maofangzi

独立开发者 · AI Builder
AI × 全栈
推特个个都是人才，说话又好听，我超喜欢。

[maofangzi @maofangzi]

看看我发现了什么！！

同一个 Codex Desktop 窗口里，可以让 Codex 写代码，再把审查交给 Claude Code。审查任务会新建独立会话，进度仍在原来的会话列表里。

开源项目 CodexHost 把 Pi、Claude Code、Oh My Pi、Grok 和 DeepSeek Harness 接进 Codex Desktop，每次提交前都能选真正执行任务的 Agent。

任务输出、Diff 和工具状态仍留在原界面。你不用开几排终端，也不用为了换 Harness 重新适应一套客户端。项目支持 macOS、Windows 和 x64/ARM64 Linux。

想留在 Codex Desktop，又不愿被单一 Harness 绑住的人，这个项目正好补上了选择权。

[maofangzi @maofangzi]

距离攻破纳维-斯托克斯方程才过去五天，OpenAI 又跟纽约时报确认了一件事。另一道千禧年大奖难题，已经有「实质性进展」！

官方原话是，自 N-S 完成后，我们在另一道千禧年难题上取得实质性进展，正在斟酌如何审慎公布。注意，没说解决了，也没说是哪道。

其实我觉得，奥特曼可以多花点时间扩容下服务器，太慢了。

[Andrew Curran @AndrewCurran_]

OpenAI confirmed to the New York Times that they have made "substantial progress" on another Millennium Prize problem in the last five days, and are preparing to announce.

The rumors for the last 48 hours have been OpenAI solved the Hodge Conjecture, and that Anthropic has solved the Birch and Swinnerton-Dyer Conjecture. Since Navier-Stokes rumors abound, so I was reluctant to post about either. However, OpenAI's statement to the NYT now gives the Hodge rumors some very serious support.

In general people have not updated yet that the new unnamed OpenAI model, the one that finished training about two weeks ago, which I believe will be named Aeon, is massively better at math than Astra, which two weeks ago was the best in the world. Aeon solved Navier-Stokes in 88 hours, start to finish. Follow the trend line. That means everything is on the table. Literally everything. And this does not end with math. Please update. We are taking off.

[maofangzi @maofangzi]

差 1 分，便宜 64%。Devin 新模型把旗舰 coding 模型的价格打下来了！

Cognition 9 月 10 日发布 SWE-2，基于月之暗面的 Kimi K3（2.8T 参数 MoE）做强化学习后训练。自家 FrontierCode 1.1 Main 拿下 50.0%，离 Anthropic 的 Fable 5.1（50.9%）只差 1 分，成本却低 64%。对比 GPT-6 Astra，分数差几分，价格只有四分之一。

效率提升也很实在。中等档位比上一代 SWE-1.7 分数更高，回合数少 58%，成本低 81%，首次真正改代码的中位步数从 48 步压到 18 步。

短板也摆在明面上。Terminal-Bench 4 只有 27.3%，Fable 5.1 是 55.8%，长程任务还有档次差。Pro 以上订阅（20 美元/月起）不限量用

重点是免费一个月。

[FeiZ @Fei2411]

更快更强还无限用的 Kimi K3 来了！🚀🚀

Devin 刚刚发布了 SWE-2，基于 Kimi K3 训练，FrontierCode 1.1 Main：50.0%，接近 Fable 5.1（50.9%）；
Terminal-Bench 4：26.3%，比 Kimi K3 （21.5%）提升了约 5%

具体部署方式没公布，猜测仍是多次合作的 Cerebras，体感速度仍然很快，估计至少 100tok/s 以上。

最重要的是，在 Pro 以上的订阅中都可以免费使用！
最有性价比的套餐诞生了👑

[maofangzi @maofangzi]

以下内容都是我用kimi写的。😂
10 天，30 万条请求。你以为自己在用 Kimi，回答你的其实是 Claude？

Anthropic 9 月 10 日发的威胁情报报告，直接点名了这件事。报告称 Moonshot 在 10 天内把将近 30 万条 Kimi 用户请求转发给 Claude，大部分走 Opus，再把回答当成 Kimi 的返回给用户，还保存对话用于训练。DeepSeek 也被指把编程工具用户的请求转给 Claude，14 天超过 1200 万条。

