Neko Itosu

Curious mind, passionate about research and how AI really works
AI / ML / Agents｜正在进一步理解智能系统如何学习、推理与认识世界
中文 / English / 日本語
二次元・数学・绘画・音乐・钢琴
† vibe coding 重度依赖 †

[Neko Itosu @WHMaoXian]

@leo114119 免费会员聊天能用已经可以了

[Neko Itosu @WHMaoXian]

Good morning Argon

[Neko Itosu @WHMaoXian]

@hosting997 Reverse engineering can involve cracking software or bypassing protections, and that could get your account flagged. Not worth risking your main account over it.

[Neko Itosu @WHMaoXian]

REA最近太火了，别被名字和营销带偏了，有两点需要注意

1:千万千万千万别让官号codex和claude去搞。

2:它根本不是一键逆向任何软件。
本质上只是让 AI 调用现有逆向工具，自动分析部分功能，根本复刻不了大型商业软件。

[Neko Itosu @WHMaoXian]

@LucisObscura520 @portertech 这样吗，我看这种bot出来之前就有了，我还一直以为他们是用浏览器自动化操作 X，再接个模型API回复的

[Xudong Han @Xudong07452910]

现在顶尖大学里教 Agent，已经开始一路教到 GPU 了。

非常推荐 Harvard 的 CS2680 Modern AI Systems 。课程从 Claude Code、Agent Loop、Eval 和 Token 成本开始，一路往下做到模型部署、Routing、KV Cache、量化和 speculative decoding。

作业设计也很有巧思，5 个作业都围着同一个 Agent 系统往下拆，先看它怎么工作，再自己搭 Harness，最后一路优化到推理层。

感觉它更像是在回答一个问题：

一个现代 AI Agent，到底是怎么从应用层一路跑到底层系统的？

做 Agent、Harness、AI Infra 的可以收藏。

https://t.co/zUUJPY4lyA

[Neko Itosu @WHMaoXian]

@portertech 那些ai评论大部分是调用api的运营的，平台允许这样。这个dot能做到一样的效果，而且还是Astra驱动还免费，只不过平台规则上是不允许不调用api的自动化运营

[Neko Itosu @WHMaoXian]

@Vincent_AINotes 估计是重置太多，都在拉满噔扛不住

[Neko Itosu @WHMaoXian]

@jngjng383250 还是注意一点比较好

[Neko Itosu @WHMaoXian]

@ajs6888 应该会，目前还没有这么强

[Neko Itosu @WHMaoXian]

Artificial Analysis 推出了一个新基准：AA-3D

主要评估AI Agent 自主构建可交互的 3D 场景的能力。

目前前四名:
Sonnet 5.5：1194 Elo
Opus 5.5：1190 Elo
GPT-6 Astra：1165 Elo
Fable 5.1：1131 Elo

四款模型在统计上没有显著拉开差距。

不过能构建 3D 世界，不代表模型真正掌握了物理世界的运行规律。
这里面还有大量编程和工具调用能力的贡献。

未来如果再结合世界模型与实时生成技术，AI 驱动的虚拟世界会非常值得期待。

[Artificial Analysis @ArtificialAnlys]

We have measured intelligence. Now it's time to see it.

We are launching AA-3D, our new benchmark for how well AI models build interactive 3D scenes. Each model works as an agent in an offline sandbox and turns a written brief into a three.js web page. A human panel of judges compares pairs of scenes built from the same brief without knowing which model made which.

AA-3D details:

➤ 55 written briefs across seven domains, including landmarks, machines, living bodies, games, and space. Each brief says what the scene must show and how people should interact with it, and leaves the rest to the model

➤ Each model gets up to three hours and 250 turns. Its sandbox has Blender, Rapier physics, Python, a headless browser to test its own work, and a CC0 library of materials, 3D models, and characters

➤ Panelists spend at least 30 seconds with each scene before they vote. One comparison in every four is a validation check

➤ Over 10,000 counted votes from over 1,000 panelists rank 21 models by Elo, with 95% confidence intervals

Key findings:

➤ Four models are statistically tied at the top: Claude Sonnet 5.5 (1194), Claude Opus 5.5 (1190), GPT-6 Astra (1165), and Claude Fable 5.1 (1131). The 95% confidence intervals of the other three overlap Claude Sonnet 5.5's

➤ AA-3D Elo rises with intelligence: each point on the Artificial Analysis Intelligence Index is worth about 18 Elo (Pearson 0.92)

➤ The best-placed open weights models, Kimi K3 (982) and MiMo-V2.6-Pro (977), rank 11th and 12th, more than 200 points behind first place

➤ A scene costs $0.20 to $32.08, depending on the model. The cheapest, GPT-6 Luna, ranks eighth

More results, more scenes, and how AA-3D works follow in the thread below.

[Neko Itosu @WHMaoXian]

@xioxioyng539598 @798145343h @ai_funss ៷>ᴗ<៷好

[Neko Itosu @WHMaoXian]

@thsottiaux @poteto hhhhh

[Neko Itosu @WHMaoXian]

@zhangyifk 啊？😮

[Neko Itosu @WHMaoXian]

省流总结:
Cursor pro用户别切换，除非你编码主要用grok（应该没有吧）

X 蓝v用户，已有旧订阅的尤其是祖传Premium+别动，保持原样。旧套餐能留着，切了就回不去了。新用户可以订阅plus或Premium。

Grok订阅用户同理，旧订阅有折扣的先别动。没折扣和新用户可以买。

[SpaceX Super Intelligence @SpaceXSI]

We've unified plans across Grok, Cursor, and X Premium.

Introducing The X Subscription.

[Neko Itosu @WHMaoXian]

@rexriviera @KandaoMaster 可以搜一下，还挺多的

[Neko Itosu @WHMaoXian]

@bei351009016 😭

[Neko Itosu @WHMaoXian]

@Dicklong1999 没错

[arXiv Bangers @arXivBangers]

Multi-Stream Attention Residuals

Multi-Stream Attention Residuals applies multi-head attention over network depth, achieving a 1.58x compute speedup over standard residuals at equal loss in tests on Mixture-of-Experts models up to 28B parameters.

https://t.co/FFelslP28s