陶哲轩博客写数学问题时，通常先把对象定义清楚，再给直觉、反例和证明轮廓。把「Organising mini-polymath2」改写成可阅读的中文笔记，重点是：问题在问什么、已知到哪一步、下一步最容易走偏在哪。原站广告、分享条和导航已去掉。

问题在问什么

This is a continuation of the preceding post regarding the setting up of the mini-polymath2 project. From the discussion, it seems that it is best to start at some fixed time, either on July 8 or July 9, to ensure an orderly opening to the project and also so that I can select which of the six questions would be most suitable for such a project.

I have set up a placeholder wiki page for the project here , though right now it is necessarily quite bare.

已知结果和反例

There wasn’t much feedback on where to host the polymath project, but I am thinking of hosting the main project over on the polymath blog , in order to take advantage of the slightly better threading format there (numbered comments and nested comments in particular, although unfortunately there is no comment editing capability), and also because it would be technically easier to have multiple moderators on that blog. When the project begins, I will also announce it in a brief

The one thing that I need to fix before the project begins is the starting time. I do not know exactly when the Kazakhstan local IMO organising committee will release the full list of questions, but presumably they should be available by, say, noon July 8 at Kazazhstan time (GMT+4), or about 8am July 8 GMT, which would probably be the earliest feasible starting time. Given that I am at GMT-8, the next reasonable starting time would be about eight hours later, at 4pm July 8 GM

证明或构造的主线

I’ve set it so that up to three answers may be selected. I am also open to other suggestions regarding the starting time. I encourage everyone who is interested in participating to select some times for the poll. Note: the times given are with respect to GMT. A list of some representative time zones and their UTC offsets (which are more or less equivalent to GMT offsets, modulo daylight savings issues) can be found here .

With this feedback, I should be able to select a reasonable starting time. The previous mini-polymath was active for about 48 hours (in particular, continuing to produce alternate solutions well after the first one was posted), and I would expect something similar here. I am hoping therefore that provided there is some effort by participants to help bring people up to speed, that one could drop in even several hours after the formal start of the project and still be able to c

阅读时建议盯住的点

先写出对象、假设和失败的例子，再进入证明。没有反例的直觉，很容易把局部技巧当成一般定理。

阅读和落地时建议先做的 5 件事

1. 用自己的语言重写定义和结论，不看原文能不能说清对象是什么。
2. 找一个最小反例或边界情形，确认假设少一条会怎样。
3. 把证明拆成可独立检验的引理，每步只保留一个新想法。
4. 若涉及计算或形式化，先写可复现的小例子，再谈一般情形。
5. 记下尚未解决的缺口：缺估计、缺构造，还是缺正确的范畴。

和智能体、形式化工具怎么接

龙虾PRO做 OpenClaw 落地时，数学笔记最有用的部分往往是「可检验的步骤」：定义、反例、引理边界。智能体适合帮忙展开计算和检索，不适合代替你决定哪条假设能扔。

本文侧重全链路风控方法论。落地时请用自身业务单据做回放验证，不要把示例阈值直接当生产策略。 相关：风控体检 · 方案资源