Garphy

[Laurent @L4UR3N7]

🚨🚨🚨 ALERTE SÉCURITÉ MAXIMALE ! Python est compromis.

🔸 Mettez immédiatement pip à la version 26.2.2 ou plus.
🔸 Dans chaque projet Python (chaque venv) : python -m pip install --upgrade pip
🔸 uv : uv self update (0.13.4 ou plus)
🔸 Mettez aussi à jour les certificats du système (apt / dnf upgrade ca-certificates) et vos images Docker

pip install est compromis.

L'autorité indienne eMudhra (emSign), qui signe les « passeports » prouvant qu'un site est bien celui qu'il prétend être, utilisait des clés secrètes ridiculement faibles : 8 clés, toutes inférieures à 256. L'une valait 4. Autrement dit, n'importe qui peut fabriquer un faux passeport valide pour n'importe quel site.

Le risque pour Python : quelqu'un sur ton réseau (wifi public, FAI, État) se fait passer pour https://t.co/KxkevWMTiB, et ton « pip install » télécharge un paquet piégé sans aucune alerte. Aucune attaque n'a été signalée pour l'instant, mais c'est trivial à faire.

Un paquet piégé peut installer n'importe quoi. Porte dérobée, virus, vol d'informations en tout genre.

La situation est terrifiante !

Rien que dans l'écosystème Python, chaque installation de CPython, chaque venv, uv, tout est exposé.

Et je ne vous parle même pas de npm, pnpm, yarn, Go, Rust.

Ma stratégie de sécurité 2026 : rester chez moi et ne me connecter à Internet que depuis un FAI de confiance.

C'est quand il y a de la mer qu'on reconnaît les vrais marins...

[Anzal Abidi @AnzalAbidi]

Crazy 72 hours in security. Three major incidents dropped back-to-back: eMudhra, PingCAP (TiDB Cloud), and PHP.

Here is the short version of what happened and who got hit.

1. eMudhra (Trusted CA compromise)

eMudhra generated 8 intermediate CA certificates in April 2025 (cross-signed by Sectigo). A bug in their HSM key-wrapping test script created a second set of P-384 keys where the private keys were single-byte numbers under 256. Operators picked the wrong key labels and signed them into real CA certs.

One of the intermediate CA private keys was literally the number 4.
The other seven: 56, 66, 115, 155, 156, 230, 237.

Three weeks later (May 2025), their CRL signatures started failing because the CRL signer was using the real keys instead of the weak ones. Instead of asking why the keys didn't match, someone switched the CRL signer to the weak keys and closed the ticket.

• Browsers are safe: Chrome, Firefox, Apple, and Microsoft already blocked them, and zero rogue certs hit CT logs.
• Servers are not: curl, Python (certifi), Go binaries, and Docker base images don't check CRLs. Update ca-certificates, certifi, and Go (1.27.3 / 1.26.10).
• 10-second check on any server:
curl https://t.co/vhiGLxNhIS
(TLS error = safe. 200 OK = still vulnerable.)

2. PingCAP / TiDB Cloud (10.6 TB breach)

Attackers found GCP and Azure credentials committed inside PingCAP's own repos. They set up a replication feed of TiDB Cloud's production control plane and exported the entire database every 6 hours for 22 straight days. 88 full exports, zero alerts. 200 GB is already out on torrents.

Beyond the 413 private repos and 20.6M customer SQL queries logged from their web SQL Editor, PingCAP engineers were pasting customer secrets into Jira tickets.

Who is affected in the leaked records:

• Live credentials exposed in support tickets:
- Pinterest: AWS keys for their pingcap-logs S3 bucket
- CardX / Silverlake (NPCI-linked env): AWS admin key in a backup log with full access to 258 external S3 buckets
- DeBank: 38 presigned S3 URLs
- An unnamed AI/dev platform: a 314 MB slow-query log attached to a ticket containing 8,027 live customer secrets (an AWS root key, a GitHub admin PAT over 1,701 repos, 7,485 AWS STS tokens, and 192 OpenAI/Anthropic/Gemini keys)

• Tenants with infra metadata, VPC CIDRs, app-server IPs, or contract pricing exposed:
- Tech: Atlassian, Databricks, Airbnb, Baidu, LINE, Nylas, Persona, PatSnap
- E-comm & Travel: https://t.co/C512KFOEq5, Shopee, Flipkart, Mercari, ZOZO, AEON, Delhivery, Klook
- Gaming: Capcom, Square Enix, Konami, Bandai Namco, Cygames, Colopl, DMM
- Finance: PayPay, SoftBank, SBI, SB Payment Service, Fujitsu, TOPPAN, Bank of China, China Life
- Crypto: Bitget (said only public trade logs were on it and they already left), Pionex, DigiPlus, NFTscan, Chainbase

• Developer profiles leaked (from a purchased https://t.co/28tjHqd5Wr table in PingCAP's BigQuery):
- 24,900 personal emails + GitHub/LinkedIn profiles of engineers at Tencent (3,327), Huawei (2,726), ByteDance (2,475), AIG (2,140), Mastercard, Flipkart, and bKash.

