Cloud Chen

[Ivan Morgillo @hamen]

Google Drive on Linux, done properly: rclone mount + working thumbnails, no flaky web client. Files show up like a native folder. Here's the setup that actually sticks 👇

https://t.co/go38CkFZaE

[saket @saketme]

Retrofit, OkHttp, and Okio are moving to a new github organization https://t.co/Y1CdbaMkf7

[Andrej Karpathy @karpathy]

Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.

[Nicos Nicolaou @nicolaou_nicos]

🚀 kotlinx.coroutines 1.11.0 is here!

✅ Kotlin 2.2.20 support
✅ New Flow & StateFlow APIs
✅ Improved structured concurrency docs
✅ Fixes for SharedFlow, flowOn, RxJava leaks & R8 GC issues
✅ Better JS/Wasm exception handling
https://t.co/alWCkckJJw

#Kotlin #AndroidDev

[Berryxia.AI @berryxia]

兄弟们！Google 刚刚更新了Code Wiki！

Google 直接把开发者最头疼的“读代码”瓶颈，一键干掉了。

他们刚刚发布的 Code Wiki，让任何一个代码仓库Github都能瞬间拥有一个永远最新、结构化、 智能可聊的 Wiki。

核心三招狠活：

- 自动扫描仓库，代码一改文档就实时更新，再也不用担心文档过时

- Gemini 驱动的智能聊天，直接把整个 Wiki 当知识库，回答问题还能超链接跳到具体代码行

- 自动生成架构图、类图、时序图，点一下就能从解释直达源码

新手第一天就能上手，老代码没人维护也能秒懂，大库小库通吃。

Google 自己都说：“读现有代码是软件开发里最大、最贵的瓶颈之一。”

现在，这个时代终于要结束了。

公共仓库已开放预览：https://t.co/3hCQKlozkl

你觉得 Code Wiki 会不会成为下一代代码理解标配，把“读代码”从痛苦变成享受？

[Chris Banes @chrisbanes]

💠 Haze 2.0-alpha01 is out!

This is a big release, splitting Haze up into layers, with the goal of enabling pluggable visual effects.

Let me know what you think!

🔗: https://t.co/VlPTqecFrx

[Detroit Tigers @tigers]

not a bad place for your first hit

[MLB @MLB]

Save that baseball

Hao-Yu Lee's first MLB hit is a game-tying single!

[看不懂的SOL @DtDt666]

兄弟们，这个牛逼，谷歌憋了个文件检测杀招，内部藏了多年终于开源！

这工具谷歌在内部跑了好些年，Gmail、云盘、安全浏览全靠它兜底，每周要校验上千亿份文件。

如今直接开源上线，名叫Magika，专扒文件的真实身份，不管怎么伪装都别想蒙混过关。

把恶意软件改名叫「简历.pdf」？Magika一眼检查是不是病毒。

把脚本伪装成图片文件？照样逃不过它的检测。

黑客靠篡改文件后缀玩的所有猫腻，它全能识破。

依托1亿份文件训练的AI模型，覆盖200多种文件类型，识别准确率99%，单文件检测仅需5毫秒。

一行命令直接安装：
 pip install magika 

当年守护谷歌亿万用户的同款神器，现在普通人也能用上了。
https://t.co/tT1YGpmIkJ

[Reviewer3 @Reviewer3]

AI is starting to outperform humans in peer review

We benchmarked 145,000+ human and AI review comments across social science (@NatureHumBehav), computer science (@iclr_conf), and life science (@eLife) papers

R3 leads frontier LLMs and humans across disciplines, providing feedback that is consequential to a major claim within a paper ~90% of the time.

Human peer review is no longer the ceiling.

[Nick Fox @thefox]

The new AI-powered Google Finance is going global and shipping to 100+ countries… it’s a total reinvention of Google Finance with AI at the core 🚀🚀🚀

+ AI-powered market research
+ Advanced charting
+ Live earnings calls
+ Expanded real-time data

Try it: https://t.co/DUtHFXM0cH

[原來是擅長破壞民族團結的朋友 @honglong0420]

哇！這個能源資料視覺化網站做得好棒
把台灣各發電廠和電網清楚地呈現出來

[Muhammad Ayan @socialwithaayan]

🚨 BREAKING: Someone just built the exact tool Andrej Karpathy said someone should build.

48 hours after Karpathy posted his LLM Knowledge Bases workflow, this showed up on GitHub.

It's called Graphify. One command. Any folder. Full knowledge graph.

Point it at any folder. Run /graphify inside Claude Code. Walk away.

Here is what comes out the other side:

-> A navigable knowledge graph of everything in that folder
-> An Obsidian vault with backlinked articles
-> A wiki that starts at index. md and maps every concept cluster
-> Plain English Q&A over your entire codebase or research folder

You can ask it things like:

"What calls this function?"
"What connects these two concepts?"
"What are the most important nodes in this project?"

No vector database. No setup. No config files.

The token efficiency number is what got me:

71.5x fewer tokens per query compared to reading raw files.

That is not a small improvement. That is a completely different paradigm for how AI agents reason over large codebases.

