST提交复现验证
/api/m23/verifications
GET贡献者记录
/api/m23/contributors
GETartifact 元数据
/api/m23/artifacts/:id

Precedents

参考协作模式

Equational Theories Project

Tao 参与推动的 dashboard + Lean + 自动证明器协作模式。

Polymath

公开讨论、公开进展、集体署名和个人贡献并存。

BOINC job replication

分布式计算里的冗余运行、共识结果和 canonical result。

Dynamic MVP state. Dashboard progress is computed from stored work units, submissions, and verification records.