annado do

[Anthropic @AnthropicAI]

This Executive Order is an important step in strengthening America’s leadership in AI.

We look forward to collaborating with the White House to support its implementation.

https://t.co/ZwDimPrp3t

[Microsoft AI @MicrosoftAI]

Seven new models launching at Build: let’s go!
Reasoning. Code. Image. Transcribe. Voice.

Built from scratch on a clean data lineage, designed for efficiency, working seamlessly as a family of models

Thread 🧵
#MSBuild

[VPN测试王｜香港免费VPN推荐 @VPNTestor]

AI 被封号最常见的 3 个原因（VPN 笔记 2026 清单）：

1️⃣ 共享机场 IP 段被滥用｜救回率 仅 20%
几十个账号挂一个出口，OpenAI 按 C 段封·你被连坐

2️⃣ Cloudflare WARP / 主流 VPN IP｜救回率 几乎 0
Express / Nord / WARP 的 IP 段全球已知·秒封

3️⃣ 短时频繁切 IP 国家｜救回率 30%
一会美区一会日区·风控直接亮红

数据中心 IP 黑名单（这些机场全踩雷）：
❌ AWS / GCP / Azure
❌ Vultr / DigitalOcean
❌ 搬瓦工 / RackNerd
❌ 香港节点（OpenAI 明确封锁）

真能稳跑 AI 工具的只剩 2 类 IP：

✅ 美/日等 原生住�� IP（家庭宽带级·独享）
✅ 美/日等 云服务器固定 IP（SSH 上去跑 Claude Code）

普通机场 + Netflix 思路买的节点，
拿去登 ChatGPT 基本就是「免费体验封号服务」。

🤔 你被 AI 平台封过号吗？
评论区报一下，我下周做「AI 工具 IP 红黑榜」👇

[Fuli Luo @_LuoFuli]

Inference Optimizations Behind the MiMo-V2.5 Series API Price Reductions

Read the full technical blog: https://t.co/B5tp4tdnim

The V2.5 model family, including MiMo-V2.5 and MiMo-V2.5-Pro, is built on a Hybrid Sliding Window Attention (Hybrid SWA) architecture, which compresses KVCache storage to roughly 1/7 that of Full Attention. However, architectural advantages rarely translate directly into measurable gains in production serving. To realize these gains, we redesigned KVCache management, tiered caching, and the prefix-cache tree; addressed key challenges in SWA KVCache handling; and optimized scheduling as well as the Prefill/Decode pipeline.

Validated on real production traffic, these optimizations have increased effective KVCache capacity by nearly 5x, with server-side cache hit rates averaging 93%–95% across mainstream harness frameworks. Together with MoE configuration tuning and multimodal inference optimizations, they enable more efficient long-context inference and form part of what makes the recent API price cuts possible.

[Young @Young_AGI]

Kimi Code---kimi-datasource plugin!

It connects professional external data sources, letting you query and analyze stock prices, economic indicators, and more — all in natural language！

From data retrieval → code analysis → full report generation, now a complete closed-loop workflow.🚀

[Lonely @Lonely__MH]

额，发现了一个 对比 Codex 短信接码平台 价格和库存的网站。

独乐乐不如众乐乐

拿去Happy吧🥳

传送门👉🏻：https://t.co/YvjNoc0ZWQ

[Geek @geekbb]

Kimi K2.6 在 Openrouter 上免费🥳

[annado do @doannado]

hi everyone

[Google @Google]

Gemini 3 has arrived! 🚀
Here are three things to know about our most intelligent mode

[GitHub @github]

Custom agents are available in Copilot CLI, making it easier for you to specialize ✨

https://t.co/BgQz4AbOzR

[AshutoshShrivastava @ai_for_success]

🚨 Huge if true...
Apple is reportedly paying Google to build a custom Gemini-based model that runs on its Private Cloud Compute servers to power Siri.
Apple plans to roll out the revamped Siri around March next year, possibly with an AI-powered web search feature.

[DogeDesigner @cb_doge]

BREAKING: Sam Altman responds to Elon Musk.

[Poonam Soni @CodeByPoonam]

🚨 BREAKING: Anthropic has developed a method to test whether AI models can understand their own thoughts.

Using a new technique called concept injection, researchers inserted specific neural patterns into Claude models, and asked if they could detect them.

Claude Opus 4.1 was able to identify injected patterns like "all caps" or "loudness" about 20% of the time, indicating early signs of AI introspection.

[Cline @cline]

A model without a knowledge cutoff.

Perplexity Sonar Pro is now available in Cline.

[Chubby♨️ @kimmonismus]

650 million active users who use Gemini weekly. The gap to OpenAI seems small, with currently 800 million active users. Why is that?

Google has a strength that OpenAI does not have: distribution. Google is in a strong position, and not least the Android operating system and its own Pixel phones are bringing Gemini into the hands of millions of users – once again: distribution. That is the key to success.

You have to hand it to Google – they thought very far ahead early on, and that is paying off.

[Qoder @qoder_ai_ide]

Qoder 0.2.10 is here 🎉

- NES-powered Tab & Tab to Jump just got smarter.
- Run parallel Quest tasks in isolated Git worktrees.
- and more.

Here's what’s new 👇

[Cognition @cognition]

Today we’re releasing SWE-1.5, our fast agent model.

It achieves near-SOTA coding performance while setting a new standard for speed. Now available in @windsurf.

[Cursor @cursor_ai]

Read about everything that's new in 2.0. More tomorrow!

https://t.co/LJ9SDfSqJ2

[GitHub @github]

Welcome to your Agent HQ 📍Orchestrate any agent, any time, anywhere.

Coding agents from @claudeai, @OpenAI, @cognition, @julesagent, @xai and more will become available in GitHub as part of your paid Copilot subscription.
https://t.co/3d8a2Cafqq

[AMD @AMD]

America’s AI Stack, powered by AMD.

Together with @ENERGY, @ORNL, @HPE and @OracleCloud, we’re proud to announce that AMD is helping advance science at scale. With AMD Instinct GPUs and EPYC CPUs at the core of the nation’s newest supercomputer and AI Factory, Discovery and Lux, the U.S. is poised to continue to lead AI innovation.