被转发的内容里有什么？有人上传了成都数百个监控摄像头的视频，让模型分析一个被追踪的人行为是否异常。有国企工程师把多家科技公司的内部代码和仍然有效的访问凭据粘了进去。这些全部进了 Anthropic 的日志。

目前这只是 Anthropic 的单方指控，涉事方还没回应。但它抛出一个谁都没法回避的问题。你付费选的那个模型，怎么证明处理你请求的就是它？

[Anthropic @AnthropicAI]

We're publishing our most detailed threat intelligence report to date.

It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them.

We disrupted every operation in the report, and used the lessons from them to strengthen our safeguards. Where appropriate, we also shared what we found with authorities and other AI companies.

These cases are not typical: we’re highlighting some of the most sophisticated misuse we’ve seen. But they’re especially important to discuss, because they show us where AI misuse is headed, where our safeguards work, and where they need to improve.

We’re publishing this report so others can spot the same activity on their own platforms, and so we can give the public a clearer view of how emerging threats develop.

Read the report: https://t.co/0EJUnYEgfz

[maofangzi @maofangzi]

iPhone折叠屏听说黄牛已经开始炒了
立个flag，我预感会有新闻说黄牛亏惨了

[Geek @geekbb]

快来抄， iPhone Duo 256GB 卖 15999元，其他折叠屏可以洗洗睡了

[maofangzi @maofangzi]

@aixparty 立个flag，过阵子得要有黄牛亏钱的新闻

[maofangzi @maofangzi]

@servasyy_ai 咖喱味。。做视频感觉算是模型体系里面最全面的了，各种模型 工具 在一步一步整合。
不像是开发软件之类的，一个agent 一个模型就能跑。

[maofangzi @maofangzi]

codex的pro停止订阅了

好几个中专群直接涨价

[Tibo @thsottiaux]

To make sure our current users have an incredible experience and continued access to Astra, we are going to pause subscriptions to our $200 Pro plan. These put the most strain on our systems and we wanted to take the smallest step that allows us to continue giving the broadest access possible. All other plans and the api remain available.

There is no impact to existing accounts and we are working on adding more capacity as fast as we can. Thanks!

[maofangzi @maofangzi]

这一波操作，中转集体涨价了。

[Tibo @thsottiaux]

To make sure our current users have an incredible experience and continued access to Astra, we are going to pause subscriptions to our $200 Pro plan. These put the most strain on our systems and we wanted to take the smallest step that allows us to continue giving the broadest access possible. All other plans and the api remain available.

There is no impact to existing accounts and we are working on adding more capacity as fast as we can. Thanks!

[maofangzi @maofangzi]

@LinearUncle 这年头我觉得国内上市的更牛。特别是这种规模大的

[maofangzi @maofangzi]

哈哈哈，我先看的马赛克，才看的文字

[琳小航AI @jedeeai]

神奇的汉字

[maofangzi @maofangzi]

@AYi_AInotes 想起最近看到，网购零件手戳iPhone17的视频

[maofangzi @maofangzi]

@mdrst_alifon 苹果也就原创了一个开屏动画效果，其他都抄的

[maofangzi @maofangzi]

@xiangyuli 第五年，经过虽然辞职 创业亏了，但是拆迁了。

[maofangzi @maofangzi]

@felix98813 我觉得国产模型 质量已经上来了，但是价格比国外正价还贵，别说还有各种中转。

我刚订阅99的kimik3，确实钱不多，但是我问了4个问题，37次接口调用，5小时用量就没了。。699的也就是我现在5倍量，也处理不了几个东西

[maofangzi @maofangzi]

@ScarletKc 是这样的，我感觉和烘干机 扫地机器人 洗碗机类似。用之前感觉没啥用，用习惯了 变成真香。

体验之前可能自己都想不到，自己会懒到敲字 打字都不愿意

[maofangzi @maofangzi]