If you use TiDB Cloud: rotate your console API keys, DB passwords, and anything you ever pasted into a support ticket.

3. PHP "PHP2Shell" (CVE-2026-109618 & CVE-2026-109619)

Two CVSS 9.2 pre-auth RCEs leaked yesterday ahead of today's 14:00 UTC patch.

Public writeups trace it to a heap Use-After-Free in PHP's native C multipart/form-data session upload-progress handler (ext/session/session.c).

Because C-level multipart parsing runs before line 1 of index.php:
• App auth, Laravel/WordPress middleware, and file_uploads = Off do not stop it.
• Only PHP 8.2–8.5 get official patches today (59% of PHP servers on Shodan are on EOL <= 8.1).
• Until you patch php-fpm: lock php_admin_flag[session.upload_progress.enabled] = off and block unexpected multipart requests at Nginx/Apache/Cloudflare.

[LinXunFeng @xunfenghellolo]

@__Crazy_Kid__ PHP高危RCE漏洞
https://t.co/vZwHvjcXTE

[大漂亮| C Labs @giantcutie666]

这个世界真奇妙

黄奇帆竟然是被之前那笔加密货币洗钱按顺藤抹出来的大瓜😅

[Garphy @Garphy]

@KELMAND1 国标1500N，而且新的团标已经要往上定到2000N，这质疑个什么劲儿？百万好车，还贴着底线做被踩爆，还有啥好说的。

[Garphy @Garphy]

@ayanamist 是吧，还是不能一刀切~

[Garphy @Garphy]

@mybrobot 丹霞山鳌鱼石

[Garphy @Garphy]

GLM5.3得到了Anthropic的认可...

[Nathan Lambert @natolambert]

Sigh.

I wrote this post below, but then I didn't post it because I am tired of the consistent pushback I get from folks with the same safety worldview as Anthropic. I guess that means I should send it. Here goes!

This post is pretty solipsistic and doesn't properly take recent events in cyber risks into it's discussion. It is really hard for me to watch how the US frontier labs like Anthropic don't have the ability to consider other approaches to safety and ways things could play out.

It insinuates that Z ai (Chinese lab, builds GLM series) doesn't really care about safety and is reckless to take their business strategy.

Saying things like "Given this evidence, we think it's likely both state and non-state actors will use models like GLM-5.3 to cause real-world harm." and "This is unlike any other similarly capable AI model, all of which were released with safeguards or through limited access programs." while closed models have been used on more of the documented cyber attacks is just bowing out of the interesting question.

A plausible view is that open model weights and closed model apis (with some safe guards) are both far closer to being easy to mis-use, rather than API models being closer to safe. The trope "Open Dangerous, Closed Safe" may be closer to "Open Unsafe, Closed Unsafe"

Closed models have stronger capabilities and stronger safeguards, but the stronger capabilities part could matter more in net harm if both the safeguards are porous.

At the same time, as the authors do acknowledge (thanks - thats progress!), open models without extreme cyber guardrails are important to rapidly diffuse cyber readiness in the economy -- as programs like project glasswing are not perfect in getting all critical industry onboarded.

There will be more issues like the time when Fable was released and Amazon found a workaround, which allowed them to access the full capabilities of the model served readily at an API.

The blog overall is reasonable in it's narrow line, but it's a very effective tool in a complicated, rapidly evolving media ecosystem to reinforce a certain type of safety thinking.

[Stanley @Stanleysobest]

脑洞真大，高手在民间，

原来鬼吹灯的机关原理是这样的。

之前看鬼吹灯看的很过瘾，

一直没搞懂里面神奇的机关原理，

这个视频一次讲清楚了👍👍👍

[Garphy @Garphy]

@ayanamist 有一个没啥用的火绒的···

[Garphy @Garphy]

@ayanamist 搞不懂，我这主要也就浏览器吃内存，主要是久了虚拟内存占多了会卡，window这内存管理就是跟屎一样

[Garphy @Garphy]

@ayanamist 会卡死，剩余内存不到10G，已提交 53.1/59.5G

[Garphy @Garphy]

@ayanamist 我chrome占用最多了，休眠几次虚拟内存就吃满满的

[Arena Physica @arenaphysica]

Can the US build a DJI? We bought a 2019 Mavic Mini and a 2023 Mini 4 Pro, tore them down in our hardware lab, and put them under a bench microscope to find out.

What we found: https://t.co/2nxcHd8cSF.