What it supports:

-> Code in 13 programming languages
-> PDFs
-> Images via Claude Vision
-> Markdown files

Install in one line:

pip install graphify && graphify install

Then type /graphify in Claude Code and point it at anything.

Karpathy asked. Someone delivered in 48 hours.

That is the pace of 2026.

Open Source. Free.

[Adit Lal @aditlal]

compose-skill v2.0.0 is out

The open-source AI skill for Compose development just went multiplatform.

What's new:
- Compose Multiplatform support — CMP architecture, expect/actual patterns, platform-specific guides (Desktop, iOS, Web/WASM)
- Design-to-Compose workflow — Figma decomposition algorithm, property mapping, design token translation
- Production crash playbook — 6 crash patterns with root cause + fix, defensive patterns like SafeShimmerItem
- 9 animation recipes — shimmer, staggered lists, swipe-to-dismiss, FAB morph, parallax header + more
- CMP source code receipts from compose-multiplatform-core

Works with Claude Code, Codex CLI, Gemini CLI, Cursor, Copilot, Windsurf, and Amazon Q.

https://t.co/lF0npCgbf3

[Android Studio @AndroidStudio]

Gemma 4 is now available in Android Studio!

By running Gemma 4 locally on your machine you get access to AI code-assistance that doesn't require an internet connection or an API key for its core operations — so you’ll never run out of token quota → https://t.co/1XFsPIDP6P

[Maher Safadi @mahersafadii]

After 8 years of Android KTX, it's being killed.

"The KTX libraries were killed because the adoption of Kotlin has been such a resounding success. All extensions have now been merged directly into their respective main library"

https://t.co/aihrn9N4OK

[Charly Wargnier @DataChaz]

🚨 Google just wiped out the entire call center industry.

→ Their new voice API handles sub-second latency voice AI
→ Multi-lingual calls perfectly.
→ 90+ languages.
→ Replaces receptionists with a Python script.

Want to profit?

Do this:
#1 Target salons, clinics or similar
#2 Build an AI agent to handle FAQs & bookings
#3 Charge $1000/mo (Human receptionists cost $3k/mo).

The window closes in 6 months or less. Build it today 😏

[Jorge Castillo @JorgeCastilloPr]

You can use Skills for agent mode in Android studio now.

Link in comments

[Dy 🌸🌈 @evowizz]

I looked into "Remote Compose". I've been seeing it for quite some time. But I got surprised by the "server-driven UI" bit, so I got curious and looked into it.

TL:DR: Remote Compose is next-gen Remote Views on Android :)

First of all, the name: The AndroidX library lives under `androidx.compose.remote`. But Android's framework *also* has its own copy of the Remote Compose player, baked into the system. They share the same protocol. It essentially makes the AndroidX library "AndroidX Compose Remote Compose"... but we'll stick to "AndroidX Remote Compose" :) This will make sense in a minute.

To make it short, Remote Compose is *not* made for server-driven UI. "Remote" here refers to remote surfaces: Widgets, Wear Tiles, etc. This is exactly what the official description says: "Remote Compose is a framework to create UI for remote surfaces."

Since API 35, `RemoteViews` has a new constructor that takes `DrawInstructions`. The framework also ships its own Remote Compose player. When a widget host (like Launcher) receives `DrawInstructions`, it creates a `RemoteComposePlayer` and feeds it the bytes. No XML inflation, no View tree.

`DrawInstructions` contain a series of bytes (operations) that tell the player what to draw. The player handles layout, drawing, touch, animations, all via Canvas. Once the bytes are decoded they form a "Document," which essentially bundles everything needed for the Player.

But what if a writer uses operations the player doesn't support? Operations are essentially versioned. The document declares its version, and the player checks if it can handle it. If not it rejects the document entirely.

Now the question is: Why does this matter? Until now, Glance translated Compose-like code into `RemoteViews`, which are View-based and limited. With Remote Compose, you skip that translation. You're directly telling the player what to draw: shapes, text, animations, touch handlers, all in a compact binary format.

AndroidX libraries like Glance Widgets and Wear Tiles can now either emit traditional `RemoteViews` (the old way) or emit `DrawInstructions` with Remote Compose (the new way) when supported by the host!

One question remains: if Android has its own player in the framework, why is there an AndroidX Remote Compose player if not for server-driven UI? That would be a fair question to be honest.

Non-system apps can't use `https://t.co/yAZ6RHqB1e.internal` classes. So third-party launchers need the AndroidX player to render Remote Compose widgets. It also means they can potentially support newer operations than whatever shipped with the device's Android version.

I don't believe the purpose of this was for server-driven UI. I believe server-driven UI is something that is possible but wasn't the primary intention.

By the way, there's a good chance we're going to see Remote Compose previews in Android Studio at some point. But this is just a guess.

[Jorge Castillo @JorgeCastilloPr]

AndroidX quietly dropped Remote Compose 👀

It’s the next level of server driven UI. It sends actual native UI drawing operations from server to client.

You update or even ship new UI from server and render it natively on Android. Powerful.