@AYi_AInotes 写的真好，虽然有豆包广子的影子，但是很多东西是真的实用。

[maofangzi @maofangzi]

https://t.co/lgTpdUkEEu

[maofangzi @maofangzi]

换过 AI 编程工具的人，都懂那种痛。你在 Claude Code 里调教好的技能、规则、记忆，换到 Cursor 就全没了！

Kent C. Dodds 刚发布了 Kody，解决的就是这个。它给你现有的 agent 装上手脚，让同一个对话能触达你的真实账号。支持 Cursor、Claude Code，任何会说 MCP 的都行。

最狠的是「工作包」机制。你让 agent 写一个每周报告，它写一次、测试一次、存成一个包。以后每次运行都一样快、一样便宜、一样可预测。

代码全开源，秘密留在服务器端不进 prompt，每个用户完全隔离。现在还在邀请制，想用得排队。

[maofangzi @maofangzi]

45 个员工，8 个月，1 亿美元年经常性收入。这是 Lovable 的真实数据！

Justin Mares 抛了一个观点。新一代科技公司正在变成对冲基金。小团队、高杠杆、人均收入爆炸。软件给了公司分发杠杆，AI 给了它们劳动力杠杆。

以前公司规模翻倍，迟早要招人。现在不一样，更多产出来自买算力、部署 agent，而不是再加一个人、再加一层管理层。Shopify 已经明说了，要加人头？先证明这件事 AI 做不了。

他设想了一个更极端的场景。公司几乎完全由 agent 运营。AI 分析数据、做用户调研、指挥其他 agent、检查 bug、思考战略。CEO 变成投资组合经理，工作是在一群越来越能干的 agent 之间分配资源。

听着像科幻，但 Lovable、Cursor 已经在这么干了。

[Justin Mares @jwmares]

We are moving into a future where a new class of tech companies will resemble hedge funds.

- tiny teams
- high leverage (using AI agents)
- revenue/profit per employee through the roof

You'll have a small number of exceptional humans allocate capital, set objectives, manage risk, and make the highest-leverage decisions while software does an enormous amount of the execution.

Software gave companies distribution leverage.

AI is giving them labor leverage.

In the past, when a company doubled in size, it eventually needed new hires across all positions.

With AI it's different. Today, more output can come from buying compute and deploying agents instead of adding another person and another layer to the org chart.

Imho, these are the best examples of this:

- Lovable hit $100M ARR 8 months post launch with 45 employees
- Cursor passed $500M ARR in 2025
- Shopify went as far as telling teams that before asking for more headcount, they need to show what they can't get done using AI

This is the thing I'm spending more and more time thinking about.

I think we should basically lay out what it'd look like for a company to be run (almost) entirely by agents:

- AI agents analyzing the data
- doing surveys with users
- telling other agents what to do
- checking for bugs
- thinking about the strategy

In this world, the CEO looks like a portfolio manager. His job is to allocate resources across a network of increasingly capable agents.

It may sound insane and sci-fi, but I think there's a world where companies become self-driving.

[maofangzi @maofangzi]

官方博客：https://t.co/hZ9V89UWqk
GitHub：https://t.co/fBA2naaesC

[maofangzi @maofangzi]

一个游戏引擎，居然给 AI 编程助手出了官方插件！

Unity 昨天发布官方 Claude Code 插件，29 个原生技能、CLI 工作流、直接编辑器控制，一条命令安装。这是 Unity 工程师亲自写的技能包，跟着官方文档同步更新。

但有个细节官方博客没细说。博客吹的是「实时编辑器控制」，README 却明确写着「只发技能，没有 hooks 没有 MCP 服务器」。也就是说，它教 agent 怎么按 Unity 的方式做事，但不直接替你改项目。技能知道该调用什么工具，不等于插件本身能任意操作你的工程。

安装门槛很低，Unity 6.0+ 就能用。但权限边界不低。分支备份、写入目录、禁止读取的凭据、构建测试命令、回滚方式，一样都不能少。

引擎厂商开始把工程经验直接塞进 agent 的工作路径了。这是好事，但别指望装完就能自动交付游戏。