[AYi @AYi_AInotes]

有��燃炸，这才是数学最狂暴的力量，全屏静音把这十几秒流动的拓扑轨迹看完，去感受一下什么叫真正的智识震撼！

这段在推特上狂揽 180 万播放的动态数学动画，直接把全世界玩了几十年的魔方，从根源上降维打击得体无完肤！

兄弟们你知道一个普通的三阶魔方，到底藏着多少种绝望的混乱吗？
整整 4300 亿亿种（$$4.3 \times 10^{19}$$ 种状态）！
如果你每秒钟转一种组合，就算从宇宙大爆炸那一刻转到现在，你连全盘状态的百分之一都试不完！
绝大多数人一辈子都在被动当人肉打字机——死记硬背七八套公式、一层一层转，转错一步全盘报废；
但在顶级数学家眼里，这个塑料块从诞生第一天起就不是玩具——
它是一个纯粹的对称置换群，是一张横跨高维空间的巨型「凯莱图（Cayley Graph）」！

看懂这段视频背后令人战栗的数学暴力美学，你的认知会被当场连根拔起：

1️⃣ 空间映射：把玩具变成宇宙级拓扑图：
魔方的每一个物理状态，在图论里就是一个孤立的“节点”；
你的每一次顺时针或逆时针旋转，就是连接两个节点之间的“有向边”！
4300 亿亿种混乱被编织成了一张巨大的超图；
解魔方的本质，根本不是把红色对齐红色、白色对齐白色，
而是在这张巨型网络里，寻找一条从当前节点通往“绝对对称点”的最短路径！
2️⃣ 算术级神迹：“上帝之数是 20”：
这是人类计算机科学史上最浪漫的公理之一：
数学家用了等效上万年的计算时间，跑完了这张图的每一个角落，
最终得出了震撼全人类的结论——这张包含 4300 亿亿个节点的巨网，它的最大直径，居然只有整整 20！
这意味着什么？
无论魔方被世界上最刁钻的人打得多乱、看起来多么病入膏肓，
在上帝视角里，距离完美还原，最多永远只需要 20 步！
3️⃣ 真正的降维解法：群论分层（Kociemba 算法）：
计算机绝不会像普通人一样从底层开始拼十字；
它利用抽象代数，先把整个置换群压缩投影到一个仅有 20 亿状态的“子群”里，
先解决朝向，再解决位置；
别人在平地像无头苍蝇一样乱撞，数学家在高维空间里直接开虫洞！

这段视频最让人热血沸腾的，不仅仅是魔方被攻破，
而是它用无懈可击的几何美感告诉你：
宇宙中看似最复杂的混沌，在足够高维度的规律面前，本质上不过是极其纯净的对称折叠。

生活和商业里的很多绝境也是一模一样：
低维的视角里全是死循环、全是撞墙、全是不可逾越的阻力；
但只要你站在图论的上帝视角往前看，
最黑暗的混乱背后，其实一直静静躺着一条不超过 20 步的最短测地线！

[LinearUncle @LinearUncle]

研究了下这两天很火的 typesafeai 的 jev 模型，没想到用起来这么简单(看图）。

从 API 看，就是提供详细的上下文（state），再给它几个问题和对应选项，让它据此判断选择选项。

已经有人用来炒股了，让jev根据实时提供的股票数据进行买入还是卖出的决定！

目前vercel gateway已经上架了jev API，可��使用。

官方文档地址：
https://t.co/Ed6ge9t1yK

vercel gateway:
https://t.co/uSD0iJS2nS

[Niels Rogge @NielsRogge]

For anyone curious how Jev works, I made a visual explanation using @claudeai :)

This is based on the Qwen2.5-RLCD model which @harshagundal released on @huggingface

The idea is to replace autoregressive LLM generation by a single Transformer decoder (of a pre-trained LLM), which processes the context + JSON schema only once. The keys and values of those tokens are cached.

Next, for each field of the JSON schema, we:
1. pass its field suffix tokens through the Transformer decoder again (reusing the KV-cache)
2. obtain a final hidden state, which we pass through the language modeling head
3. we obtain scores, also called logits, for all tokens in the vocab of the LLM
4. we only look at the scores of the tokens we care about for the given field, and pass those through a softmax to obtain probabilities which sum to 1
5. we take the token with the highest probability.

The benefits of this are that:
1. it's fast (we don't need to generate the JSON schema token by token)
2. it's 100% valid JSON (we don't need to rely on the model to generate a valid schema)

[思维怪怪 @0xLogicrw]

前 OpenAI 研究员 Diogo Almeida 创办的 TypeSafe AI 推出新模型 Jev。它有点像一个能读懂自然语言的超级分类器，不生成文本，只返回选项、分数和概率，专门给软件做判断。

普通大模型需要一个 token 一个 token 往外生成，Jev 则可以并行给出多个结果。TypeSafe 还用���的 RLCD 方法训练它的置信度，让模型在给出判断的同时，告诉程序自己有多大把握。

TypeSafe 用安全告警、Agent 审查、发票处理和客服做了 4 组自动化工作流测试。在这套自建评测里，Jev 最高比 Claude Sonnet 5 快约 193.6 倍，比 Claude Opus 5 便宜约 444.6 倍。

TypeSafe 把 Jev 称为首款「System One Model」（快速决策模型），名字来自卡尼曼《思考，快与慢》里的 System 1。它专门处理分类、打分、路由和下一步决策这类快速判断任务。

[yishan @tspy]

软件工程师 fanhao 走��方硬件路线成功复刻 Microduck

已实机完成官方9种强化学习的策略

对，就是之前那个花了12000 RMB 复刻机器鸭的 fan 哥

虽然是个软件工程师，无奈动手能力超强，顺利完成

走官方硬件路线复刻的同学，可以参考

相关资料已经上传GitHub
https://t.co/kS16pBrOyy